Description

The Problem: Our current documentation (be it technical specs, onboarding guides, or product manuals) is often static, difficult to search, and overwhelming for new users. Finding a specific answer usually requires reading through pages of text.

The Solution: This project aims to test Google’s NotebookLM as a "Knowledge Engine" for our internal documentation. By ingesting our existing PDFs, Google Docs, and Markdown files into NotebookLM, we want to create a semantic search interface that answers questions based only on our source material (reducing hallucinations).

We will also experiment with the "Audio Overview" feature to see if we can turn dry technical documentation into engaging "podcasts" for faster team onboarding.

You do not need to be an AI engineer to join this team! We are looking for:

Content Curators (PMs/Writers): People who know where our best (and worst) documentation lives to feed the model.

Prompt Testers (QA/Devs): People to try and "break" the model by asking complex edge-case questions to see how well the citations work.

Creatives: To assess the quality of the AI-generated audio summaries and determine if they are useful for training.

Frontend/Design (Optional): To mock up how a "Chat with Docs" button might look in our existing UI.

Goals

The Pilot: Create at least 3 distinct "Notebooks" based on our data (e.g., one for "Engineering Onboarding," one for "API Documentation," and one for "HR Benefits").

The "Podcast" Test: Generate Audio Overviews for our most complex documentation to see if audio learning is a viable alternative for knowledge transfer.

The Benchmark: Run a "Search vs. Ask" test. Compare how long it takes to find an answer using our current search bar vs. asking NotebookLM.

The Verdict: Deliver a one-page recommendation on whether this tool should be adopted by the wider team or if we should build a custom RAG solution.

Resources

NotebookLM Tool: notebooklm.google.com

Source Data Folder: TBD

Project Kanban/Tracker: TBD