How to build an Employee Learning and Development Agent

Workflow Overview

1. User Input

• Node: Chat Input

• Purpose: The user starts by typing a question or challenge related to leadership, character, or management (e.g., “Struggling to lead certain members of my team. Any sources?”).

2. Knowledge Retrieval

• Node: Great Managers Playbook (Knowledge Base)

  • Purpose: Searches a curated company playbook PDF for relevant advice, frameworks, and best practices for managers.

• Node: Module 1 (Knowledge Base)

  • Purpose: Searches a second company document (“The Elements of Great Managing”) for additional research-backed management strategies.

Both knowledge base nodes receive the user’s question and return the most relevant excerpts from their respective documents.

3. AI Leadership Coach

• Node: Leadership Coach LLM

• Purpose:

  • Receives the user’s question and the relevant content from both knowledge bases.

  • Uses a specialized prompt to:

    • Only answer using company-approved documents.

    • Always cite the source (doc + section/page).

    • Refuse to answer if the info isn’t in the docs.

    • Keep answers concise and positive.

    • Suggest HR if confidence is low.

    • Embed a helpful YouTube video if the question is about leading people.

• How it works:

  • The AI combines the user’s question and the retrieved knowledge, then generates a short, actionable, and well-cited response.

4. Output to User

• Node: Output

• Purpose:

  • Displays the AI’s answer (with citations and, if relevant, a video) directly to the user.

5. Conversation Logging (Optional/Advanced)

• Node: Recording LLM

• Purpose:

  • Takes both the user’s question and the AI’s response, formats them as a JSON object for record-keeping or analytics.

  • (In your setup, this node is ready to send the conversation to a Google Sheet, but the actual write action is not shown in the current flow.)

How the Data Flows

1. User types a question →

2. Question is sent to both knowledge bases →

3. Relevant excerpts are sent to the Leadership Coach LLM →

4. LLM generates a concise, cited answer →

5. Answer is shown to the user

6. (Optionally, the conversation is formatted for logging/analytics.)

Key Features

• Citations: Every answer is backed by company documentation.

• Guardrails: No personal, legal, or medical advice; only company-approved info.

• Video Support: For people-management questions, a helpful video is embedded.

• Logging Ready: The flow is set up to log Q&A pairs for future review or analytics.

In summary:
This workflow is a robust, company-compliant leadership coaching assistant. It ensures that all advice is sourced from your internal documents, is concise, and is ready for both user-facing support and internal analytics.