Reimagining Workplace Learning: Building a Multi-Agent AI Learning Partner - Jen Shannon

The Problem with Learning at Work

In most organizations, learning systems are disconnected from day-to-day workflows. LMS courses sit in silos. Internal documentation is scattered or outdated. New hires are onboarded with checklists instead of conversations. And development opportunities? Often one-size-fits-all, and rarely personalized.

Meanwhile, the pace of work accelerates. Roles evolve quickly. Employees across departments (from engineers to HR professionals) need immediate access to relevant, reliable knowledge to do their jobs better.

So what if learning wasn’t a destination, but an experience embedded directly into your workflow?

My Capstone: A Multi-Agent Learning Assistant for the Flow of Work

For my Gen AI Capstone, I prototyped a multi-agent learning assistant designed to support employees across all departments. This assistant isn’t just a chatbot it acts more like an internal learning partner. It understands your role, connects you to internal knowledge, supplements answers with external resources, and offers tailored development suggestions.

The project uses:

Retrieval-Augmented Generation (RAG) for document understanding
Grounding via Google Search for external enrichment
Embeddings + Vector Database (ChromaDB) for semantic search
Multi-agent orchestration to manage conversational flow

How It Works

The assistant is made up of several specialized agents:

Internal Docs Agent: Uses RAG to retrieve answers from internal documentation (Python docs used as a proxy)
External Search Agent: Uses Gemini + Google Search grounding to find trusted tutorials and articles
Career Agent: Suggests skill growth paths based on your role and query
Manager Agent: Orchestrates the conversation, logs the session, and adapts the flow

All wrapped in a modular, conversational interface that simulates realistic employee interactions.

Behind the Scenes: Orchestrating the Agents

The ManagerAgent is the conductor of the system. It routes each user query through the right agents (internal documentation, external search, and career guidance) then synthesizes a learning recommendation.

Here’s a trimmed version of the orchestration logic:

def handle_query(self, user_query, role=None, demo_mode=True):
    print(f"👤 User: {user_query}")
    self.session_history.append({"User Query": user_query})

    # Step 1: Internal Docs
    internal_answer = self.internal_agent.answer_with_reference(user_query)
    self.session_history.append({"Internal Knowledge": internal_answer})

    # Step 2: External Resources
    if self._needs_external_search(internal_answer):
        external_answer = self.external_agent.search(user_query)
        self.session_history.append({"External Resources": external_answer})

    # Step 3: Career Guidance
    if demo_mode:
        self.user_role = role or "Data Scientist"
    career_answer = self.career_agent.suggest_skills(self.user_role, user_query)
    self.session_history.append({"Career Development": career_answer})

    # Step 4: Recommendation + Summary
    self._generate_learning_recommendation(user_query)
    self._display_session_summary()

This logic ensures a dynamic, context-aware experience that adapts to the user’s role and the nature of their question, just like a smart internal advisor would.

Example Scenarios

Demo 1: Software Engineer Onboarding A new engineer asks, “When should I use list comprehension instead of a for loop?”

The Internal Agent explains performance and readability benefits based on Python docs
The External Agent pulls relevant tutorials and benchmarks
The Career Agent suggests further topics to master: time complexity, functional programming, and pandas optimization

Demo 2: Product Manager Exploring AI A PM asks, “What is LangChain and how could it help our product?”

Internal Agent finds no results (correctly)
External Agent pulls recent LangChain use cases and official docs
Career Agent suggests learning prompt chaining and LLM integration workflows

Demo 3 (Concept): HR Specialist Creating an Onboarding Plan

The agent retrieves internal onboarding docs (future extension)
Enriches them with external HR best practices
Recommends professional development in knowledge management

Takeaways

This project started as a tech demo and evolved into a practical vision: what if every employee had a personalized AI learning assistant built into the systems they already use?

With modular design and thoughtful orchestration, this prototype shows how we can:

Deliver contextual learning moments
Personalize career development
Reduce onboarding friction
Turn static docs into dynamic knowledge systems

What’s Next

This project lays the groundwork for a production-ready internal assistant:

Connect to Confluence, Google Drive, or Notion
Persist user preferences and history
Integrate with internal LMS catalogs
Extend to customer enablement and partner training

The future of learning at work isn’t a separate portal. It’s an intelligent system that meets you where you are and grows with you.

Explore More

Thanks to Google, Kaggle, and the GenAI Intensive team for an incredible course and opportunity to bring this vision to life.