# Welcome to CATAPA Documentation ## CATAPA Documentation This documentation provides the technical and functional specifications for the CATAPA ecosystem. The platform is divided into two primary service areas: ### CATAPA HRIS A centralized Human Resource Information System for organizational management and compliance. * **Personnel Management:** Core employee data, organizational structures, and job histories. * **Payroll & Statutory:** Engine for payroll processing, PPh 21 (tax) calculation, and BPJS compliance. * **Time & Attendance**: Management of leave policies, shift scheduling, and attendance logs. More about CATAPA HRIS here [CATAPA Services](/catapa/catapa-hris/catapa-services) ### Digital Employee An agentic AI application designed to execute office-based tasks within a human workforce. Unlike standard chatbots, which are passive and limited to runtime context, the Digital Employee is designed for proactivity and autonomy. #### Functional Characteristics * Agentic Behavior: Operates as an active participant in workflows, capable of taking initiative rather than only responding to prompts. * Continuous Context: Combines human-like contextual understanding with machine-level recall and 24/7 availability. * Integration: Designed to operate across existing office job functions at scale. More about Digital Employee here [Digital Employee Concept](/catapa/digital-employee-de/overview/digital-employee-overview) ### Jump right in
Cover image
CATAPA HRISHandle Payroll with ease/files/f22Gy3qyHy0pLGPyDL1s/pages/UNUlTPYK8IccbqwhtGCa
Digital EmployeeSomething exciting is coming your way!https://images.unsplash.com/photo-1644088379091-d574269d422f?crop=entropy&cs=srgb&fm=jpg&ixid=M3wxOTcwMjR8MHwxfHNlYXJjaHwxfHxkaWdpdGFsfGVufDB8fHx8MTc2OTcwNjI4NHww&ixlib=rb-4.1.0&q=85/pages/Mwq3NeMPlVYxnrB9DNOu
# About CATAPA CATAPA was founded on 21 April 2017, with its first application focusing on payroll. The platform was created to help companies leave manual processes that often cause problems such as poor documentation, long processing time, and difficulty in collecting and managing data. Over time, CATAPA learned that clients need more than payroll. Because of that, CATAPA continues to grow into an **Integrated HR Solution**, a one-stop platform to support end-to-end human resource operations. # CATAPA Services CATAPA provides several applications and modules that help companies manage their employees more efficiently. Today, CATAPA includes the following key modules: * **Organization**\ Manage company structure, job levels, and organization hierarchy. * **Employee**\ Store and manage all employee information in one place. * **BPJS Ketenagakerjaan**\ Automatically calculate BPJS Ketenagakerjaan contributions for both company and employee. The system also provides government-required reports. * **BPJS Kesehatan**\ Calculate BPJS Kesehatan contributions and provide necessary reports. * **Payroll**\ Manage payroll for active and resigned employees based on company rules. CATAPA also provides payslips and all required payroll reports. * **PPh 21**\ Calculate and manage income tax according to company policies. CATAPA provides 1721-I (monthly and yearly), 1721-A1 PDFs, and Coretax upload files (xml: BPA1, BPMP, BP21, BP26). * **Termination**\ Manage data and processes for employees who leave the company. # CATAPA Uptime SLA CATAPA targets **99% uptime per month**.\ Exclusions include: * Scheduled maintenance on Saturday and Sunday, 00:00 – 06:00 WIB. * Force majeure events such as: * Cloud platform or network provider disruptions (AWS, GCP, etc.) * Local, national, or international network issues * Actions by third parties that intentionally cause service disruption (e.g., DDoS) # Compliance with Government Regulations All CATAPA modules are designed to follow Indonesian government regulations, including working hour standards, BPJS calculations, and PPh 21 calculations. CATAPA’s team also monitors regulatory changes regularly to ensure the application remains up to date. # Customization CATAPA provides complete HR features by default.\ If your company requires a special or unique feature, you may contact **** to discuss customization needs. # How to Start Using CATAPA To begin using CATAPA: 1. Send an email to . 2. Include the information below: 1. Company Name: 2. PIC (Person in Charge): 3. Authority: 4. Phone Number: 5. Email: 6. Line of Business / Industry: 7. Company Address: 8. Number of Employees: 9. Required Modules: 10. How did you hear about CATAPA?: 11. Previous Payroll System: 3. The CATAPA team will follow up on the incoming email for the next steps. 4. CATAPA supports data migration, making it easier for companies moving from another HR or payroll system, and is ready to assist throughout the migration process. 5. Companies interested in learning more can request a demo session based on the CATAPA team’s availability. # FAQ and User Guide Need guidance or support? The CATAPA Help Center offers step-by-step articles for every feature.
1. [All categories and modules](https://help.catapa.com/categories) 2. [Troubleshooting and error message explanations](https://help.catapa.com/search?query=pesan%20error)
# CATAPA Chat # Overview **Claudia** serves as the central workspace where users interact with CATAPA’s AI-powered HR assistant. Designed with clarity and ease of use in mind, Claudia enables HR teams, managers, and employees to access information, perform HR tasks, analyze data, and navigate regulatory guidance, all within a consistent and intuitive interface. Whether accessed through the web platform or via WhatsApp, Claudia offers a seamless and responsive experience across different use cases and user roles. More than a simple chatbot, Claudia functions as a **Digital Employee**, equipped with specialized assistant modes for regulations, HR analytics, employee self-service, and system guidance. These capabilities allow organizations to automate routine tasks, accelerate decision-making, and deliver accurate, context-aware insights directly within day-to-day workflows. Behind the scenes, Claudia integrates with CATAPA’s HRIS ecosystem, ensuring secure data access, role-based permissions, and reliable performance. This section provides an overview of how each interface component supports efficient communication, task execution, and data-driven HR operation, forming the front-facing layer of CATAPA’s AI-enabled HR transformation. # Feature Overview **Last update**: March 12, 2026 Claudia is an enterprise-grade **Agentic AI Digital Employee Platform** built specifically for HR operations, designed with a ✨**privacy-first foundation** and seamless integration into the CATAPA ecosystem. It empowers HR teams, managers, and employees through specialized assistant modes, advanced data analysis, and secure AI-driven automation available across web and WhatsApp. Claudia brings together the power of the **CATAPA Agentic AI System**, enabling capabilities such as the **Regulatory Assistant (KG-Powered)**, **HR Assistant (KG-Powered)**, **Employee Assistant**, **Help Center Assistant**, and **Advanced Reports**, all while supporting Bahasa Indonesia and strong privacy safeguards. With the upcoming **Digital Employee** experience, Claudia moves beyond traditional chatbots to deliver proactive, AI-native workforce augmentation.
With ✨a growing suite of intelligent features, this documentation organizes Claudia’s capabilities into clear categories for easy navigation. Each item includes a brief description for quick reference, while dedicated sections offer detailed explanations and step-by-step guidance whenever you need deeper understanding. ## Highlights :book: ### Accessing Claudia :rocket: * [**Login with CATAPA**](/catapa/catapa-hris/catapa-chat/getting-started) **-** Sign in securely using your existing CATAPA account to access Claudia on web or mobile. * [**Connecting CATAPA Account to Claudia**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/accessing-claudia/connecting-catapa-account-to-claudia) **-** Link your CATAPA account so the HR Assistant and Employee Assistant can access and analyze your company’s HR data. * [**Accessing Claudia via WhatsApp**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/accessing-claudia/accessing-claudia-via-whatsapp) **-** Chat with Claudia directly through WhatsApp using secure company ID, username, and OTP verification. * [**Sign Out**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/accessing-claudia/sign-out) **-** Safely end your session when you no longer need to use Claudia. ### Navigation and Chat List :file\_folder: * [**Interface Navigation**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/interface-navigation) **-** View the home page, sample questions, and the sidebar that stores all your previous conversations. * [**Search Chat** ](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/search-chat)**-** Quickly locate past conversations by typing keywords in the search bar. * [**Select Assistant**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/select-assistant) **-** Switch between the four assistant modes depending on your needs. * [**Tips for Using Claudia Effectively**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/tips-for-using-claudia-effectively) **-** Follow best practices to get the most accurate and helpful responses from Claudia. * [**Security and Privacy**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/faq/security-and-privacy) **-** Learn how Claudia protects your data through encryption, masking, and role-based access. * [**Clear Chats**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/clear-chats) **-** Remove all existing chat histories when you need a fresh start or want to maintain privacy. * [**Sources**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/sources) **-** View the official regulatory references and documents included in Claudia’s answers. ### Submitting and Approving Requests :page\_facing\_up: * [**Submissions Through Claudia Virtual Assistant**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/select-assistant/employee-assistant/submitting-and-approving-requests/submissions-through-claudia-virtual-assistant) **-** Submit HR-related requests such as leave, attendance corrections, absences, cancellations, and overtime directly through chat. * [**Approvals Through Claudia Virtual Assistant**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/navigation-and-chat-list/select-assistant/employee-assistant/submitting-and-approving-requests/approvals-through-claudia-virtual-assistant) **-** Managers can view all pending requests and approve them individually or in bulk. ### Settings and Report :gear: * [**Change Theme**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/settings-and-report/change-theme) **-** Adjust Claudia’s visual theme for a more comfortable viewing experience. * [**Change Language**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/settings-and-report/change-language) **-** Switch Claudia’s interface language based on user preference. * [**Report an Issue**](/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/settings-and-report/report-an-issue) **-** Submit technical feedback or report problems directly to the CATAPA support team. Claudia inherits features from GLChat, an **enterprise-grade Agentic AI chatbot platform.** The complete list of features from GLChat can be accessed here: # Getting Started Claudia is a virtual HR assistant designed to support your day-to-day activities. It helps HR teams, employees, and managers access information quickly, make informed decisions, and perform various self-service tasks with ease. You can access Claudia through the launcher icon available in the CATAPA application or through the Employee Self-Service (ESS) platform. ### How to Access Claudia If you are already using CATAPA, you can access Claudia through the launcher icon displayed on your company’s CATAPA dashboard. Alternatively, Claudia can be accessed directly via the link [claudia.catapa.com](https://claudia.catapa.com). ### Logging in to Claudia #### **1. Select "Login with CATAPA"** On the Claudia login page, click **"Login with CATAPA"** to begin the authentication process.

Figure 1 - Claudia Login Page

#### 2. Enter Your Company Name and Credentials Next, enter your company name, followed by your CATAPA username and password, exactly as you would when logging into CATAPA.

Figure 2 - CATAPA Login Page

#### 3. Access the Claudia Dashboard Once the login process is successful, you will be redirected to Claudia’s main interface.

Figure 3 - Claudia Home Page

# Features and How-To Guide *Last Update: Mar 2026* Welcome to the Claudia Virtual Assistant User Guide.\ This guide provides detailed, step-by-step instructions for using Claudia so employees, supervisors, and HR teams can maximize efficiency in managing HR tasks. While the Overview section explains Claudia’s core capabilities, this guide focuses on **how to operate each feature in real workflows.** You’ll find clear instructions, screenshots, and tips for both everyday tasks and advanced use cases. To improve navigation, features are grouped into the following categories: ## Accessing Claudia
NoFeatureDescription
1Connecting CATAPA Account to ClaudiaThis section explains how to link your CATAPA account to Claudia so the HR Assistant and Employee Assistant can access and analyze your company’s HR data.
2Accessing Claudia via WhatsappThis section guides you through connecting your company account and interacting with Claudia directly through WhatsApp.
## Navigation and Chat List
NoFeatureDescription
1Interface NavigationThis section introduces the main elements of Claudia’s interface, including the home page and the navigation area, which allows you to access previous conversations or begin a new chat easily.
2Search ChatThis section explains how to quickly find past conversations using the built-in search feature, allowing you to locate specific chats by entering keywords in the conversation history page.
3Select AssistantThis section explains Claudia’s four assistant modes and how to switch between them. Each mode offers specialized capabilities, allowing users to choose the most suitable assistant based on their task or information needs.
4Tips for Using Claudia EffectivelyThis section introduces Claudia’s four assistant modes and shows how to switch between them based on your needs.
5Security and PrivacyThis section outlines how Claudia safeguards user data through role-based access, encryption, and automatic masking to ensure secure and private interactions.
## Submitting and Approving Requests
NoFeatureDescription
1SubmissionsThis section explains how employees can submit HR requests directly through Claudia, including examples of how each submission is processed.
2ApprovalsThis section explains how managers can review and approve pending employee submissions in Claudia, either all at once or selectively, using simple commands.
## FAQ
NoFeatureDescription
1General InformationThis section provides an overview of what Claudia is, the assistant modes available, and the scope of topics Claudia can support within HR and regulatory workflows.
2Getting StartedThis section explains how to begin interacting with Claudia, including starting new conversations and switching between assistant modes to match your needs.
3Accuracy, Reliability, and ReferencesThis section explains how Claudia handles information accuracy, how users can verify answers, where references come from, and what to do when Claudia cannot provide a response.
4Usage Limits and System BehaviorThis section outlines the usage limits applied in Claudia and explains how the system notifies users when they are approaching their allowed quota of interactions.
5Security and PrivacyThis section explains how Claudia protects sensitive data through encryption, masking, and role-based access to ensure secure and authorized data handling.
6Feedback and SupportThis section explains how users can provide feedback to help improve Claudia and how to reach CATAPA Support for further assistance.
# Organization Admin Dashboard The **Admin Dashboard** acts as the operational backbone of **Claudia**, providing administrators with complete visibility and control over the platform’s configuration and usage. Each module within the dashboard is designed to **simplify setup**, **streamline management**, and **accelerate configuration workflows**. For complete guide for the admin dashboard, you can visit the GitBook here: # Accessing Claudia # Connecting CATAPA Account to Claudia To enable the HR Assistant and Employee Assistant to access your company’s data in CATAPA, you must first connect your CATAPA account to Claudia. Follow the steps below: #### 1. Open the Settings Menu Click your account name located at the bottom-left corner of the screen, then select **Settings**.

Figure 4 - Settings Menu

#### 2. Select the Connector Option In the Settings page, choose the Connector menu.

Figure 5 - Settings Page > Connector

#### 3. Connect to CATAPA In the CATAPA section, click the Connect button to begin the integration process.

Figure 6 - Selecting the CATAPA Connector

#### 4. Log In to Your CATAPA Account You will be redirected to the CATAPA login page. Enter your CATAPA username and password to authorize the connection.

Figure 7 - CATAPA Login Page

#### 5. Successful Connection Once the connection is successful, the Connect button will change to Disconnect, indicating that Claudia is now linked to your CATAPA account.

Figure 8 - Settings Page with CATAPA Connector Successfully Connected

Congratulations! You can now use Claudia to analyze and access your company’s HR data seamlessly. # Accessing Claudia via Whatsapp Interacting with Claudia is now even more convenient, as you can chat with her directly through WhatsApp. Follow the steps below to start using Claudia on WhatsApp: #### 1. Greet Claudia for the First Time If this is your first interaction with Claudia via WhatsApp, she will greet you and provide instructions to enter your company name once you initiate the conversation.
**Note:** Only companies that are subscribed to Claudia can use the WhatsApp feature. #### 2. Enter Your Company ID Registered in CATAPA When Claudia asks for your company ID, simply reply with your company’s registered ID.
#### 3. Enter Your Company Username Claudia will then request your CATAPA username. Respond by typing your username.
#### 4. Enter the OTP Sent to Your Registered Email An OTP (One-Time Password) will be emailed to the address associated with your Claudia account. Enter the OTP in the chat to continue.
#### 5. View Available Claudia Applications (Assistant Modes) You can view all available assistant modes by typing the command **/listapp.**\ Claudia will display a list of applications along with their corresponding IDs.
#### 6. Select the Application You Want to Use To switch to a specific assistant mode, use the command: **/setapp \[application\_id].** Choose the appropriate ID from the list provided in the previous step.
If the application you are looking for does not appear, please contact your company’s HR team to adjust your access settings. #### 7. Start Chatting with Claudia via Whatsapp Once you have selected your preferred assistant mode, you can begin asking questions and interacting with Claudia just as you would in the web version.
**Important Note:** A single phone number can only be associated with one Claudia account. # Sign Out To sign out, simply click your profile menu in the bottom-left corner and select the **Sign Out** option.
# Navigation and Chat List # Task Scheduler ## Schedule a Task You can schedule a task in GLChat. The tasks can vary, such as: * Send a prompt to LLM * Send instructions to AI Agent * Perform a search (Web/Connector) * Conduct deep research To create or schedule a task, type a prompt in the chat input box. After selecting one of the tasks above, use the + Button and choose the schedule message.
**Note**: Currently, you cannot schedule a task for a prompt that contains a file attachment. There are several types of schedules: * Once * Interval (minutes) * Interval (hours) * Daily * Weekly * Weekdays * Monthly * Yearly Once you are firm with your choice, hit the Save button. A toast message informing you that the task is already created will appear, along with a link to open the Task List.
Once the task meets the scheduled time, it will be executed as a new conversation.
## View Task List To view task lists, open the toast notification that appears after scheduling a task, or simply go to the "Schedule Task" menu in the profile section at the bottom left. This menu includes two tabs: **Scheduled** * Displays all created tasks and their execution status. * Options available: Set task to inactive, edit, or delete a task.
**Completed** * This section shows all tasks that have already been executed. * You can also click the completed task to go to the conversation containing the scheduled task. ## Edit Task To edit a task, select "Edit" from the Task List. This will redirect you to a page where you can modify the task or its prompt.
Note, to change the schedule, simply click the link "Change schedule here".
Once you are done with the edit task, click the check button. A toast message will appear informing you that the adjustment was successful. You can check the task via the Task List menu. # Guest Mode Guest Mode allows users to access application without logging in, enabling quick, temporary interactions while maintaining data privacy and session isolation Guest Mode is designed for users who need instant accesswithout requiring an account or authentication. It’s ideal for demos, trials, or public deployments where ease of access is critical. In this mode, user sessions are temporary. Once the browser is closed or the session expires, all conversation history and uploaded data are automatically cleared to protect privacy. Currently only limited features are available in Guest Mode other than conversational interaction such as: anonymize personal data and question library # Interface Navigation ### 1. Home Page

Figure 1 - Claudia Home Page View

When you access Claudia, you will be welcomed by the home page, which displays several sample questions (as shown in Figure 1). These examples can help you begin interacting with Claudia more easily. Some of the available sample questions include: 1. “How is the TER tax calculated for an unmarried employee earning 12 million?” 2. “How much year-end tax should be paid if the monthly income is 10 million?” 3. “Why does year-end tax increase even when the TER rules have already been applied?” 4. “What are the rights and obligations of a tax withholder according to PMK 168?” These examples serve as a reference to help you formulate your own queries when communicating with Claudia. ### 2. Navigation Area The left side of the screen serves as the navigation area. This section displays your conversation history with Claudia. You can select any previous conversation to continue interacting from where you left off. If this is your first time using Claudia, you can start a new conversation by clicking the “New Chat” button. Each conversation session will be saved, enabling you to revisit information or responses that Claudia has provided in the past.

Figure 2 - Navigation Area

# Search Chat If you need to locate a conversation from earlier interactions, you can use the search feature available on the conversation history page. Simply click the magnifying glass icon in the navigation area (as shown in Figure 6). Once the history page appears, type a keyword related to the conversation title into the search box. Claudia will then display a list of conversation titles that closely match your keyword.

Figure 3 - Conversation History Page View

# Select Project Claudia offers various assistant modes, referred to as Projects, each designed to help you handle different topics and operational needs. Every assistant has a unique role, specific functions, and its own conversation history, allowing you to interact more efficiently based on the type of support you require. The following are the four available Claudia Projects: 1. Regulation Assistant 2. HR Assistant 3. Employee Assistant 4. Help Center Assistant To change assistant modes: 1. Hover your cursor over the navigation panel and select the Project box. 2. A list of available assistant modes will appear. 3. Choose the assistant mode that best matches your current needs.

Figure 4 - Claudia Project Selection Interface

# Regulatory Assistant In this mode, Claudia acts as a regulatory advisor, helping you understand the latest government regulations and laws in Indonesia, particularly those related to HR. Equipped with comprehensive knowledge in taxation, labor regulations, and payroll, Claudia supports you in maintaining compliance with applicable rules. **How the Regulation Assistant Works** Using this assistant is simple. You only need to describe the issue or question you want to clarify, and Claudia will provide recommendations based on the relevant regulations. Each explanation includes references to official regulatory documents, ensuring that the information remains accurate and reliable. Using this assistant is simple. You only need to describe the issue or question you want to clarify, and Claudia will provide recommendations based on the relevant regulations. Each explanation includes references to official regulatory documents, ensuring that the information remains accurate and reliable. **Features** **Regulatory Questions** Claudia allows you to obtain clear and concise answers to various regulatory questions with ease. Furthermore, Claudia does not simply quote government regulations; it also incorporates reasoning capabilities, enabling it to understand context and provide logical explanations behind each answer. This ensures that you not only receive the correct solution but also gain a clearer understanding of the underlying regulatory principles. **Trusted Sources** Every response includes references to official documents, such as laws or government regulations to ensure information accuracy and reliability.

Figure 5 - Example Response from the Regulation Assistant

**Important Note** Claudia is powered by artificial intelligence (AI), which means inaccuracies may still occur. Although the information provided always refers to official regulatory documents, we strongly recommend that you: 1. Verify the details provided, especially when they involve critical information such as names of individuals or locations. 2. Consult legal or accounting professionals when making important decisions related to regulatory compliance. # HR Assistant **To be able to use the HR assistant, make sure to connect your CATAPA account to Claudia by following this guide.** In the HR Assistant mode, Claudia helps you analyze various types of HR-related data, particularly those involving employees, payroll, attendance, and time management. When presenting the results, Claudia provides context-based interpretations and suggests actionable steps that you can take based on the findings. **How the HR Assistant Works** HR assistant works by analyzing data in your CATAPA account on your behalf. The privilege and permission to the data is based on your role in CATAPA. For example, if you have administrator role in your CATAPA account then you will be able to analyze all data. Another example if your account is limited to employees in certain area, then the data will be limited to **Please note** that currently, we only **anonymize and deanonymize data during the generation of responses**. **Features** **PII Masking:** We implement PII Masking so that the AI does not directly have access to your employee's personal data. Currently, the supported PII types include: 1. KTP 2. NPWP 3. PROJECT 4. ORGANIZATION 5. EMPLOYEE\_ID 6. FAMILY\_CARD\_NUMBER (KK) 7. FACEBOOK\_ACCOUNT 8. LINKEDIN\_ACCOUNT 9. BANK\_ACCOUNT 10. BPJS ACCOUNT (BPJS Kesehatan) 11. EMAIL\_ADDRESS 12. PHONE\_NUMBER 13. PERSON Supported Types of Analysis: 1. **Employee data analysis** 2. **Payroll and salary analysis** 3. **Attendance and time-tracking analysis** The detailed list of supported data that can be analyzed are listed at the bottom of this page. Below is the example question and result from Claudia:

Figure 6 - Example Response from the HR Assistant

**Important Note** 1. Ensure that your data is complete for optimal analysis results. 2. Claudia analyzes data based on what is available in your system, and incomplete data may lead to inaccurate insights. 3. Claudia syncs your data daily everyday, so today's data update will be reflected on tomorrow's **Claudia HR Assistant supported data** | Data | Field | | ---------------------------------- | ------------------------------- | | Bank Account | ID | | Bank Account | Employee ID | | Bank Account | Account Number | | Contact Information | ID | | Contact Information | Email | | Custom Data Metadata | ID | | Custom Data Metadata | Name | | Custom Data Metadata | Type | | Custom Data | ID | | Custom Data | Employee ID | | Custom Data | Custom Data Metadata | | Custom Data | Value | | Custom Data | Date | | Education | ID | | Education | Employee ID | | Education | Major | | Education | Level | | Education | Institution Input | | Employee | ID | | Employee | Start Date | | Employee | Location of Hire | | Employee | Manager ID | | Employee | Name | | Employee | Join Date | | Employee | Active Status | | Employee | Identification Number | | Employee Details | ID | | Employee Details | Place of Birth | | Employee Details | Name | | Employee Details | Blood Type | | Employee Details | Marital Status | | Employee Details | Gender | | Employee Details | Date of Birth | | Employee Details | Family Card Number | | Employee Details | Race | | Employee Details | Religion | | Employment Status | Employee ID | | Employment Status | Location | | Employment Status | Organization | | Employment Status | Job Title | | Employment Status | Job Level | | Employment Status | Employment Type | | Employment Status | Employment Status Type | | Employment Status | Contract Start | | Employment Status | Contract End | | Employment Status History | Employee ID | | Employment Status History | Location | | Employment Status History | Organization | | Employment Status History | Job Title | | Employment Status History | Job Level | | Employment Status History | Employment Type | | Employment Status History | Employment Status Type | | Employment Status History | Effective Date | | Employment Status History | Contract Start | | Employment Status History | Contract End | | Employment Status Types | ID | | Employment Status Types | Name | | Organization | ID | | Organization | Name | | Job Title | ID | | Job Title | Name | | Job Level | ID | | Job Level | Name | | Location | ID | | Location | Name | | Employment Type | ID | | Employment Type | Name | | City | ID | | City | Name | | Education Major | ID | | Education Major | Name | | Education Level | ID | | Education Level | Name | | Family | ID | | Family | Employee ID | | Family | Name | | Family | Gender | | Family | Relation | | Job Experience | ID | | Job Experience | Employee ID | | Job Experience | Job Title | | Job Experience | Job Level | | Job Experience | Start Date | | Job Experience | End Date | | Salary Item | ID | | Salary Item | Code | | Salary Item | Name | | Salary Payment | ID | | Salary Payment | Employee ID | | Salary Payment | Total THP | | Salary Payment | Payment Date | | Salary Payment | Salary Payment Summary | | Salary Payment Summary | ID | | Salary Payment Summary | Transferred Amount | | Salary Payment Summary | Payment Date | | Salary Payment Summary | Company Bank Account | | Salary Payment Summary | Number of Transferred Employees | | Salary Payment Summary | Number of Processed Employees | | Termination | ID | | Termination | Employee ID | | Termination | Effective Date | | Termination | Reason | | Termination | Approval Status | | Termination Reason | ID | | Termination Reason | Name | | Family Relation | ID | | Family Relation | Name | | Family Relation | Relation Type | | Race | ID | | Race | Name | | Religion | ID | | Religion | Name | | Employee Salary Template | ID | | Employee Salary Template | Effective Date | | Employee Salary Template | Salary Template | | Company Bank Account | ID | | Company Bank Account | Bank Branch | | Company Bank Account | Name | | Bank Branch | ID | | Bank Branch | Bank | | Bank | ID | | Bank | Name | | Additional Income Group | ID | | Additional Income Group | Code | | Additional Income Group | Name | | Additional Income Group | Type | | Additional Income Process Employee | ID | | Additional Income Process Employee | Additinoal Income Process | | Additional Income Process Employee | Employee ID | | Additional Income Process Employee | Total | | Additional Income Process Employee | Total After PPh | | Additional Income Process | ID | | Additional Income Process | Additional Income Group | | Additional Income Process | Additional Income Date | | Additional Income Process | Payment Date | | Additional Income Process | Type | | Salary Calculation | ID | | Salary Calculation | Employee Workday | | Salary Calculation | Normal Workday | | Salary Calculation | Employee Salary Template | | Salary Calculation Detail | ID | | Salary Calculation Detail | Salary Calculation | | Salary Calculation Detail | Amount | | Salary Calculation Detail | Salary Template Detail | | Salary Template Detail | ID | | Salary Template Detail | Salary Item | | Salary Template Detail | THP | | Attendance Detail Recapitulation | ID | | Attendance Detail Recapitulation | Date | | Attendance Detail Recapitulation | Attendance | | Attendance Detail Recapitulation | Shift | | Attendance Detail Recapitulation | Shift Time In | | Attendance Detail Recapitulation | Shift Time Out | | Attendance Detail Recapitulation | Attendance Time In | | Attendance Detail Recapitulation | Attendance Time Out | | Attendance Detail Recapitulation | Employee ID | | Attendance Detail Recapitulation | Requested Overtime | | Attendance Detail Recapitulation | Paid Overtime | | Attendance Detail Recapitulation | Overtime Description | | Attendance | ID | | Attendance | Employee ID | | Attendance | Date | | Attendance | Time In | | Attendance | Time Out | | Attendance | Offset | | Attendance | Description | | Attendance | Attendance Status In | | Attendance | Attendance Status Out | | Attendance Status | ID | | Attendance Status | Code | | Attendance Status | Name | | Attendance Status | Attendance Type | | Shift | ID | | Shift | Code | | Shift | Name | | Shift | Time In | | Shift | Time Out | | Shift | Break Start | | Shift | Break End | | Presence Record | ID | | Presence Record | Employee ID | | Presence Record | Date | | Presence Record | Time | | Presence Record | Time UTC | | Presence Record | Offset | | Presence Record | Location Status | | Presence Record | Photo Status | | Presence Record | Latitude | | Presence Record | Longitude | | Presence Record | Approval Status | | Presence Record | Rejection Status | | Presence Record | Description | | Presence Entry | ID | | Presence Entry | Employee ID | | Presence Entry | Date | | Presence Entry | Time In | | Presence Entry | Time Out | | Presence Entry | Offset | | Presence Entry | Approval Status | | Presence Entry | Rejection Status | | Presence Entry | Description | | Presence Entry | Attendance Status In | | Presence Entry | Attendance Status Out | | Leave Balance | ID | | Leave Balance | Employee ID | | Leave Balance | Year | | Leave Balance | Start Date | | Leave Balance | End Date | | Leave Balance | Extended Date | | Leave Balance | Remaining | | Leave Balance | Extended Remaining | | Employee Roster | ID | | Employee Roster | Date | | Employee Roster | Shift | | Employee Roster | Employee ID | | Workgroup Workday | ID | | Workgroup Workday | Date | | Workgroup Workday | Shift | | Workgroup Workday | Workgroup | | Workgroup | ID | | Workgroup | Code | | Workgroup | Name | | Employee Identity Card | ID | | Employee Identity Card | Employee ID | | Employee Identity Card | Identity Card Number | | Employee Identity Card | Identity Card | | Identity Card | ID | | Identity Card | Name | | Identity Card | Identity Card Type | # Employee Assistant **To be able to use the Employee assistant, make sure to connect your CATAPA account to Claudia by following this guide.** **How Employee Assistant works** The Employee Assistant allows both employees and managers to perform HR-related submissions independently. In addition, managers can directly analyze and review data related to their subordinates. The managers and subordinate relation is based on your account's setup in CATAPA. **Features** 1. **Self Service Request:** 1. Absence request 2. Leave requests 3. Overtime request 4. Presence correction 2. **Manager Assistant:** Managers can review and approve employee submissions, as well as analyze subordinate data. **PII Masking:** We implement PII Masking so that the AI does not directly have access to your employee's personal data. Currently, the supported PII types include: 1. KTP 2. NPWP 3. PROJECT 4. ORGANIZATION 5. EMPLOYEE\_ID 6. FAMILY\_CARD\_NUMBER (KK) 7. FACEBOOK\_ACCOUNT 8. LINKEDIN\_ACCOUNT 9. BANK\_ACCOUNT 10. BPJS ACCOUNT (BPJS Kesehatan) 11. EMAIL\_ADDRESS 12. PHONE\_NUMBER 13. PERSON

Figure 7 - Example Response from the Employee Assistant

# Submitting and Approving Requests Welcome! Claudia Virtual Assistant is designed to simplify the submission and approval process within your organization. This guide provides complete instructions covering navigation, submission steps, and approval actions. #### Navigation in Claudia Virtual Assistant **Getting Started** 1. Open Claudia by clicking the following link: [**claudia.catapa.com**](https://claudia.catapa.com) 2. Log in to your account\ Enter your CATAPA credentials to access Claudia’s features. 3. Select the **Employee Assistant**\ After logging in, choose the application named **Employee Assistant** to begin interacting with Claudia. #### **Important Notes** * Ensure that you have the appropriate access level to perform submission or approval actions. * Provide complete information so Claudia can process your request immediately without needing additional clarification. # Submissions Through Claudia Virtual Assistant ### 1. Leave Request **How to Submit:** Type your request clearly, for example: > *“I would like to apply for a half-day leave tomorrow from 09:00 to 13:00.”* Claudia will immediately process your request and provide a confirmation.\ You may add an optional description if needed. **Example:** You: *“I would like to apply for a half-day leave tomorrow from 09:00 to 13:00.”*\ Claudia: *“Your half-day leave on \[date] from 09:00 to 13:00 has been submitted. Would you like to add a description?”*\ You: *“No, thank you.”*\ Claudia: *“Your leave request has been sent for approval.”* ### 2. Attendance Correction **How to Submit:** Provide the correction details, for example: > *“I would like to correct my attendance on October 10 to clock in at 08:30 and clock out at 17:00.”* Claudia will process the correction and provide confirmation.\ You may add an optional description if necessary. **Example:** You: *“I would like to correct my attendance on October 10 to clock in at 08:30 and clock out at 17:00.”*\ Claudia: *“Your attendance correction for October 10 has been submitted with clock-in time 08:30 and clock-out time 17:00. Would you like to add a description?”*\ You: *“No.”*\ Claudia: *“Your attendance correction has been sent for approval.”* ### 3. Absence Report **How to Submit:** Specify the type of absence and the date, for example: > *“I would like to report an absence today due to illness without a doctor’s note.”* Claudia will record your report and provide confirmation.\ You may also add an optional description. **Example:** You: *“I would like to report an absence today due to illness without a doctor’s note.”*\ Claudia: *“Your absence for ‘Sick without doctor’s note’ on \[date] has been recorded. Would you like to add a description?”*\ You: *“No.”*\ Claudia: *“Your absence report has been saved.”* ### 4. Attendance Cancellation **How to Submit:** State the attendance entry you wish to cancel and the reason, for example: > *“I would like to cancel yesterday’s attendance due to personal reasons.”* Claudia will register the cancellation and provide confirmation. **Example:** You: *“I would like to cancel yesterday’s attendance due to personal reasons.”*\ Claudia: *“Your attendance on \[date] has been cancelled for the following reason: personal reasons.”*\ You: *“Thank you.”* ### 5. Overtime Request **How to Submit:** Specify the type of overtime, the duration, and your compensation preference, for example: > *“I would like to record 2 hours of pre-shift overtime today, and I prefer payment.”* Claudia will process your request and send a confirmation. **Example:** You: *“I would like to record 2 hours of pre-shift overtime today, and I prefer payment.”*\ Claudia: *“Your 2 hours of pre-shift overtime on \[date] has been recorded with payment as the preferred compensation. Would you like to add a description?”*\ You: *“No.”*\ Claudia: *“Your overtime request has been sent for approval.”* # Approvals Through Claudia Virtual Assistant ### 1. Request the List of Pending Submissions To view all items awaiting your approval, type: > *“Show pending overtime requests.”* Claudia will then display a list of submissions that require your review. ### 2. Approve All or Approve Selectively * To approve all pending requests at once, type:\ \&#xNAN;**“Approve all.”** * To approve only specific items, mention the employee’s name or provide particular details related to the submission. Claudia will confirm once the approvals are completed successfully. #### Example Interaction You: *“Show pending overtime requests.”*\ Claudia: *“Here are the overtime requests waiting for your approval: 1. \[Employee Name], \[Overtime Details]...”*\ You: *“Approve all.”*\ Claudia: *“All overtime requests have been approved. Thank you.”* # Help Center Assistant The Help Center Assistant guides CATAPA users by providing information and step-by-step instructions regarding features and modules. **Features** 1. **CATAPA Service Guides:** Access information about modules such as Payroll, BPJS, and other features across all help center content in CATAPA 2. **Sources**: Claudia will give you the relevant source, as a Claudia will intelligently use multiple sources to give you answers The supported documents are all of the user guide available in help.catapa.com You may ask the question in the input box, and Claudia will answer along with the source as the screenshot below:
# Tips for Using Claudia Effectively To get the most out of your interaction with Claudia, consider the following tips: 1. **Ask Specific Questions**\ The more specific your question, the better the response will be. For example, instead of asking “Show payroll data,” you can provide more detail such as “Show overtime payments for October 2023.” 2. **Use Everyday Language**\ Claudia understands natural, conversational language. You do not need to use specific keywords, simply ask your question as you would when speaking to a colleague. 3. **Learn from the Sample Questions**\ The questions provided in the question library are excellent references for understanding the types of prompts you can use and the structure that works best. 4. **Utilize the Conversation Panel**\ All of your previous interactions with Claudia are recorded and can be accessed anytime, allowing you to revisit earlier responses whenever needed. # Security and Privacy Claudia is designed with strong attention to data security and user privacy. The following principles ensure that your information remains protected at all times: 1. **Role-Based Access**\ The data and assistant modes you can access are tailored to your user role, ensuring that each user only sees information they are authorized to view. 2. **Data Protection**\ All data and interactions are encrypted and securely stored to prevent unauthorized access. 3. **Data Masking**\ Claudia automatically masks personal information to protect sensitive data during processing and interaction. # Clear Chats You can delete conversation by click **View All** from the left sidebar, choose chat that you want to clear, and then click the red **Delete** button. You also can clear all chats by click **Select All** and then click the red **Delete** button. This will remove chat history.
# Sources You can see the **references or sources** for each response by clicking the Source button at the bottom of the response. A dialog box on the right sidebar will display a list of sources. You can click each link to verify the information directly from the source.
# Settings and Report # Change Theme You can change the theme from the left sidebar by going to Settings and selecting a theme. Currently, there are three options: **Light**, **Dark**, and **System**.
# Change Language You can change the language by going to **Settings** from the left sidebar and selecting your preferred language. Currently, the available languages are **English** and **Bahasa Indonesia**.
# Report an Issue You can report an issue from the left sidebar by going to **Help Center** and selecting **Report an Issue.** This will redirect you to Feedback Form Claudia where you can submit your report.
# Manage Shared Chat Manage and organize your shared chat in one place You can access Shared Chat Management menu in Profile Setting
In this menu, you can: 1. View the list of chats you shared before 2. Remove the shared chat link 3. Shortcut to go to the chat room # Connector as Search Option Connectors let you search directly within your own tools and data sources (e.g., Gmail, Google Calendar, Google Drive, GitHub) instead of only relying on the web or knowledge base. This helps you quickly find emails, files, events, or code without leaving the chat ## **Prerequisite** Before using connectors in search, you must connect your identity for each service, currently supported connectors: * Gmail * Google Calendar * Google Drive * GitHub Connectors can be linked in two ways: * **Profile Settings** → Connect your account once and reuse across sessions. * **Tool Dropdown → Search Option** → Connect directly when you select a connector. ### How to Use Connectors in Search
1. Open the Tools dropdown in chat input. 2. In the Tool dropdown, choose **Search** and toggle ON the connector(s) you want to search. 3. Enter your prompt and send. 4. Results will be retrieved from the selected connector(s). # Sign Out To sign out, simply click your profile menu in the bottom-left corner and select the **Sign Out** option.
# FAQ # General Information ### What is Claudia? Claudia is a virtual HR assistant designed to help HR teams, managers, and employees quickly access information related to regulations, payroll, attendance, and other employee data. It simplifies administrative tasks and delivers relevant answers based on the questions you ask. ### What assistant modes are available in Claudia? Claudia offers four assistant modes to support different needs: * **Regulation Assistant:** Provides information on government regulations, including taxation, labor laws, and payroll rules. * **Help Center Assistant:** Offers guidance on how to use CATAPA. * **HR Assistant:** Helps analyze employee data, payroll, attendance, and time management. * **Employee Assistant:** Enables employees and managers to perform self-service tasks such as leave applications and employee data analysis. ### Can Claudia assist with topics outside HR and regulations? Currently, Claudia focuses on HR, regulations, payroll, and time management. However, as the system continues to evolve, additional topics and capabilities may be introduced in future updates. # Getting Started #### How do I start a conversation with Claudia? To begin a conversation, click the **New Chat** button on the home page. If you want to continue a previous conversation, simply select it from the chat history list on the left side of the screen. #### How do I switch assistant modes in Claudia? To switch assistant modes, choose one of the available assistants displayed on the screen (Regulation Assistant, HR Assistant, or Employee Assistant). Make sure to select the mode that best matches your question to receive an accurate response. # Accuracy, Reliability, and References ### What should I do if Claudia provides inaccurate information? Although Claudia is designed to provide accurate responses, errors may still occur. If you receive information that seems incorrect, verify it with official sources or consult relevant professionals. You can also send feedback using the **Thumbs-up** or **Thumbs-down** icons in each chat. ### Can Claudia provide references or source information? Yes. In Regulation Assistant mode, Claudia provides references to official documents such as laws and government regulations. If an answer is based on external sources, those references will also be included. ### Is Claudia always accurate when giving regulatory advice? Claudia uses trained data and the latest available regulations to provide answers. However, due to the complexity and frequent updates in regulations, it is advisable to verify important information or consult legal or accounting professionals before making critical decisions. ### What should I do if Claudia cannot answer my question? If Claudia cannot provide a sufficient answer, try rephrasing your question using clear and specific details. If the issue persists, you may submit feedback through the Thumbs-up or Thumbs-down buttons, or contact CATAPA Support for further assistance. # Usage Limits and System Behavior ### Are there usage limits in Claudia? Yes. Claudia applies usage limits to ensure fairness for all users. Each user is allowed a specific number of conversations and data analyses within a certain period. If you are nearing your limit, Claudia will notify you. # Security and Privacy ### Can Claudia handle sensitive data? Yes. Claudia is built with strong security features to protect sensitive information. All data entered into the system is encrypted and masked before being processed by AI. Additionally, **Role-Based Access** ensures that only authorized users can view certain data. # Feedback and Support ### How can I provide feedback to improve Claudia? Claudia continuously improves based on user feedback. You can share your feedback about answers or features using the Thumbs-up or Thumbs-down buttons in each chat. Your input is valuable in helping enhance Claudia’s performance. ### How can I contact support if I need further assistance? If you need additional help or face technical issues, you may contact CATAPA Support at **150150** or via the email address provided in the application. # Digital Employee Features
Claudia is your digital employee designed to automate and enhance human resources operations. As a digital employee, Claudia integrates seamlessly with your tools and processes, handling tasks from daily operations to team communications. Below are key capabilities, focused on practical use cases for HR teams. ### Administrations and Operations Claudia excels at processing HR data and documents to provide insights and automate workflows, integrating with systems like CATAPA for streamlined operations: * Analyze resumes, performance reviews, or employee surveys to extract key insights, such as skill gaps or sentiment trends, helping in talent assessment. * Perform actions like updating employee records, tracking leave requests, or generating reports directly in your HRIS. * Manage candidate pipelines by moving applicants through stages (e.g., screening to interview), applying role-based access controls (RBAC) to ensure only authorized team members view sensitive data. * Update data in real-time, such as syncing new hire info across systems, reducing manual errors during onboarding. ### Communication Channels Claudia communicates naturally across multiple platforms, acting just like a team member to handle inquiries and updates: * Read and reply to emails for tasks like responding to job applications, sending offer letters, or following up on employee queries. * Engage via WhatsApp for quick chats with candidates or staff, such as confirming interview times or sharing policy updates. * Interact through the Claudia web chat UI for internal HR support, answering questions on benefits or payroll in real-time conversations. ### Customizable Actions Based on Your Business Process Claudia adapts to your unique HR workflows through its agentic pipeline, allowing tailored automation: * Configure sequences for processes like recruitment, where Claudia screens resumes, schedules interviews, and notifies teams based on your company's steps. * Adjust for compliance needs, such as automating approval chains for promotions or terminations. * Integrate custom triggers, like alerting managers when there are actionable in your current system. ### Popular App Support Claudia connects with widely used tools to enhance HR efficiency: * Integrate with Google Workspace or Microsoft 365 for email, calendars, and document sharing during onboarding. * Link to collaboration platforms like Slack or Microsoft Teams for team announcements and feedback collection. * Support for file storage services (e.g., Google Drive, OneDrive) to organize HR files securely. * Connect to code repositories like GitHub for tech hiring processes, such as reviewing candidate portfolios. ### Long-Term Memory Claudia remembers interactions and preferences for personalized support by retaining employee interaction history to provide context-aware responses (e.g., recalling a staff member's previous leave requests or approvals). ### Automated and Scheduled Background Processes Claudia works behind the scenes to handle routine tasks without constant oversight: * Run scheduled reports and delivered via email or chat. * Automate background checks or reference verifications for new candidates. * Monitor HR metrics in the background from your HRIS. # Chat Configuration # Upload Document / File There are two ways of uploading a document or file: 1. **Directly in the chat box** 2. **From the Chat Control panel** on the top-right corner ### Attach File in the Chat Box
You can upload a document or file by clicking the **Attach File** icon in the chat box.\ There are two ways to add files: 1. Upload from your computer 2. Add directly from Google Drive (if your account is connected) ### Upload From Computer
\ When you choose to upload from your computer, a window will appear where you can drag and drop your files or click **Select** to pick them manually. You can upload up to five files at once, and each file can be up to 10 MB in size The types of files you can upload depend on both the configuration of the **Document Processing Orchestrator (DPO)** module and the **LLM you have selected**. Some examples include: 1. Image (.png, .jpg, .jpeg, .webp) 2. Spreadsheet files (.xlsx, .csv) 3. Audio files (.wav, .mp3) 4. Document files (.pdf and .docx) 5. Log or markup files (.txt, .md, and .log) 6. Source Code (.java, .py) 7. Video (supported by specific LLM such as Google's **Gemini**) The maximum file size allowed is 10 MB per file. ### Upload From Google Drive
If you choose to add files from Google Drive, the system will prompt you to sign in and grant access the first time you use this option. After authentication, or if you have already connected your Drive before, you can immediately browse your Google Drive files and select the ones you want to attach. ## Upload File from Chat Control You can also attach files through the **Chat Control** panel by clicking the **Files** option in the top-right corner.
This opens the file management sidebar, where you can: * View all files you have already uploaded * Click **Upload File** to add new files from your computer or Google Drive (the same interface as the chat box attachment) This method is useful if you want to upload files **before** writing your message or manage previously uploaded files in one place. ## Preview Uploaded File Once you send a message with an attachment, you can **preview the file directly in the chat area** by clicking on the file preview.
After that a popup will open showing the full file content. # Web Search You can browse the internet for more up-to-date information by enabling the **Web Search** tool. To do this, click the **Tools** button in the chat box, select **Search**, and then toggle **Web** on. Once it is enabled, simply type your question in the chat box and send it — the answer will include information from the web.
# Deep Research Unlock insights with our deep research feature, diving into comprehensive data analysis and uncovering hidden patterns to inform your decisions.
Deep Research allows you to get a more thorough, multi-step exploration of a topic compared to a standard chat response. To enable it, click the **Tools** button in the chat box, select **Deep Research**, and then choose between the available mode. * **Essentials** – A quicker research mode that gathers key information and summarizes the main points. * **Comprehensive** – A more detailed research mode that runs longer, reviews more sources, and generates a deeper, more extensive answer. Once selected, simply type your question in the chat box and send it — the system will perform a structured research process and return a richer, well-organized response. Current capability of Deep Research: 1. Perform information retrieval to internet. 2. Doing deep research of the topic based on prompt sent. 3. Includes in-line citations to indicate the origin of specific facts or claims. These citations are clickable and will open the original source link directly. 4. Provides a complete list of references used at the end of the response. Each reference is clickable and links to the original source for easy access. # Switch Large Language Model (LLM) You can switch models by clicking the **Models** options in the chat box. From there, you can select the desired model.
Or, you can change the model during the message regeneration.
# Task Scheduler ## Schedule a Task You can schedule a task in Claudia. The tasks can vary, such as: * Send a prompt to LLM * Send instructions to AI Agent * Perform a search (Web/Connector) * Conduct deep research To create or schedule a task, type a prompt in the chat input box. After selecting one of the tasks above, use the arrow button beside the send button to schedule the message.
**Note**: Currently, you cannot schedule a task for a prompt that contains a file attachment. There are several types of schedules: * Once * Interval (minutes) * Interval (hours) * Daily * Weekly * Weekdays * Monthly * Yearly
Once you are firm with your choice, hit the Save button. A toast message informing you that the task is already created will appear, along with a link to open the Task List. Once the task meets the scheduled time, it will be executed as a new conversation.
## View Task List To view task lists, open the toast notification that appears after scheduling a task, or simply go to the "Schedule Task" menu in the profile section at the bottom left.
This menu includes two tabs:**Scheduled** * Displays all created tasks and their execution status. * Options available: Set task to inactive, edit, or delete a task. **Completed** * This section shows all tasks that have already been executed. * You can also click the completed task to go to the conversation containing the scheduled task. ## Edit Task To edit a task, select "Edit" from the Task List. This will redirect you to a page where you can modify the task or its prompt. Note, to change the schedule, simply click the link "Change schedule here". Once you are done with the edit task, click the check button. A toast message will appear informing you that the adjustment was successful. You can check the task via the Task List menu.

# Chat Input and Responses # Step Indicator Capability to show steps / process that run during user send message until the response is generated. It will show the steps involved from backend and also the total duration.
# Message Interaction ## Send Message You can send a message by click "Send Message" button or hit Enter button once you already type the message in the chat input area ## Stop Response
If you change your mind and **want to stop the message** you sent earlier, simply click the **Stop Response** button to cancel it. ## Edit Input
If you want to **edit your message**, simply click **Edit**, type your new message, and then click the **Send** button or press **Enter** on your keyboard.You can easily revisit previous or recent questions by clicking the arrow on each message. ## Copy Prompt/Response
You can easily copy the response by clicking the **Copy** button at the bottom of each response. The response is in Markdown format. ## Thumbs Up / Thumbs Down
You can rate a chatbot response with **Thumbs Up (👍)** or **Thumbs Down (👎)** and optionally add **tags** and a short **reason**. Your feedback is used by the **Memory** system to improve future replies (tone, length, clarity, language). ## Regenerate Message
You can easily regenerate the answer by clicking the **Regenerate Message** button and selecting the method you want to use. By **default**, the "**Try Again**" option will **use the most recent method you selected** to generate the message. ## Export Response You can export an individual chat bubble if you want to save or share its contents outside the application.
To do this, click the **three dots (⋮) menu** located below the bubble you want to export. From the menu that appears, choose **Export**, and then select the format you want — for example, **PDF**. The system will generate the file and allow you to download it to your computer. ## Mask Personal Data The **Mask Personal Data** option allows you to quickly hide sensitive information in the chat response, such as names, emails, or phone numbers, before sharing or exporting the conversation.
When you click this option from the eye-icon (:eye:) menu under a chat bubble, the system will automatically anonymize or mask identifiable details in that specific response (e.g., names, emails, phone numbers). ## Read Aloud The **Read Aloud** option lets the system speak the content of the selected chat bubble.
Clicking this option will trigger text-to-speech playback, so you can listen to the response instead of reading it. **Limitation**: It only supports up to **100 characters (including spaces)** to prevent exceeding the quota. ## Understanding Chat History (Short-Term Memory) Chat history refers to the record of past conversations or messages exchanged between users in a chat application. It serves multiple purposes, such as providing context for ongoing discussions, facilitating the retrieval of information, and ensuring accountability in communication.
**Key Benefits** 1. **Contextual Reference**: Maintains a record of past interactions, enabling users to recall previous conversations. 2. **Information Retrieval**: Allows users to search and retrieve information exchanged in past chats quickly. 3. **Accountability**: Provides a transparent log of communications that can be used for verifying information or resolving disputes. Understanding the significance of maintaining and managing chat history effectively can greatly enhance communication and efficiency within a team. # Sources You can see the **references or sources** for each response by clicking the Source button at the bottom of the response.
A dialog box on the right sidebar will display a list of sources. You can click each link to verify the information directly from the source. # Related Queries When using Web Search, the chatbot can provide **recommended related questions based on the given context**.
Simply select one of the related questions, and it will automatically fill the chat box. Then you need to click the **Send** button or press **Enter** on your keyboard. # Question Library A centralized repository of questions that combines admin configured prompts and or auto-generated FAQ, for guided conversation starters.
Questions in question library are organized by categories for easier navigation and discovery. In addition, the Question Library appears contextually in the chat input area, giving users quick access even after multiple turns in an ongoing conversation
When clicked, the question can directly processed, or editable, depends on the configuration in the admin dashboard. # Show Activity and Thinking This feature displays intermediate steps or background actions performed by the AI while it processes a request. For example, users can see when the system is analyzing a query, fetching data from multiple sources, or reasoning through multi-step instructions. It helps users understand how the AI arrives at its conclusions and provides reassurance that the model is actively working on their request.
Currently, **Show Activity and Thinking** is available in several features including **AI Agent**, **Deep Research**, and **Web Search,** where it reveals activities such as retrieving documents, summarizing results, or planning next actions.
# Resources # Relevant Documents To support your understanding of Claudia’s capabilities and the broader ecosystem of CATAPA Digital Employees, the following documents provide deeper insights into the architecture, principles, and real-world impact of Claudia. Each resource is available as a hyperlink and offers additional context to complement this documentation. ### 1. CATAPA HR Digital Employees [CATAPA HR Digital Employees](https://docs.google.com/presentation/d/1iyHUohUDC_Uvt8eAqPy-aTL9NFk5e0uPOIESbVy4ZuE/edit?usp=sharing) This document introduces the concept of **Digital Employees** in CATAPA, describing how AI-native agents like Claudia enhance HR operations through superhuman capabilities and photographic memory. ### 2. Claudia Introduction - Master Slides [Claudia Introduction - Master Slides](https://docs.google.com/presentation/d/1WqP1Vd9DF-SDFMl7dEYG0vHbyQRPn6gBSeuWHfw-IL8/edit?usp=sharing) This introductory deck presents Claudia as a **proactive Digital Employee**, not just a typical chatbot, capable of performing continuous HR analysis and generating reports 24/7. It summarizes Claudia’s core strengths, including Predictive Analysis, HR Payroll evaluation, and regulatory support. # Release Notes You can find list of historical release note from Claudia in this page # 12 March 2026 On March 12th, 2026, we are releasing a new version of CATAPA Chat based on GLChat version 0.8.023. *** ## Main Release #### Multi-Tenant Platform Support – New CATAPA Chat now supports a multi-tenant platform architecture, allowing multiple organizations to operate independently within the same system while keeping their data, configurations, and access securely separated. Each organization can manage its own environment—including applications, users, agents, and configurations—without affecting other organizations. **User Action**: Not required. *** #### [Admin Dashboard for Organization Management](https://gdplabs.gitbook.io/catapa/catapa-hris/catapa-chat/features-and-how-to-guide/organization-admin-dashboard) – New A new Admin Dashboard is now available in CATAPA Chat Settings, providing organization administrators with a centralized place to manage their environment. From this dashboard, administrators can configure applications, manage users, roles, and permissions, set up AI agents, integrate external tools or APIs, and manage knowledge bases. Additional settings such as pipelines and presets can also be configured to support organizational workflows. **User Action:** Not required. *** #### Global Admin Dashboard for Platform Governance – New A Global Admin Dashboard has been introduced for the CATAPA Chat and CATAPA teams to manage the platform across organizations. This dashboard provides centralized visibility and control for managing organizations, assigning administrative access, configuring global defaults, and monitoring subscriptions and usage across the platform. While each organization manages its own environment through the Admin Dashboard, the Global Admin Dashboard enables platform-level oversight and governance. **User Action:** Not required. # 30 January 2026 On January 30th, 2026, we are releasing new version of CATAPA Chat / Claudia, Based on GLChat version 0.7.0021 **Main Release** *** **Security & Data Protection – Improvement** * The app now runs on newer core platform versions with the latest security patches. * Document indexing now masks identifiable information (PII), improving how sensitive data is handled behind the scenes. **User Action** Not required. *** **Jump to Specific Moments in Video with Clickable Timestamps – New** You can now jump directly to specific moments in a video by clicking on the timestamps displayed. This makes it easier to review important sections without manually scrubbing through the whole video. **User Action** Not required. This feature appears automatically on videos that include timestamps. *** **More Comfortable Chat Experience – Improvement** The chat input box now expands and becomes scrollable, so you can write and review longer messages more easily. Text formatting (line breaks, lists, etc.) is also displayed more neatly and consistently. **User Action** Not required. Just use the chat as usual. *** **Smoother Loading States & More Accurate Latest Message Display – Improvement** * Loading indicators have been refined so “loading” screens feel smoother and are less likely to look stuck. * The display of the latest message branch in a conversation has been fixed, so you see the most recent replies more reliably. **User Action** Not required. *** **Clearer File Upload Behavior – Improvement** The upload area (dropzone) is now only active where file uploads are actually supported. This reduces confusion and errors when trying to upload files in places that don’t accept them. **User Action** Not required. Simply upload files where the upload area is visible and active. *** **Overall Stability Improvements – Improvement** Multiple fixes have been applied to reduce errors in edge cases, such as when the system temporarily can’t reach storage or when processing long‑running tasks and streams. You should experience a more stable app with fewer unexpected errors. **User Action** Not required. *** **Other Releases** The following changes are backend and infrastructure related. They don’t change how you use the app day to day, but they help keep the app fast, secure, and reliable: * Performance improvements in how sessions and background processes are handled. * Optimizations in how attachments/files are processed behind the scenes. * Updates to various technical components and dependencies to address vulnerabilities and improve stability. **User Action** Not required. All improvements are applied automatically. # 23 January 2026 This update is based on GLChat version 0.7.011 #### Key enhancements 1. Improved task interaction: Tasks are now clickable, enabling faster navigation and a more seamless workflow.
2. Default application preference: Users can set a default application in profile settings to streamline access and reduce repetitive steps.
3. Enhanced knowledge base capabilities: Support for more flexible knowledge base configurations has been added to improve how knowledge sources are organized and managed. 4. Refined scheduling experience: Updates to scheduling fields and behavior provide a clearer and more consistent experience when setting scheduled activities.
#### Reliability and performance improvements 1. Stronger session stability: Fixes were implemented to improve session retrieval and ensure organizational context is consistently applied. 2. Better rendering and content handling: Improvements were made to reduce issues when processing large images and to optimize markdown rendering for more consistent display. 3. Backend and pipeline stability: Multiple fixes improve processing reliability, logging traceability, and overall system robustness. 4. Interface and mobile refinements: A range of UI adjustments address visual issues and improve the experience across devices, including iOS. #### Security update 1. Security vulnerability mitigation: This release includes an important security update addressing a denial-of-service risk associated with the underlying runtime, improving platform safety for all users. # Overview
A **Digital Employee (DE)** is an AI-powered **digital colleague** designed to **amplify and extend human capability — not replace people**. Unlike traditional AI tools that respond only when prompted, a Digital Employee operates within a **defined business role**, owns **ongoing work** (not just one-off tasks), and is accountable for delivering **measurable outcomes**. A Digital Employee combines the strengths of both software and human workers.\ It has ***persistent identity*** (suce as its own email), ***rich in context***, ***proactive***, equipped with ***photographic memory***, and ***available 24/7***, while maintaining machine-level speed, consistency, and scalability. Humans remain essential for **judgment, strategy, and oversight**, while Digital Employees increase **operational speed, capacity, and consistency** across the organization. ### Why Digital Employees Exist Traditional AI systems are powerful but limited in how they contribute to real work. * **Chatbots** are conversational interfaces that answer questions. * **AI agents** can execute tasks when instructed. * **Human employees** bring context, judgment, and initiative. However, none of these alone fully solve the challenge of **scaling knowledge work while maintaining quality and consistency**. Digital Employees address this gap by combining the best aspects of both humans and AI systems. They function as **persistent AI workers** embedded into business processes — helping teams move faster while maintaining reliability and oversight. ### Key Benefits of a Digital Employee A Digital Employee is designed to behave more like a **colleague** than a traditional software tool. #### 🔁 *Rich in Context* A Digital Employee understands the broader business context of the work it performs.\ It can combine information from multiple sources to support decision-making and operational tasks. #### ��*Proactive* Rather than waiting passively for instructions, a Digital Employee can identify issues, surface insights, and initiate actions based on defined responsibilities. #### ��*Photographic Memory* A Digital Employee maintains consistent memory across interactions, allowing it to retain context and knowledge without the inconsistencies of human recall. #### ⏱️*Available 24/7* Digital Employees operate continuously without downtime, enabling organizations to handle workload spikes and time-sensitive operations.
## Human Employee and Digital Employee Digital Employees are designed to **work alongside human employees**, combining human-like contextual understanding with machine-level speed and reliability.
CapabilityHuman EmployeeDigital Employee
Rich in context✅ Yes✅ Yes
Proactive✅ Yes✅ Yes
Photographic memory❌ No✅ Yes
Available 24/7❌ No✅ Yes
Instant scaling during busy periods❌ No✅ Yes
Instant upskilling rollout❌ No✅ Yes
Handling multiple tasks in parallel⚠️ Limited✅ Yes
Consistent rule execution⚠️ Limited✅ Yes
Knowledge retention⚠️ Limited✅ Yes
Predictable cost structure⚠️ Limited✅ Yes
Strategic judgment & creativity✅ Yes❌ No
Empathy & relationship building✅ Yes❌ No
Digital Employees allow organizations to **increase operational capacity without increasing operational complexity**, while human employees remain responsible for **strategy, judgment, and decision-making**. # Digital Employee Concept A **Digital Employee (DE)** is designed to function similarly to a human knowledge worker inside an organization. Rather than acting as a simple chatbot or automation tool, a Digital Employee operates as a **role-based digital colleague** that can perform ongoing work, collaborate with people, and interact with enterprise systems. To achieve this “employee-like” presence, a Digital Employee is built from several core capabilities that mirror how a human employee operates. ### The Human Analogy One way to understand a Digital Employee is to compare it to a **human worker**. A human employee typically has: * an identity (name, role, contact information) * knowledge and memory * skills and expertise * tools to perform tasks * governance and supervision A Digital Employee is designed with the same structure.
ComponentHuman AnalogyPurpose
IdentityFace & IDGives the DE a name, role, and presence in the organization
Memory & CognitionBrainEnables reasoning, context awareness, and knowledge retrieval
SkillsKnow-howDefines how the DE performs repeatable workflows
ToolsHandsAllows the DE to interact with systems and perform actions
GovernanceAccess cardControls permissions, oversight, and safety rules
ConfigurationKeysDetermines system access and behavioral settings
AgentsTeam membersSpecialized agents that help complete tasks
These components allow a Digital Employee to operate as a **persistent organizational worker**, rather than a temporary AI session. ### How Digital Employees Work A Digital Employee typically acts as a **coordinator** that can delegate work to specialized AI agents. The workflow generally follows this pattern: 1. A user requests a task. 2. The Digital Employee interprets the request. 3. It selects the relevant **skills and tools**. 4. It may delegate subtasks to specialized agents. 5. Results are returned to the user or executed in connected systems. This architecture enables Digital Employees to handle **complex workflows**, not just simple questions. # Choosing the Right Solution: Chatbots, AI Agents, Digital Employees, and Humans ### Summary Organizations today have multiple ways to use AI in their operations. These solutions are not competing technologies — they solve **different types of problems**. Understanding the difference helps organizations choose the **right solution for the right job**. In general: * **Chatbots** help answer repetitive questions and retrieve knowledge in a low-cost way. * **AI Agents** help automate specific workflows or tasks behind the scenes. * **Digital Employees** take ownership of ongoing operational work with a built-in security, guardrails, and clear escalation plan to human. * **Human Employees** focus on making complex decisions, judgment, strategy, relationships, and leadership. Together, they form a **human–digital workforce**, where each type of worker contributes according to its strengths.
### Comparison The table below summarizes how these four types of workers differ. | Dimension | Chatbot | AI Agent | Digital Employee | Human Employee | | ----------------------- | -------------------- | -------------------- | --------------------------- | ----------------------------- | | **Primary focus** | Conversation | Task execution | Role ownership | Strategic work | | **Typical interaction** | User asks a question | User triggers a task | User delegates work | Self-directed | | **Scope of work** | Narrow | Narrow–moderate | Broad within a defined role | Broad and adaptive | | **Persistence** | Session-based | Task-based | Continuous operation | Continuous | | **Proactivity** | ❌ No | ⚠️ Limited | ✅ Yes | ✅ Yes | | **Memory** | ⚠️ Limited | ⚠️ Limited | ✅ Photographic memory | ⚠️ Human memory | | **System interaction** | Usually read-only | Yes via tools | Yes within role boundaries | Yes | | **Accountability** | None | Task completion | Role-based accountability | Organizational accountability | ### Chatbots Chatbots are designed primarily for **conversation and knowledge retrieval**. They are best suited for answering questions, summarizing documents, and helping users access information quickly. **Example use cases** * Internal knowledge base assistant * HR policy questions * Document summarization * Research and drafting Chatbots are ideal when the goal is to **help people think, learn, or write faster**, but they typically **do not perform operational tasks directly**. *** ### AI Agents AI Agents extend chatbots by adding the ability to **execute actions using tools and workflows**. Instead of only answering questions, an agent can interact with systems such as calendars, databases, or APIs. **Example use cases** * Scheduling meetings * Updating tickets * Generating reports * Sending automated notifications AI Agents are best suited for **structured tasks with clear inputs and outputs**. They are powerful operators but usually focus on **specific workflows rather than owning a full role**. *** ### Digital Employees Digital Employees represent the next step: AI systems that **operate within a defined business role**. A Digital Employee combines multiple agents, tools, and knowledge sources into a **single digital worker** with: * identity (name, email, role) * persistent memory * defined responsibilities * human supervision This allows organizations to delegate **ongoing operational work**, not just individual tasks. **Example use cases** * Meeting documentation and follow-ups * Recruiting coordination * Payroll compliance monitoring * Operational reporting Digital Employees are designed to **work alongside human employees**, helping teams scale operational capacity. *** ### Human Employees Human employees remain essential for tasks that require: * judgment and decision-making * creativity and strategy * relationship building * leadership and accountability Digital Employees and AI tools are designed to **support human work**, not replace it. *** ### Choosing the Right Solution | If your goal is… | The best solution is… | | ----------------------------------------- | --------------------- | | Answer questions or summarize information | Chatbot | | Execute a defined task or workflow | AI Agent | | Delegate ongoing operational work | Digital Employee | | Handle strategic decisions and leadership | Human Employee | This layered approach allows organizations to combine different capabilities within the same platform. *** ### Example Consider a meeting workflow: | Task | Best Solution | | ----------------------------------------------- | ---------------- | | Ask questions about previous meeting notes | Chatbot | | Schedule a meeting | AI Agent | | Manage meeting notes, summaries, and follow-ups | Digital Employee | | Decide on strategic next steps | Human Employee | # How Digital Employees Are Specialized
Digital Employees can perform different roles across an organization — such as recruiting assistants, payroll analysts, or project management assistants. To make this possible, Digital Employees are built using a **shared foundation** combined with **role-specific specialization**. {% hint style="info" %} **Why This Matters** This architecture allows organizations to: * deploy new Digital Employees faster * reuse existing infrastructure * maintain consistent governance and security * scale Digital Employees across departments As a result, organizations can build an expanding **digital workforce** where different Digital Employees support different business functions, while maintaining a consistent and secure platform. {% endhint %} *** ### The Shared Foundation Just like human employee who has common skills like reading email and writing documents, every Digital Employee starts from the same foundation (called *Core*). This foundation provides the capabilities that all Digital Employees need in order to operate within an organization. Examples include: * identity (name, role, and system accounts) * memory and contextual understanding * access to enterprise systems * ability to perform tasks through tools * governance and security controls Because these capabilities are already built into the platform, new Digital Employees do not need to recreate them. In technical term, we built these capabilities in Digital Employee Core. #### The Tools & Connectors Just as human employees leverage tools for work, DE utilizes functions and APIs in the forms of Tools and Connectors in order to interact with systems and perform actions. The wide range of Connectors allows DE to serve many industries and solve increasingly diverse problems.

GDP Labs has access to many Tools and Connectors. As per this writing, we have identified and performed security verification on 129 Connectors and 1,397 Tools from official sources—of which we self-host 29 Connectors and 386 Tools.

*** ### Role Specialization On top of the shared foundation, Digital Employees can be specialized for specific roles. This specialization includes: * domain knowledge * role-specific workflows * specialized skills * access to relevant tools and systems This allows organizations to create Digital Employees tailored to different functions. Examples include: * **Digital Employee for Recruitment** (Claudia✨ - HR Recruiter) * **Digital Employee for Payroll** (Claudia✨- HR Payroll Analyst; Claudia✨- HR Payroll Officer) * **Digital Employee for Project Management** (Pamela✨ - Project Manager) Each of these Digital Employees shares the same foundation but performs different types of work. *** ### Example Consider a **Digital Employee for Recruitment**. Shared foundation capabilities: * identity and email account * memory of past interactions * access to document and knowledge systems Role specialization: * screening candidate CVs * scheduling interviews * generating interview summaries * coordinating hiring workflows By combining the shared foundation with recruiting-specific capabilities, the Digital Employee can support the hiring process efficiently. # Benefits A clear example of the business value of Digital Employees is **DE HR Recruiter**, which assists and automates GDP Labs’ end-to-end talent recruitment process. 1. **Before Digital Employee:** Full human involvement across **14 recruitment steps** 2. **After Digital Employee:** Humans primarily focus on the **3 interview stages**, while the remaining steps shift to **supervision and approvals**. This reallocation meaningfully cuts manual effort and administrative overhead, allowing the hiring team to spend more time on high-value judgment and candidate interaction. The outcome is a measurable improvement in efficiency: with most operational steps automated or assisted, DE HR Recruiter delivers an estimated [**86% increase in hiring velocity**](https://docs.google.com/presentation/d/1sRmrh0mF2H4dIj48UDo_Ggiz58Ze3crJ_DiWskxtr34/edit?slide=id.g388fef65ac2_0_0#slide=id.g388fef65ac2_0_0). In practical terms, this means roles can be filled faster with less strain on HR capacity—enabling GDP Labs to scale hiring throughput without proportionally scaling headcount. # Chatbot vs AI Agent vs Digital Employee vs Human Employee ### Summary Organizations today have multiple ways to use AI in their operations. These solutions are not competing technologies — they solve **different types of problems**. Understanding the difference helps organizations choose the **right solution for the right job**. In general: * **Chatbots** help answer repetitive questions and retrieve knowledge in a low-cost way. * **AI Agents** help automate specific workflows or tasks behind the scenes. * **Digital Employees** take ownership of ongoing operational work with a built-in security, guardrails, and clear escalation plan to human. * **Human Employees** focus on making complex decisions, judgment, strategy, relationships, and leadership. Together, they form a **human–digital workforce**, where each type of worker contributes according to its strengths. #### **TL; DR** 1. Use **Chatbots** when you just need a simple, low-cost way to answer repetitive questions. 2. Use **AI Agents** when you want to automate specific workflows or tasks behind the scenes. 3. **“Hire” a Digital Employee** when you want something that: 1. Fits naturally into your organization (name, email, job title, supervisor). 2. Can reliably run real work processes end-to-end. 3. Comes with **built-in security, guardrails, and human oversight**. 4. Can be deployed **in days, not months**, using a standardized foundation (DE Core). 4. Keep **Human Employees** focused on what they do best: complex decisions, relationships, creativity, leadership—and let Digital Employees handle the repetitive, structured work around them. | | Chatbot | AI Agent | Digital Employee | Human Employee | | ------------------ | ---------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Ideal Use Case** |

• Customer FAQs

• Password reset help

• Store hours and basic info

• Lead capture forms

|

• Auto-routing tickets

• Generating reports

• Data entry/validation • Scheduled notifications

• Workflow orchestration

|

• First-line IT support

• Recruiting coordination

• Invoice processing

• Meeting scheduling

• Employee onboarding

• Customer success follow-up

|

• Strategic planning

• Complex negotiations

• Creative leadership

• People management

• Crisis response

• Relationship building

| #### **Quick Comparison** | Dimension | Chatbot | AI Agent | Digital Employee | Human Employee | | -------------------------- | ------------------------------------------------------ | ------------------------------------ | -------------------------------------------------- | ------------------------------------------------------ | | **Architecture** | Single agent, simple logic | Can be single or multi-agent | Usually multi-agent system | Human brain (parallel processing) | | **Scope** | Conversation-focused (very narrow) | Task-focused (narrow) | Role-focused (broader) | Role-focused with adaptability (broadest) | | **Temporal model** | Session-based: user asks → bot responds → session ends | Triggered → Execute → Done | Continuous operation with ongoing responsibilities | Continuous: manages multiple priorities over time | | **State persistence** | None or minimal (forgets after session) | Minimal (just for the task) | Rich (maintains context over days/weeks/months) | Very rich (years of experience and relationships) | | **Proactivity** | Zero (purely reactive) | Reactive to triggers | Monitors, reminds, follows up without prompting | High initiative: identifies problems and opportunities | | **Integration depth** | Minimal or none (isolated channel) | Point integrations (specific APIs) | Deep integration across multiple systems | Complete access to all tools and information | | **Organizational context** | Customer service tool | Technical asset / backend automation | Organizational member with identity and role | Legal employee with rights and responsibilities | #### **Detailed Comparison** ***Identity & Presence*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | ------------------------------------ | ----------------------------------- | ------------------------------------- | ------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------- | | **Who is it in your organization?** | Just "Support Bot" in a chat window | Usually invisible backend system | Has a name (e.g., "Alex - Recruiting Assistant"), email address, job title, reports to a manager | Legal person with employment contract, full organizational identity | | **Does it feel like a team member?** | No – feels like a tool or kiosk | No – most people don't know it exists | Yes – appears on org chart, has a "desk" in your systems | Yes – physical presence, personal relationships | ***Work Style & Autonomy*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | ------------------------------ | ----------------------------------------------------- | ------------------------------------------------------------- | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- | | **How it works** | **Reactive**: Sits idle until someone asks a question | **Task-based**: Executes when triggered by a system or person | **Proactive**: Monitors its work queue, follows up, reminds people, escalates issues | **Fully autonomous**: Self-directed, can reprioritize, handle ambiguity | | **Can it work independently?** | No – needs constant questions from users | Partially – can complete specific tasks end-to-end | Yes, within boundaries – handles routine work with occasional human check-ins | Yes – trusted to make judgment calls and strategic decisions | | **Does it take initiative?** | Never – purely responsive | Rarely – only within programmed workflows | Sometimes – can notice patterns and flag issues (e.g., "I see 10 candidates applied this week") | Regularly – identifies problems and proposes solutions | ***Memory & Context*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | -------------------------------- | ---------------------------------------------------- | ---------------------------------------------------- | -------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------- | | **Remembers past interactions?** | Basic or none – often starts fresh each conversation | Limited – remembers within a single task or workflow | Yes – maintains context over weeks/months (e.g., remembers candidate from last month) | Yes – rich personal memory and company history | | **Learns preferences?** | Minimal – may use basic rules | Some – can adapt to patterns in data | Yes – learns how you like things done, team conventions, organizational norms | Yes – deeply understands culture, unspoken rules, relationships | | **Accesses company knowledge?** | Usually just FAQ database | Can query specific systems when programmed | Accesses multiple knowledge bases, documents, prior conversations—like a unified brain | Accesses everything plus informal knowledge, relationships, institutional memory | ***Skills & Capabilities*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | -------------------------- | ------------------------------------------------ | --------------------------------------------------- | -------------------------------------------------------------------- | ------------------------------------------------------------------------ | | **How skilled is it?** | Follows scripts – can't deviate | Executes predefined workflows with some flexibility | Has modular skills that can be added/updated (like training courses) | Deep expertise, can learn new skills, transfers knowledge across domains | | **Can handle complexity?** | No – struggles with anything outside simple Q\&A | Moderate – good for structured, repeatable tasks | Good for well-defined knowledge work with clear processes | Excellent – handles ambiguity, nuance, and novel situations | | **Quality of work** | Consistent but limited | Consistent within scope, may miss edge cases | Generally reliable but requires spot-checking (\~90-95% accuracy) | Variable but uses judgment to maintain quality | ***Tools & System Access*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | ----------------------- | ---------------------------------------------- | --------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | --------------------------------------------------- | | **What can it access?** | Usually nothing, or just reads an FAQ database | Can call specific APIs: send emails, update databases, create tickets | Full toolkit: email, calendar, HR systems, CRM, document repositories—like digital "hands" | All software systems plus physical tools and spaces | | **Integration level** | Minimal – lives in isolation | Medium – connected to specific systems per use case | Deep – integrated across multiple systems like a real employee | Complete – can use any tool or system given access | ***Governance & Oversight*** | | Chatbot | AI Agent | Digital Employee | Human Employee | | ------------------------------------ | --------------------------------------------- | -------------------------------------- | ----------------------------------------------------------------------------------------- | ---------------------------------------------------------- | | **Who oversees it?** | IT team maintains it | Developers or automation team | Has a designated manager/supervisor, just like human staff | Direct manager plus HR oversight | | **How do you control what it does?** | Basic filters (e.g., "don't discuss pricing") | Permission settings and workflow rules | Role-based access controls, guardrails, approval requirements for sensitive actions | Company policies, performance reviews, professional ethics | | **Can you see what it did?** | Basic logs (mainly for debugging) | Technical logs that engineers read | Business-friendly audit trail: every action, timestamp, reason—understandable by managers | Performance records, but relies on trust and communication | | **What if it makes a mistake?** | User gets frustrated, may contact support | May break a workflow or cause errors | Escalates to human for correction; mistake is logged and reviewed | Takes accountability, learns, corrects course | # List of DE Digital Employees can be specialized for different roles across an organization. Each Digital Employee is built on the same shared foundation, known as the **Digital Employee Core (DE Core)**. This core provides the essential capabilities required for a Digital Employee to operate within an organization. On top of this foundation, each Digital Employee adds role-specific knowledge, workflows, and tools tailored to a particular business function. This page provides an overview of the Digital Employees currently available or under development. *** ### Available Digital Employees The following Digital Employees are currently available or being developed.
Digital EmployeeFunctionTypical ResponsibilitiesCover image
Project Manager DEProject coordinationMeeting documentation, information retrieving/files/Kq9t4MHZXWgLc4h0PZII
Recruitment Officer DETalent acquisitionCV screening, interview coordination, candidate follow-ups/files/d2jVUw2r1Lg5PqadSe5n
HR Compliance AnalystPayroll compliance checkingDetect payroll anomalies, verify statutory compliance, generate reports/files/d2jVUw2r1Lg5PqadSe5n
HR Payroll Officer DEPayroll operationsValidate payroll inputs, coordinate data updates, prepare payroll processing/files/d2jVUw2r1Lg5PqadSe5n
Each Digital Employee operates within a defined role and supports human employees in handling operational work more efficiently. ### Expanding the Digital Workforce New Digital Employees can be created for additional roles across the organization. Examples of future Digital Employees include: * IT Support Digital Employee * DevOps Digital Employee * Customer Support Digital Employee Because all Digital Employees share the same foundation, new roles can be introduced quickly as organizational needs evolve. # DE Core ### Introduction The **Digital Employee Core (DE Core)** is the foundation platform that powers all Digital Employees. Rather than building each Digital Employee from scratch, DE Core provides a shared set of capabilities that allow Digital Employees to operate within an organization. These capabilities include identity, memory, system integrations, and governance controls. By using DE Core, organizations can create new Digital Employees quickly while maintaining consistent security, reliability, and operational standards. *** ### What DE Core Provides DE Core provides the fundamental capabilities required for Digital Employees to function as digital colleagues within a workplace. These capabilities include:
CapabilityDescription
IdentityGives each Digital Employee a recognizable role and presence in the organization
MemoryEnables Digital Employees to retain knowledge and context across interactions
System IntegrationAllows Digital Employees to access enterprise systems and data
Task ExecutionEnables Digital Employees to perform actions using tools and workflows
GovernanceEnsures Digital Employees operate within defined permissions and security policies
These shared capabilities allow organizations to deploy many Digital Employees while maintaining a consistent platform. *** ### Why DE Core Matters By centralizing these capabilities, DE Core allows organizations to: * deploy new Digital Employees faster * maintain consistent governance and security * reuse existing integrations and infrastructure * scale Digital Employees across departments # DE Core Capabilities Your Digital Employee combines the speed and reliability of automation with human-like reasoning and context awareness. It can: * **Identify itself** with a name, email, and job title * **Take action** using integrated tools (email, calendar, documents) * **Think and adapt** by breaking down complex requests, asking for clarification, and learning from memory * **Delegate to specialists** for complex workflows * **Operate safely** under governance controls with human oversight By understanding these core capabilities, you can confidently collaborate with your DE as a trusted member of your team. *** ### 1. Identity: Who Your Digital Employee Is Every Digital Employee has a professional identity, just like a human colleague: #### **Name** Your DE has a recognizable name (e.g., "Claudia"), making it easy to refer to naturally in conversations—just say "Ask Claudia" instead of using technical IDs. #### **Email Address** Each DE has its own corporate email address (e.g., `de.claudia@gdplabs.id`). This allows it to: * Send and receive official work emails * Participate in calendar invitations as an organizer or attendee * Integrate with access control systems #### **Language** Your DE can be configured to communicate in a specific language, ensuring seamless collaboration with global teams. All its messages, reports, and emails will be in the language you set. #### **Job Title** The DE's job title (e.g., "HR Recruiter Assistant") clearly defines its role and helps colleagues understand its responsibilities and limitations. #### **Job Description** A summary of what the DE can do helps you know when to involve it. For example: "Assists with recruitment processes, including CV screening, interview scheduling, and candidate communication." #### **Job Instructions** Behind the scenes, your DE follows explicit operating procedures that define: * What actions it can take independently * When it should ask for help * What information it must never share *** ### 2. Tools: How Your Digital Employee Takes Action DEs use **Tools**—individual integrations with enterprise systems—to perform specific tasks. #### **Single Tool Execution** Your DE can execute discrete actions through integrated APIs, such as: * Sending an email via Gmail * Creating a calendar event via Google Calendar * Querying a database **Example:** When scheduling an interview, the DE uses the Google Calendar tool to check availability, the Gmail tool to send the invitation, and a database tool to retrieve candidate details. **Key Features:** * Tools have clear inputs and expected outputs * Errors are handled gracefully (e.g., timeouts trigger fallback actions) * All tool usage is logged for transparency #### **Pre-Built Tool Integrations** Your DE comes with ready-to-use integrations for common enterprise systems, including: * **Google Drive** – File storage and retrieval * **Google Mail** – Email management * **Google Docs** – Document creation and editing * **Google Calendar** – Event scheduling **Built-in Tools** include specialized utilities like: * GitHub search (commits, issues, pull requests) * PDF/DOCX readers * Web search * Python code execution * Data validation tools **Benefits:** * Saves development time—integrations are pre-built and tested * Uses secure OAuth 2.0 authentication with minimal permissions * Includes documentation and health checks #### **PDF Generation & Email Distribution** Your DE can generate PDFs from Markdown documents and send them via email. **Example:** Creating a report from notes and distributing it to stakeholders. **Capabilities:** * Convert Markdown to PDF while preserving structure * Send emails with PDF attachments * Track delivery status (sent/failed) per recipient *** ### 3. Agents: Delegation to Specialists Your DE can delegate tasks to specialized subordinate agents. #### **Subordinate Agent Delegation** For complex workflows, the main DE delegates to specialists with focused expertise. **Example:** When evaluating a CV, the main DE delegates to `cv_eval_agent` (the specialist), receives the score back, and decides whether to proceed with an interview. **Key Features:** * Structured task handoff with clear input/output * Full audit trail (main DE → specialist → result) * Graceful error handling (timeouts trigger escalation) #### **Configurable Maximum Steps** To prevent infinite loops, the system limits delegation depth (default: 5 levels). **Example:** If Agent A delegates to Agent B, which delegates back to Agent A, the system terminates with a clear error. #### **Minutes of Meeting (MoM) Agent** A specialized agent automatically generates and distributes meeting minutes. **Workflow:** 1. Retrieves meeting summary from Meemo (meeting recordings platform) 2. Extracts key points, decisions, and action items 3. Formats into a standardized MoM document (DOCX/PDF) 4. Distributes via email to all participants **Example:** * **After a weekly team meeting**, the MoM agent generates a PDF titled "MoM\_Weekly\_Team\_Sync\_2025-01-20.pdf" * **Email subject:** "\[MoM] Weekly Team Sync - 2025-01-20" * **Email includes:** Brief summary, action item highlights, and attached PDF **Key Features:** * Applies company branding (logos, formatting) * Handles incomplete data by escalating to supervisor * Logs all actions for compliance *** ### 4. Memory & Cognition: How Your Digital Employee Thinks Your DE doesn't just execute commands—it reasons, learns, and adapts. #### **Request Decomposition** Your DE can break down complex instructions into clear, ordered steps. **Example:** * **You say:** "Schedule a 30-minute interview with the candidate next week and send me confirmation." * **DE breaks it down:** 1. Look up the candidate's calendar 2. Find a 30-minute slot next week 3. Create a calendar event 4. Send calendar invites 5. Send email confirmation #### **Tool & Skill Selection** The DE automatically selects the right tools for each step. **Example:** For arranging a meeting, it uses: * Calendar Lookup tool → Check availability * Event Creation tool → Create the meeting * Email tool → Send confirmation #### **Ambiguity Handling** When a request is unclear, your DE asks for clarification instead of guessing. **Example:** * **You say:** "Schedule a meeting with Alex next week." * **DE responds:** "I found two people named Alex (Alex Tan and Alex Garcia). Which one should I invite? Also, how long should the meeting last?" #### **Escalation** When the DE encounters a problem it can't solve (e.g., tool errors, unresponsive users, security constraints), it escalates to a human supervisor. **Example:** * **Situation:** Urgent meeting request, but you haven't responded to clarification questions for 4 hours. * **DE Action:** Sends a summary to your supervisor: "I could not complete the urgent meeting request due to ambiguity. The user is unresponsive. Please intervene." #### **Knowledge Base Access** Your DE retrieves information from configured documents and policies to provide accurate, fact-based answers. **Example:** * **You ask:** "What is our paternity leave policy?" * **DE responds:** "According to the Employee Handbook (Section 5.3), eligible employees receive 2 weeks of paid paternity leave." **Key Features:** * Cites sources (document name, section, page) * Indicates when information is unavailable * Respects access permissions (you only see what you're authorized to view) #### **Private User Memory** Your DE remembers your personal preferences privately—no one else can access them. **Example:** You tell the DE: "When summarizing my daily emails, always list the ones from the engineering lead first." The DE applies this only to your summaries. #### **Shared Team Memory** Your DE also stores information meant to be shared across your team. **Example:** After a meeting, your manager says: "The Q3 report deadline has moved to October 15th." Later, when another colleague asks, "When is the Q3 report due?", the DE correctly responds with the updated deadline. **Private vs. Shared:** * **Private:** Personal preferences, health details, financial matters * **Shared:** Work updates, team logistics, organizational policies *** ### 5. Governance: Security & Oversight #### **Governance Module** (Details in separate technical documentation) Your DE operates under strict controls: * **Role-Based Access Control (RBAC):** Only accesses tools it's authorized to use * **Guardrails:** Prevents harmful actions (e.g., won't share personal contact information) * **Audit Logs:** Every action is recorded for transparency * **Human-in-the-Loop (HITL):** Escalates critical decisions to supervisors # Use Case Scenarios Unlike domain-specific Digital Employees that solve business problems, **DE Core solves the "Digital Employee creation problem"**—it shifts the challenge from "How do we build an AI agent?" to "What should our Digital Employee do?" This abstraction enables: * **Business stakeholders** to think in terms of "hiring employees" rather than "configuring agents" * **Developers** to focus on role logic rather than infrastructure plumbing * **Organizations** to scale AI adoption without exponential engineering costs * **GDP Labs** to establish a platform business model in the emerging AI workforce market The ideal use cases for DE Core are therefore **meta-use cases**—scenarios where the primary value is enabling other use cases to exist efficiently and consistently. | Title | Scenario | Business Value | | ------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | One-Day Digital Employee Creation | A developer needs to build a custom Digital Employee for their organization's specific workflow (e.g., "Invoice Processing Assistant" or "Customer Onboarding Bot"). Using DE Core, they initialize a project, configure identity and tools, define job instructions, and deploy to production—all within 8 hours instead of weeks of ground-up development. |
  • 80% Development Time Reduction: Focus on business logic, not infrastructure
  • Consistent Architecture: Every DE follows proven patterns for security, memory, and communication
  • Faster Time-to-Value: Organizations see ROI in days, not months
| | Digital Workforce Standardization Across Organizations | A large company wants to deploy 10+ Digital Employees across departments (HR, IT, Finance, Legal) with consistent governance, security, and audit capabilities. DE Core ensures all DEs share the same authentication system (SSO/LDAP integration), audit logging format, and HITL escalation workflows, while allowing role-specific customization through extensions. |
  • Unified Governance: Single compliance framework across all Digital Employees
  • Operational Consistency: IT teams manage all DEs using the same tools and processes
  • Scalable Security: RBAC and guardrails applied uniformly—reduces security risks
| | Digital Employee Platform-as-a-Service | GDP Labs offers "Digital Employee-as-a-Service" to clients. Using DE Core's multi-tenant architecture, they can spin up isolated Digital Employees for different clients (e.g., law firms, accounting firms, startups) with client-specific branding, data isolation, and billing. Each client gets their own "Digital Employee workspace" without GDP Labs rebuilding the foundation. |
  • Platform Business Model: Recurring revenue from DE Core subscriptions Rapid
  • Client Onboarding: New clients get their DE operational in hours
  • Economies of Scale: Single infrastructure supports hundreds of Digital Employees
| | Community-Driven Digital Employee Skills | Developers build custom Skills and Tools for DE Core (e.g., "Salesforce Integration Skill" or "Slack Bot Tool") and publish them to the DE Core marketplace. Other developers can install these extensions (`de-core skills install salesforce-crm`) without writing integration code from scratch, creating a vibrant ecosystem similar to WordPress plugins or VS Code extensions. |
  • Accelerated Innovation: Community contributions expand DE capabilities exponentially
  • Reduced Development Costs: Organizations reuse existing Skills instead of custom development
  • Network Effects: More developers → more Skills → more valuable platform
| | Digital Employee Migration and Modernization | Organizations with existing automation (RPA bots, custom scripts, basic chatbots) want to modernize to AI-powered Digital Employees. DE Core provides migration paths—import existing workflows, upgrade to LLM-based reasoning, and add missing capabilities (memory, HITL, audit logging) without starting over. |
  • Investment Protection: Preserve existing automation logic while gaining AI capabilities
  • Gradual Transformation: Migrate incrementally instead of risky "big bang" replacements
  • Competitive Upgrade Path: Stay current with AI advancements using established framework
| | Enterprise-Grade Digital Employee Governance | Regulated industries (finance, healthcare, legal) need Digital Employees that meet strict compliance requirements. DE Core provides built-in GDPR data handling, SOX-compliant audit trails, HITL approval workflows, and immutable action logs. Organizations can deploy AI automation while satisfying regulatory audits. |
  • Compliance by Design: Regulatory requirements built into the framework—not bolted on
  • Audit Readiness: Complete action trails available for regulatory reviews
  • Risk Mitigation: Guardrails and HITL prevent AI from making unsupervised high-risk decisions
| | Standardized Quality Assurance & Testing Framework | A developer finishes building a new "Procurement DE" responsible for ordering office supplies. Instead of writing custom test scripts or manually clicking through workflows to verify it works, the developer runs the built-in DE Core test suite command (e.g., `de-core test` or `de-core validate`). The framework automatically: **Validates Configuration:** Checks `config.yaml` for structural errors or missing credentials. **Runs Integration Tests:** Simulates tool execution (e.g., mocking the "Purchase Order API" to ensure the DE forms valid requests without actually spending money). **Performs Red-Teaming:** Automatically attempts "prompt injection" attacks to ensure the DE's guardrails hold up against malicious inputs. **Checks Compliance:** Scans the DE's memory logic to ensure GDPR data erasure commands function correctly. |
  • Production Reliability: Ensures no Digital Employee is deployed with broken configuration or logic errors, preventing operational disruptions.
  • Security Assurance: Automated red-teaming and vulnerability scanning (SAST/DAST) catch security gaps before they are exposed to the real world.
  • Reduced QA Overhead: Eliminates the need for manual "chat testing," reducing the Quality Assurance cycle from days to minutes.
  • Standardized Benchmarks: Every DE is held to the same performance standards (e.g., response latency <3s) before being allowed into production.
| # Release Notes Browse DE Core release notes by date. Open an entry to view the full release note. 1. [19 December 2025](/catapa/digital-employee-de/list-of-de/de-core/release-notes/december-19-2025) 2. [5 January 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/january-5-2026) 3. [2 February 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/2-february-2026) 4. [25 February 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/25-february-2026) 5. [9 March 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/25-february-2026-1) 6. [16 March 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/16-march-2026) 7. [2 April 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/2-april-2026) 8. [16 April 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/16-april-2026) 9. [30 April 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/30-april-2026) 10. [6 May 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/6-may-2026) 11. [20 May 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/20-may-2026) 12. [9 June 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/9-june-2026) 13. [17 June 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/17-june-2026) 14. [19 June 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/19-june-2026) 15. [1 July 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/1-july-2026) 16. [3 July 2026](/catapa/digital-employee-de/list-of-de/de-core/release-notes/3-july-2026) # 19 December 2025 #### Overview This release introduces the first set of foundational capabilities in **DE Core**, the framework used to build “Digital Employees” as enterprise-ready autonomous workers. DE Core now supports a standardized employee profile, core reasoning to turn requests into actions, execution through integrated tools, outbound communication, and basic multi-agent delegation—helping teams deliver Digital Employees faster and more consistently. This is an early foundational release of DE Core intended to accelerate building role-specific Digital Employees on top of a standardized base. Additional enterprise governance capabilities (e.g., broader safety, compliance, and oversight mechanisms) will continue to evolve in subsequent releases. #### What’s new **Digital Employee Profile (Identity)** You can now define a Digital Employee with a complete, human-friendly profile so it can be recognized and referenced like a real team member: * **Name** for natural interaction. * **Corporate email address** as its organizational identifier for enterprise workflows. * **Language setting** to ensure the DE understands and responds in the configured language. * **Job title & job description** to clarify scope and support routing/delegation in multi-agent setups. **Operating Guidelines (Job Instructions)** Digital Employees can be configured with explicit **job instructions**—their “employee handbook”—to define: * Allowed vs. prohibited actions. * Delegation rules to specialist agents. * Escalation conditions. * Consistent response templates and boundaries. **Core Reasoning (Cognition)** DE Core now supports baseline reasoning needed for reliable execution: * Breaking down user requests into actionable steps. * Selecting the appropriate available capabilities to complete each step safely and correctly. * Accessing a configured knowledge base to provide fact-based answers (where enabled/configured). **Tools, MCP, and Integration Setup** Digital Employees can now “do work” by executing tools with structured inputs/outputs, including: * Support for executing **single-purpose actions** via tools (with contracts and error handling expectations). * **Pre-built integrations** for common enterprise systems (e.g., email and calendars) to reduce setup time. * **Default model selection:** By default, DE Core uses **GPT 5 mini**, and the user/client can configure a different model if needed. * **Tool allowlisting:** When an MCP provides a set of tools, the user/client can configure **which tools are allowed** for the model to use (so the DE only has access to the permitted subset). **Communication** Adds foundational communication support so the DE can send outbound messages using configured channels as part of completing tasks and workflows. **Delegation to Specialist Agents (Multi-Agent)** A Digital Employee can now delegate tasks to subordinate specialist agents and incorporate their results back into the workflow. This includes: * Delegating structured tasks to a specialist agent. * Receiving results with success/failure status. * Safety limits to prevent runaway or looping delegation chains. # 5 January 2026 #### Overview This release strengthens DE Core’s enterprise readiness with **proactive automation**, **human oversight**, and **auditability**. Digital Employees can now run scheduled work, escalate unresolved issues to a human supervisor, record immutable audit trails, and generate and email PDF reports from Markdown. *** #### What’s new **Scheduled automation (recurring jobs)** Digital Employees can be configured with a scheduler that triggers tasks, instructions, skills, or workflows automatically at predefined times. * Supports **interval-based triggers** (e.g., every N hours/days/weeks). * Supports **calendar-based triggers**, including **monthly** and **annual** schedules. * Scheduled runs enforce the DE’s configured rules and governance context. * Failures (e.g., missing credentials, tool timeouts, upstream API issues) are handled gracefully and can be escalated to the supervisor. * Schedules can be enabled/disabled without removing them. **Escalation to a Human-in-the-Loop (HITL) supervisor** When a Digital Employee encounters a roadblock it cannot resolve (e.g., tool errors, user unresponsiveness on urgent tasks, or out-of-scope complexity), it can escalate via **email** to a designated supervisor. * Escalations include a structured incident summary (requester, requested action, tools/resources used, and reason for escalation). * Escalation conditions are configurable per Digital Employee, including the option to disable escalation entirely. **Immutable audit logging (tamper-evident)** All DE actions can be recorded in an **append-only, tamper-evident** audit trail to support compliance, debugging, and accountability. * Captures key traceability details including prompts/outputs (where applicable), tool/agent invocation order, parameters, and results. * Logs are queryable (e.g., by user, date range, action type, tool, success/failure). * Logs are exportable (e.g., JSON/CSV) and support retention policies (minimum 2 years). **Supervisor assignment (reporting & oversight)** Each Digital Employee can be configured with an assigned supervisor (name and email) who becomes the destination for: * Escalations and error notifications. * HITL approval requests (where enabled in your setup). **PDF report generation from Markdown + email delivery** DE Core can now generate a **PDF from a Markdown input** and send it via email as an attachment. * The Markdown-to-PDF conversion preserves the main content structure. * Email sending returns per-recipient delivery status (**sent/failed**). # 2 February 2026 ### **What's New** #### **1. User x DE Memory** * DE can now store user-specific memory per DE instance. * Memory is isolated: * Not shared across different users. * Not shared across different DEs (even for the same user). * Memory can be added, recalled, and updated directly through regular conversation. * Memory is retained for as long as the DE is active. #### **2. Guardrails** * Introduced both **Input Guardrails** and **Output Guardrails**. * Supported capabilities: * Banned phrase detection * Pattern detection * Topic-based filtering * Prompt injection prevention * Custom validation rules support * :warning: Guardrails can only be configured and used after the DE pipeline has been fully set up. # 25 February 2026 ### **What's New** #### **1. DE Pipeline Architecture** * Redesigned DE from a single-coordinator agent architecture to a **pipeline-based architecture**. * Enables more structured orchestration, clearer stage separation, and better flexibility. #### **2. PII Anonymization Handling** * Adjusted system prompt to ensure anonymized user inputs can still be correctly interpreted and processed by the DE. * Improves robustness when PII masking or anonymization is applied upstream. #### **3. Fix MoM agent name** * Resolved issue where different coordinator agents had MoM sub-agents with the same name. * Prevented incorrect MCP configs for MoM agents caused by them having the same name. #### **4. Fix model config validation** * Fixed model config validation logic to support model names containing '/'. * Prevented incorrect model configs for agents. # 9 March 2026 ### **What's New** #### **1. Refactor DE Core SDK** * Refactor agent config validator * Remove unused and redundant dependency #### # 16 March 2026 ### **What's New** #### **1. Skills Support for v0.0.x** Digital Employee Core v0.0.x now supports skills by passing the parameter during init or using helper methods like `add_skills()` and `remove_skills()`. Support for v0.1.x will be released in the near future. # 2 April 2026 ### **What's New** #### **1. Skills Support for v0.1.x** Digital Employee Core v0.1.x now supports skills via Agent Builder using `add_skills()`. See [documentation](/catapa/developer-documentation/digital-employee-de-pipeline/guides/advanced-examples/skills-configuration). #### **2. PTC Support for v0.0.x** Digital Employee Core v0.0.x now supports PTC by passing `ptc=PTC(enabled=True)` during Agent init. See [documentation](/catapa/developer-documentation/digital-employee/advanced-examples/programmatic-tool-calling-ptc-configuration). Support for v0.1.x will be released in the near future. ### **What's Fixed** #### **1. Local Run with Remote Skills for v0.0.x** Now remote skills will be fetched and run properly during local runs in v0.0.x. # 16 April 2026 ### **What's New** #### **1. Build Verification Tests (BVT) before deployment to AIP** Digital Employee Core now executes BVT before deploying to AIP servers. By default, MCP connections are checked. The BVT class can be extended to add more BVT tests. See [documentation](/catapa/developer-documentation/digital-employee/advanced-examples/build-verification-tests-bvt). # 30 April 2026 ### **What's Improved** #### **1. Model object support** Digital Employee Core now supports [Model object from GLAIP SDK](https://gdplabs.gitbook.io/gl-aip/how-to-guides/language-models-guide#advanced-custom-model-configuration) as an acceptable input for the `model` argument. #### **2. Improved MCP config with AIP for BVT in v0.0.x** The MCP config resolver logic in BVT is now more consistent with how AIP resolves MCP config. BVT now recognizes strings (`id` or `name`), MCP objects, and MCP dictionary as keys in `mcp_configs`. # 6 May 2026 ### **What's Improved** #### **1. Bump `glaip-sdk` version to 0.8.25** Updates glaip-sdk to version >= 0.8.25 for improved compatibility and features. If you are using user token delegation (`gl_connectors_token` in custom tools, for example) in your DE, now you can test the functionality in your local runs. ### **What's Fixed** #### **1. Python Library Vulnerability** Addresses deprecated packages and security vulnerability in DE Core Python library dependencies. If you are using `gllm_plugin`'s `tool_plugin` decorator for your tools, you should remove it. The tools will still run as expected as a LangChain-native tool. Ensure that you wrap the tool with `Tool.from_langchain(...)` before attaching to Digital Employees. # 20 May 2026 ### What's Fixed 1. **\[DE Core Pipeline] Corrected initialization arguments in usage examples** Fixes incorrect init args in DE Core usage examples so they run as expected out of the box. 2. **\[DE Core Pipeline] Fixed pipeline step input mapping** Resolves an issue with the input map configuration for pipeline steps to ensure correct behavior during execution. # 9 June 2026 ### What's New 1. **\[DE Core Pipeline] Programmatic Tool Calling (PTC) support** `DigitalEmployeeAgent` and its builder now support Programmatic Tool Calling (PTC), enabling tools to be invoked programmatically during Digital Employee execution. ### What's Improved 1. **\[DE Core & DE Core Pipeline] Refreshed skills usage examples** Updates the skills usage examples in DE Core for accuracy and clarity. 2. **\[DE Core] Simplified local development setup** The Makefile no longer requires a login step to run, and the coverage command has been corrected — making it easier to get started locally. ### What's Fixed 1. **\[DE Core] Fixed broken examples** Fixes broken or incorrect examples in DE Core. 2. **\[DE Core] Added validation for schema-bearing tool configs** Tool configurations that carry a schema are now validated on attachment, preventing misconfigured tools from being added to Digital Employees silently. # 17 June 2026 ### What's New 1. **\[DE Core Pipeline] Event emitter support in `DigitalEmployee.invoke`** `DigitalEmployee.invoke` now emits events during execution, allowing callers to listen and react to what's happening inside a Digital Employee in real time. # 19 June 2026 ### What's Improved 1. **\[DE Core] Pinned dependency versions for reproducible builds** Dependency versions in DE Core are now explicitly pinned, ensuring consistent and reproducible builds across environments. 2. **\[DE Core Pipeline] Removed unused dependencies** Cleans up unused dependencies from the DE Core v0.1.x branch to reduce package size and eliminate unnecessary overhead. # 1 July 2026 ### What's Improved 1. **\[DE Core] Documented supported `run` and `arun` kwargs** Added documentation for the supported keyword arguments of `run` and `arun`, referencing the GLAIP SDK Agent APIs as the authoritative source. ### What's Fixed 1. **\[DE Core] Moved tool config validation after merging** Schema validation for tool configs now runs on the final merged output of `DigitalEmployee.build_tool_config()`, allowing multiple config loaders to collectively provide a single schema-bearing tool config. # 3 July 2026 ### What's New 1. **\[DE Core Pipeline] GLChat integration input formatter** Added `digital_employee_core.integrations.glchat` with helpers (`extract_content_text`, `extract_message`, `extract_chat_history`) that parse the GLChat Responses API payload shape into the message string and chat history list expected by `DigitalEmployee.invoke`. 2. **\[DE Core Pipeline] Config loaders support in agent builder** Added fluent `add_config_loaders` support to `DigitalEmployeeAgentBuilder`, replaying registered loaders during `build()` while preserving the agent's base-loader-first behavior. # Claudia✨ - DE HR Below is the comparison table between human employee and digital employee, in the context of HR
CapabilityHuman EmployeeDigital Employee
Run payroll calculations✅✅
Check payroll anomalies before payday⚠️ Limited✅
Handle payroll exceptions and complaints✅⚠️ Limited
Update employee master data✅✅
Generate HR letters/documents✅✅
Track probation, contract renewal, and deadlines⚠️ Limited✅
Answer routine HR policy questions✅✅
Handle sensitive employee issues✅❌
Screen CVs against basic requirements⚠️ Limited✅
Schedule interviews and send reminders✅✅
Assess culture fit and motivation✅❌
Sell the role and convince candidates✅❌
Keep a perfect audit trail⚠️ Manual✅
Apply a new HR policy everywhere at once❌✅
Work 24/7 during cut-off or hiring spikes❌✅
Scale instantly during peak HR workload❌✅
Take accountability for final HR decisions✅
Chase missing timesheets / payroll inputs⚠️ Manual✅
Spot duplicate / inconsistent employee data⚠️ Limited✅
Prepare payroll files for bank / vendor / tax submission✅✅
Submit payroll files to bank / coretax✅❌
Monitor approval bottlenecks⚠️ Limited✅
Auto-escalate overdue HR cases✅✅
Generate onboarding / offboarding checklists✅✅
Validate employee documents⚠️ Limited✅
Answer candidate status questions✅✅
Keep candidates warm during long hiring cycles⚠️ Limited✅
Summarize interview notes into a standard format✅✅
Read subtle candidate signals✅❌
Handle sensitive offer negotiation✅❌
Make final exception calls✅❌
Build employee trust in difficult moments✅❌
# Reginald✨ - DE HR Recruiter Generic Reginald✨ is a Digital Employee designed to support HR recruitment operations by managing candidate processing and interview scheduling tasks efficiently. # Candidate Processing Assistant ### Introduction Reginald✨ - DE HR Recruiter Generic - Candidate Processing automates the intake of new candidate applications received via email. Trigger her in GL Chat and she will: * Pull unprocessed application emails from the recruiter inbox * Extract candidate details (name, role, email, WhatsApp number, attachments) * Drop attachments into the right Google Drive folder per role * Add the candidate to the role's Candidate Tracker Sheet * Detect duplicates and skip them * Send a processing summary back to GL Chat She handles up to **50 emails per run**. *** ### Benefits * **Less manual work** — no more downloading CVs, creating folders, or copy-pasting candidate data * **Fewer duplicates** — automatic check against the existing tracker per role * **Organized documents** — every candidate gets a standard folder under the right role * **Single-glance visibility** — every run ends with a Success / Failed / Warnings summary in GL Chat *** ### Quick Start #### What you need before triggering 1. The candidate emails must already be in the recruiter inbox 2. Each email's subject must follow the format `Application - Full Name - Role` — Customizable 3. The role in the subject must match a `Position` in the Hiring & Employee Maintain sheet whose progress is **not** "Done" #### How to trigger In GL Chat, open Reginald✨ - DE HR Recruiter Generic and send: ``` process new email application ``` > 📸 *Screenshot placeholder — GL Chat trigger (to be provided)* #### What happens next 1. Reginald retrieves up to 50 unprocessed application emails 2. For each email, she extracts candidate data and checks for duplicates 3. She creates a Drive folder, uploads attachments, and adds a row to the role's Candidate Tracker Sheet 4. She labels the processed email `DE-Recruiter/Recorded` so it isn't picked up again 5. She replies in GL Chat with a summary #### Sample summary ``` Processed 18 new email applications. Success: 15 candidates Roles: - Product Manager: 7 - Data Analyst: 5 - UX Designer: 3 Failed: 2 candidates - Kevin Tan — Duplicate candidate - Jane Lim — Duplicate candidate Warnings: 1 candidate - John Doe - Product Manager - Missing attachment ``` *** ### Concepts #### Email processing rules
RuleValue
Max emails per run50
Email must containRole name in subject
Email must NOT containLabel DE-Recruiter/Recorded
#### Email subject format (Customizable) ``` Application - Full Name - Role ``` Example: `Application - John Doe - Product Manager` #### Email data extraction
Field in tracker sheetSource
Candidate NameEmail subject
RoleEmail subject
Email AddressEmail sender
WA NumberEmail signature (if available; blank if not)
AttachmentsEmail attachments
#### Role folder mapping
ConditionAction
Role folder existsUse the role folder
Role folder missingUse Uncategorized folder
Role matching is case-insensitive (`product manager` matches `Product Manager`), but extra spaces are **not** normalized — `Product Manager` (double space) will not match and the candidate will be filed under `Uncategorized`. #### Duplicate detection A candidate is flagged as duplicate if **any** of these match an existing row in the role's tracker sheet: * Candidate Name * Email Address * WA Number #### Outcome categories | Category | When it happens | Email labeled? | Folder created? | Tracker updated? | | -------------------------------- | -------------------------------------- | -------------- | --------------- | ---------------- | | **Success** | Everything processed cleanly | ✅ Yes | ✅ Yes | ✅ Yes | | **Warning — Missing Attachment** | Candidate has no attachments | ✅ Yes | ✅ Yes | ✅ Yes | | **Failed — Duplicated** | Candidate matches existing tracker row | ❌ No | ❌ No | ❌ No | | **Failed — Processing issue** | System error mid-process | ❌ No | ❌ No | ❌ No | Failed emails stay unlabeled so the next run can retry them. #### Attachment renaming
If filename contains...Rename to...
CV or ResumeCV - Full Name
PortfolioPortfolio - Full Name
Anything elseKeep original name
The first file containing `CV` or `Resume` is recorded as the candidate's CV link. *** ### Security & Permissions * **Trigger access** — anyone with access to Claudia✨ - DE HR Recruiter Generic in GL Chat can run her * **Inbox source** — the recruiter email address is configured server-side; HR doesn't pass it at trigger time * **Read/write scope** — Claudia✨ reads from Gmail and the Hiring & Employee Maintain sheet; writes to Google Drive (folders/files), Candidate Tracker Sheets, and Gmail labels * **PII** — candidate name, email, and WhatsApp number are stored in Google Drive and Google Sheets; access is governed by the Recruitment folder's sharing settings *** ### Admin & Setup Complete this checklist once before triggering Claudia for the first time. #### Setup checklist **Folder preparation** Sample folder template: [Claudia HR Recruiter Template](https://drive.google.com/drive/u/0/folders/1AiwRjdjjP2Xw7lNrfADiTFihclqwWtgz) * [ ] Create a folder in Google Drive, follow the **required folder structure** described below. * [ ] Inside each role subfolder, create a `Candidates` subfolder * [ ] Create an `Uncategorized` folder at the root of `Recruitment` (required — fallback for roles without a folder) * [ ] Inside `Uncategorized`, create a `Candidates` subfolder * [ ] Share the `Recruitment` folder with the relevant HR Recruiters and with Reginald's service account **Spreadsheet preparation** * [ ] In each role subfolder, create a `Candidate Tracker Sheet - {Role}` Google Sheet * [ ] Set up the required columns in each tracker sheet (see format below) * [ ] Confirm the `Hiring & Employee Maintain` sheet exists and lists the roles Reginald should match (with `Main Progress` ≠ Done for active roles) **Gmail preparation** * \[ ] Create the Gmail label `DE-Recruiter/Recorded` in the recruiter inbox * \[ ] Confirm Claudia has access to read the recruiter inbox and apply labels **Access & permissions** * \[ ] Recruiter email address is configured in Claudia's environment * \[ ] Claudia has Google Drive API access (read/write the Recruitment folder tree) * \[ ] Claudia has Google Sheets API access (read Hiring & Employee Maintain; read/write Candidate Tracker Sheets) * \[ ] Claudia has Gmail API access (read emails, apply labels) **Before each trigger / per candidate** * \[ ] Candidate has emailed the recruiter inbox with subject format `Application - Full Name - Role` * \[ ] The role in the subject matches an active row in Hiring & Employee Maintain *** #### Required folder structure ``` Recruitment/ ├── {Role 1}/ e.g., Product Manager │ ├── Candidates/ │ │ └── {Full Name - Email}/ ⭐ created by Claudia per candidate │ └── Candidate Tracker Sheet - {Role 1} ⭐ updated by Claudia ├── {Role 2}/ e.g., Data Analyst │ ├── Candidates/ │ └── Candidate Tracker Sheet - {Role 2} ├── Uncategorized/ required — fallback when role folder missing │ ├── Candidates/ │ └── Candidate Tracker Sheet - Uncategorized └── Hiring & Employee Maintain ⭐ used by Claudia to validate role names ``` > 📸 *Screenshot placeholder — Recruitment folder structure (to be provided)* Notes: * Reginald creates the per-candidate folders (`{Full Name - Email}/`) and updates the tracker sheets — HR sets up the parent structure * The Hiring & Employee Maintain sheet sits at the root and is shared with the Interview Scheduling flow *** #### Candidate Tracker Sheet — required columns One sheet per role, located in the role's subfolder.
ColumnSource / Value
Candidate NameFrom email subject
SourceHardcoded to Email for this flow (other values: Glints, TIA — out of scope)
StatusHardcoded to Screening for new entries
CV LinkDrive link to the first file containing CV or Resume; blank if none
Folder LinkDrive link to the candidate's folder
Email AddressFrom email sender
WA NumberFrom email signature; blank if not extractable
> 📸 *Screenshot placeholder — Candidate Tracker Sheet example (to be provided)* #### Per-candidate folder naming Format: `Full Name - Email` Example: `John Doe - john@email.com` Location: `Recruitment/{Role}/Candidates/` *** ### Integrations | Service | Purpose | | ---------------------------------------------------- | ------------------------------------------------------------------- | | **GL Chat** | Trigger channel + processing summary | | **Gmail** | Source of application emails; label management for processed emails | | **Google Drive** | Per-candidate folder creation; attachment storage | | **Google Sheets** (Candidate Tracker Sheet per role) | Candidate records — one sheet per role | | **Google Sheets** (Hiring & Employee Maintain) | Source of valid role names | *** ### Troubleshooting / FAQ #### Why was a candidate marked as Failed — Duplicated? The candidate's name, email, or WhatsApp number already exists in the role's Candidate Tracker Sheet. Reginald skips the email entirely — no folder, no tracker row, no label. The email stays in the inbox unlabeled. #### Why was a candidate marked as Warning — Missing Attachment? The application email had no attachments. Reginald still records the candidate (folder + tracker row) and labels the email — the warning is informational so HR knows to follow up for the CV. #### Why didn't Reginald pick up an email? Likely one of: * The subject doesn't contain a recognized role name from Hiring & Employee Maintain * The role's `Main Progress` is set to `Done` (Reginald ignores closed roles) * The email is already labeled `DE-Recruiter/Recorded` * Run hit the 50-email cap — remaining emails will be picked up next run #### What if the role folder doesn't exist? The candidate goes into `Recruitment/Uncategorized/Candidates/`. HR can move them later when the role folder is created. #### What if Reginald crashes mid-run? Emails that weren't successfully processed remain unlabeled and will be retried on the next run. No partial labeling — labels are applied only after a candidate is fully recorded. #### What if a candidate emails twice in the same run? The first email is processed normally. The second one is detected as a duplicate (matching name / email / WA number within the same role's tracker) and lands in **Failed — Duplicated**. #### What's the canonical Candidate Name? The name in the **email subject** is the source of truth — used for the folder name (`Full Name - Email`) and the tracker row. Names that appear differently in the email body or signature are ignored. #### How does Claudia match the role from the email subject? * **Case** — handled automatically (`product manager` matches `Product Manager`) * **Extra spaces** — not normalized; an email with `Product Manager` (double space) won't match and will be filed under `Uncategorized` #### What order are emails processed in? **Oldest first.** When the 50-email cap is hit, the newest emails are deferred to the next run. #### What if the inbox is empty? Claudia returns a clean summary with zero counts. No error. #### What happens to non-CV/non-Portfolio attachments? They're saved to the candidate folder with their original filename — no rename. #### What if a candidate has multiple CVs? Claudia picks the **first** file containing `CV` or `Resume` and records that as the CV Link. Other CV files stay in the folder. #### Can I rerun for a single failed email? Manually remove the `DE-Recruiter/Recorded` label from that email — that's the only way. There's no per-email retrigger from GL Chat. *** ### Release Notes **v1.0.0 — 28 April 2026** * Initial release * GL Chat trigger: `process new email application` * Email retrieval with role-name + label filtering (max 50 per run) * Email parsing: candidate name, role, email, WA number, attachments * Per-candidate Drive folder creation with role-folder fallback to `Uncategorized` * Attachment upload + CV/Resume/Portfolio rename * Candidate Tracker Sheet update per role (hardcoded Source = Email, Status = Screening) * Duplicate detection on Name / Email / WA Number * Gmail label `DE-Recruiter/Recorded` on successful records * Processing summary in GL Chat (Success / Failed / Warnings) Out of scope for this release: CV screening, interview scheduling, candidate communication, offer process, Glints/TIA automation. *** # Interview Scheduling Assistant ### Introduction Reginald✨ - Generic Interview Scheduling Assistant is a digital employee that schedules interviews for you. Trigger her from GL Chat and she will: * Find a panel of interviewers based on the job title * Locate calendar slots where everyone is free * Hold the slots, collect interviewer confirmations, and email the candidate * Send the final Google Calendar invite with a Google Meet link She currently supports two interview stages: * **User Interview** — panel of 2–3 interviewers (parallel format) * **VP Interview** — single interviewer *** ### Benefits * **No ATS needed** — Simply store candidates details on Google Drive * **Less manual work** — no more checking 2–3 calendars one by one * **Fewer double-bookings** — slots are held the moment they're proposed * **Faster candidate response** — candidates receive options instantly via email * **Consistent communication** — every email follows the same template * **No tribal knowledge** — interviewer assignments live in one Google Sheet, not in someone's memory *** ### Quick Start #### What you need before triggering 1. The **job title** — must match a `Position` in the Hiring & Employee Maintain sheet 2. The **candidate email address** 3. The **interview type** — `User` or `VP` 4. The **HR escalation email** — where Reginald✨ sends progress updates and any error alerts #### How to trigger Send a free-form message in GL Chat. Example: > Schedule interview for **Software Engineer**, candidate email: ****, type: **User**. Send any escalation to ****. > 📸 *Screenshot placeholder — GL Chat trigger example (to be provided)* #### What happens next 1. Claudia✨ looks up interviewers in the Hiring & Employee Maintain sheet 2. Claudia✨ finds 2 available 1-hour slots in the next 48 hours 3. Claudia✨ creates 2 tentative calendar holds with Google Meet links 4. Interviewers receive calendar invites and RSVP (Accept / Decline) 5. Once **all** interviewers accept both holds, Claudia✨ emails the candidate with Option 1 and Option 2 6. Candidate replies → Claudia✨ finalizes the chosen slot, deletes the unused one, and **sends one confirmation message in GL Chat** #### Where you'll see updates
ChannelWhat you'll receive
GL ChatOne initial confirmation message when the trigger is processed
Escalation emailAll progress updates, status notifications, and error alerts
> 📸 *Screenshot placeholder — final GL Chat confirmation (to be provided)* *** ### Concepts #### Interview types
TypePanel sizeFormat
User2–3 interviewersParallel (everyone joins the same slot)
VP1 interviewerSingle
#### Tentative hold A calendar event marked as **tentative** that reserves the slot while Claudia✨ waits for confirmation. Holds are deleted automatically if confirmation fails. #### Business hours (Customizable) * Monday–Friday, **09:00–17:00 WIB** (UTC+7) * Weekends are excluded * Each slot is **1 hour** * Search window: **next 48 hours** from the time of trigger #### Confirmation rules * **Interviewers** — ALL must accept BOTH tentative holds on their Google Calendar within **4 hours** * **Candidate** — must reply with `Option 1`, `Option 2`, or propose a new date * **Reschedule limit** — max **2 reschedule loops**; the 3rd reschedule is escalated to HR #### Candidate reply outcomes
ReplyReginald✨'s action
"Option 1" or "Option 2"Finalize the chosen slot
UnclearSend a clarification email
Propose new dateDelete holds and restart slot-finding (same interviewers)
No reply after 48 hoursSend reminder email
No reply after 72 hoursEscalate to HR
*** ### Security & Permissions * **Trigger access** — anyone with access to the Claudia✨ - DE HR Recruiter Generic in CATAPA Chat can trigger Claudia✨ * **Position Master Sheet ownership** — HR maintains it * Claudia✨ only reads/writes data needed to schedule interviews: * **Read** — Hiring & Employee Maintain sheet, interviewer Google Calendars * **Write** — Google Calendar events, Gmail (interview emails to candidate + HR) * Claudia does **not** check the candidate's calendar — only interviewer calendars *** ### Admin & Setup Complete this checklist once before triggering Claudia for the first time. #### Setup checklist **Folder preparation** * [ ] Create folders in Google Drive, follow the **recommended folder structure** below. * [ ] Inside each role subfolder, create a `Candidates` folder and a `Candidate Tracker Sheet - {Position}` spreadsheet * [ ] Create an `Uncategorized` folder at the root of `Recruitment` as a catch-all for unclassified items * [ ] At the root of `Recruitment`, place the `Hiring & Employee Maintain` sheet * [ ] Share the `Recruitment` folder with the relevant HR Recruiters **Spreadsheet preparation** * [ ] Open the Hiring & Employee Maintain sheet in the Recruitment Google Drive folder * [ ] Confirm the sheet contains the columns Report Position, User, and VP * [ ] For every position you'll schedule, fill in the Report Position cell (e.g., Product Manager) * [ ] Fill in the User column with 2–3 interviewer emails, comma-separated * [ ] Fill in the VP column with exactly 1 interviewer email * [ ] Verify all interviewer emails are valid Google Workspace accounts **Access & permissions** * [ ] Claudia✨ has read access to the `Hiring & Employee Maintain` sheet * [ ] Claudia✨ has Google Calendar API access for every interviewer listed * [ ] Claudia✨ has Gmail API access to send emails using Claudia✨'s email account * [ ] HR Recruiters who will trigger Claudia✨ have access to CATAPA Chat and to the DE **Before each trigger** * [ ] Job title to be used matches a `Position` row in the sheet * [ ] Candidate email is correct and active * [ ] Interview type decided — `User` or `VP` * [ ] Escalation email is ready — this is where progress updates and any error alerts will land *** #### Recommended folder structure ``` Recruitment/ ├── {Role 1}/ e.g., Product Manager │ ├── Candidates/ CV files, candidate documents │ └── Candidate Tracker Sheet - {Role 1} ├── {Role 2}/ e.g., SDE │ ├── Candidates/ │ └── Candidate Tracker Sheet - {Role 2} ├── {Role 3}/ e.g., AIE │ └── ... ├── Uncategorized/ required — catch-all for unclassified items └── Hiring & Employee Maintain ⭐ used by Claudia ``` > 📸 *Screenshot placeholder — Recruitment folder structure (to be provided)* Notes: * Claudia✨will read **`Hiring & Employee Maintain`** sheet and **`Candidate Tracker Sheet - {Role}`** sheet. *** #### Hiring & Employee Maintain sheet — required format This sheet (also referred to as the **Position Master Sheet**) is Claudia's source of truth for interviewer assignments. Required columns Claudia reads:
ColumnPurposeExample
PositionJob title — matched against the GL Chat inputProduct Manager
UserEmail(s) of User Interview panel — comma-separated for 2–3 interviewersanita.rahma@your-company.com, dony.aditya@your-company.com
VPEmail of VP Interviewer (single)maria.renatha@your-company.com
Other columns are for HR's own tracking and are **not** used by Claudia. > 📸 *Screenshot placeholder — Hiring & Employee Maintain sheet example (to be provided)* Rules to keep in mind: * For `User` interview: 2–3 interviewer emails per row * For `VP` interview: exactly 1 interviewer email per row * Same interviewer email may appear in both `User` and `VP` — that's fine * Keep the sheet up to date — Claudia always reads the live version *** ### Integrations
ServicePurpose
GL ChatTrigger channel + single final confirmation
Google SheetsSource of interviewer assignments
Google CalendarSlot-finding, tentative holds, confirmed events, RSVP tracking
Google MeetAuto-generated meeting link on confirmed events
GmailCandidate emails (proposal, clarification, reminder) and HR progress + escalation emails
*** ### Troubleshooting / FAQ #### Why did I get an escalation email? Escalation emails are sent for any error condition. Common reasons: * **Invalid input** — wrong email format, missing job title, or interview type ≠ `User`/`VP` * **Job title not found** in the Hiring & Employee Maintain sheet * **Invalid interviewer data** — empty or malformed emails in the sheet * **No availability** — fewer than 2 valid slots in the next 48 business hours * **Interviewer didn't respond** in 4 hours, or someone declined * **Candidate didn't reply** for 72 hours * **3rd reschedule** attempt by the candidate * **API failure** after one retry (Google Calendar or Gmail) #### Why don't I see progress updates in GL Chat? By design — GL Chat shows only the final scheduling confirmation. All in-progress updates and errors go to the escalation email so your chat stays clean. #### What does Claudia do when an interviewer declines? Both holds are deleted immediately and an escalation email is sent. Claudia does **not** auto-pick a replacement — panels are fixed in this version. #### What if only 1 slot is available in the next 48 hours? Claudia escalates with details. She does not proceed with a single slot. #### Can the candidate reply in any format? Claudia looks for `Option 1` or `Option 2` (case-insensitive, partial match). If unclear, she sends a clarification email asking for explicit wording or a new proposed date. #### Can I cancel mid-flow? Not in this version. If you need to stop a flow, contact your DE admin. #### Does Claudia handle WhatsApp or document attachments? No. Candidate communication is **email only**, with no attachments. #### Does Claudia auto-trigger the next interview stage (User → VP)? No. HR must trigger each stage separately. #### What if the same interviewer is listed twice? No problem — Claudia handles duplicates gracefully. #### What if Gmail fails to send? Same pattern as Google Calendar: Claudia retries once. If it still fails, she escalates to HR. *** ### Release Notes **v1.0.0 — 28 April 2026** * Initial release * GL Chat free-form trigger with job title, candidate email, and interview type * Hiring & Employee Maintain sheet integration * Multi-interviewer panel slot-finding (User Interview) * Single-interviewer slot-finding (VP Interview) * Tentative calendar holds with Google Meet links * Interviewer confirmation via Google Calendar RSVP (4-hour timeout) * Candidate email flow with 2 options + reply handling * 48-hour reminder, 72-hour HR escalation * Maximum 2 reschedule loops * Auto-generated Google Meet link on confirmed events * Single GL Chat confirmation; all other updates routed to escalation email --- [Next Page](/catapa/llms-full.txt/1)