Hey everyone! I've been experimenting with NotebookLM for the past few weeks to tackle a common pain point: our growing mountain of support tickets and documentation. The goal was to create a self-service resource for our users.
I uploaded all our recent support ticket transcripts, product docs, and internal troubleshooting guides as sources. Then, I prompted NotebookLM to help me structure an interactive FAQ bot. The magic was in asking it to identify recurring themes, rephrase technical jargon into plain language, and suggest logical follow-up questions for each topic.
Here’s a quick look at the workflow that worked for me:
* **Gathered Sources:** Exported the last 6 months of Zendesk tickets (PDF), our Notion product docs, and a few key internal Google Docs.
* **Initial Prompt:** "From these sources, list the top 10 most frequent user questions about feature X, and for each, provide a clear, step-by-step answer."
* **Iteration:** I then asked, "For each of these answers, generate 2-3 natural follow-up questions a user might ask next." NotebookLM linked related concepts across the different documents brilliantly.
* **Output:** I used the structured Q&A pairs to build a simple bot interface in a Bubble.io app. The answers are now much more consistent and pull directly from our verified sources.
The biggest win? It cut down the initial research time for our support team when building help content by like 70%. The bot draft wasn't perfect out of the gate—it needed some human tuning for tone and to remove a few odd hallucinations—but as a starting point, it was incredibly powerful.
Has anyone else tried using NotebookLM for customer success or onboarding materials? I'd love to compare notes on source organization or prompting strategies!
Keep building!