AI gave me more meeting notes than I wanted to read. So I started listening.

AI and meeting transcripts have made it much easier to capture what happens at work. For me, that created another problem: I still had to read, sort and make sense of everything we were capturing.

I’ll admit it. I’m not a great fan of reading. I much prefer listening to something or watching a programme when I’m learning.

That becomes awkward in product management, where so much of the job involves absorbing information. Customer conversations, stakeholder meetings, research, requirements and follow-up notes all need attention. Having a transcript doesn’t mean I’ve understood it, connected it to other conversations, or worked out what it means for the product.

Recently, I’ve been using NotebookLM for research. Then I had a fairly obvious thought: why not put meeting transcripts in there as well?

I was already using it to work through external research. Adding stakeholder conversations felt like extending something that was already useful.

Now I’ve got my own work-related podcasts to listen to while walking. It’s opened up a way of revisiting product context that suits me much better than another pile of reading.

Capturing more information created more work

There’s an appeal to recording everything. We don’t have to rely entirely on memory, and we can return to the original conversation when something is unclear.

But the amount of material grows quickly.

A meeting becomes a transcript. The transcript becomes a summary. There are action points, follow-up messages and documents linked from the discussion. Meanwhile, the next meeting has already happened.

As a PM, you still have to work out what matters. Did someone describe a customer problem or suggest a preferred solution? Was that a decision, a request, or an idea nobody committed to? Does this conversation contradict something another stakeholder said?

Those distinctions don’t disappear because a tool has written the notes for you.

For me, the difficult part was having so much information in a format I didn’t particularly want to spend more time working through.

Using NotebookLM beyond external research

NotebookLM’s Audio Overviews generate spoken discussions based on the sources you provide. They are AI-generated summaries, rather than recordings of the original people speaking.

I’d already found NotebookLM useful for research, so bringing meetings into that process made sense to me. Instead of keeping external research in one mental bucket and stakeholder discussions in another, I could revisit related material through the same tool.

For product work, that creates some useful possibilities. A notebook focused on a particular initiative could contain approved meeting transcripts, relevant research and the current product brief. The audio can then help you revisit that context without reading every document from beginning to end.

The qualification is important: putting documents together doesn’t make them equally reliable. An external article, a stakeholder opinion and a customer interview each tell you something different. I wouldn’t want a confident-sounding discussion to blur those differences.

I’ve made something I actually want to listen to

The most useful change has been personal. I can go for a walk and listen to material connected to the work I’m doing.

There’s something appealing about having a podcast that covers your own product context. The discussions are relevant because they come from material you’ve selected, rather than hoping an episode happens to touch on the problem you’re working through.

I’m still engaging with the information. I’ve changed how I take it in.

I don’t think every walk needs to become work, and I’m not suggesting PMs should fill every spare minute with more input. But when I want to revisit a topic, this gives me an option that fits how I prefer to learn.

Give the audio a product-management purpose

For another PM trying this, I’d start with one initiative rather than uploading every meeting into a single notebook. Keep the source names and dates clear, and choose material you’re permitted to use.

Then give the overview a specific purpose. “Summarise these meetings” is a starting point, but a more useful focus might be the decisions you’re preparing to make.

Here’s an example instruction to adapt:

Create an audio overview for a product manager working on this initiative. Focus on the problems discussed, stakeholder priorities, constraints, disagreements and unanswered questions. Distinguish confirmed decisions from suggestions and requests. Keep external research separate from evidence about our own customers. Don’t invent agreement, owners or deadlines where the sources don’t establish them.

NotebookLM supports custom instructions for Audio Overviews, but a prompt like this is guidance, not a guarantee that every distinction will survive the summary.

After listening, the useful next step is to identify what needs checking: a disagreement to clarify, an assumption to test, or a question to take into the next conversation. That keeps the exercise connected to product work rather than becoming another way of consuming information indefinitely.

The podcast can’t become the official record

A polished audio conversation can sound more certain than the source material deserves. Google explicitly warns that Audio Overviews can contain inaccuracies.

For a PM, small changes in meaning can matter. “We could explore this next quarter” mustn’t become “We agreed to deliver it next quarter.” A repeated stakeholder request doesn’t establish that customers need it.

Use the overview to get oriented, then check the original source before relying on an exact commitment, quotation or decision. Listening doesn’t remove the need for some reading, or for asking people to clarify what they meant.

There’s a confidentiality boundary too. Only use meeting material your organisation permits in the approved account and tool. Check recording consent, data-handling rules and access permissions, remove unnecessary sensitive information, and treat the generated audio as confidential if its sources are confidential. Don’t assume a personal account is appropriate for work transcripts.

A more manageable way to revisit product context

This has helped me find a use for meeting transcripts beyond leaving them in a folder for later.

I already had a way to capture the conversations. NotebookLM has given me another way to engage with them, alongside the research I was already doing.

For PMs who enjoy reading, a written brief may still be the best option. For me, being able to put my headphones on and revisit the material while walking makes it easier to spend time with it. I can come back to the documents when I need to check a detail, with a clearer idea of what I’m looking for.

---

Tool references: Google’s introduction to NotebookLM Audio Overviews and Audio Overview guidance, customisation and limitations.