The Transcript Tells You More Than You Asked For
- Amy Westlake
- Jun 23
- 3 min read
Situation
For a long time, I used meeting transcripts the obvious way.
What was decided. What the action items were. Who said what about the timeline.
Facts. Record. Evidence.
I wasn't thinking about what else was in there. What the words revealed about how people actually think — how they protect their territory, signal alignment, hedge before they push back. That layer of data was sitting in every transcript I ever exported. I just wasn't reading for it.
The AI Move
At some point I had a stakeholder I kept misjudging.
Not badly. Just... off. The tone I'd land on was slightly wrong. Too direct, or not direct enough. Too much context, or not enough. I'd send something and feel the friction in the response.
I had months of transcripts in NotebookLM. I decided to try something open-ended — no template, no categories. I just asked: *What can you tell me about how this person communicates and what they respond to?*
What came back was more specific than I'd expected. Communication style. How they signal disagreement (almost never directly). What they need before they can move on a decision. The things that make them dig in.
And at the bottom: a Dos and Don'ts list.
I hadn't asked for that. It appeared anyway — tight, quick-reference, immediately useful. *Lead with the outcome before the context. Name the risk early. Don't soften it. Don't summarize what they already know.*
I saved it to a Google Doc. Added it as a source in NotebookLM. Now I load the profile any time I'm drafting something to this person — the message that comes back isn't generic, it's calibrated — or pull up the Dos and Don'ts before I walk into a meeting with them.
The Shift
I have profiles on most of the stakeholders I work with regularly now.
Each one started the same way: a few transcripts, one open-ended question. Each one gets updated after any critical meeting — a reaction I didn't expect, a pattern that showed up again, something I want to remember going into the next conversation.
What changed wasn't just my communication. It was what I was noticing in meetings.
Once you've built a profile for someone, you start reading their behavior differently in real time. You notice when they hedge — which is different from when they're genuinely uncertain. You notice when they're signaling resistance before it becomes explicit. You get faster at recognizing the pattern because you've already named it.
That's the shift that surprised me. I wasn't just communicating better. I was paying different attention.
The Pattern
Every meeting generates behavioral data about the people in it.
Most of that data lives nowhere. In your gut. In half-formed impressions that don't survive the week. In the vague sense that something works with this person and something else doesn't.
A transcript is a behavioral dataset. Not a record of what was said — a record of how people communicate when they think they're just talking. How they frame disagreement. How they protect territory. How they signal that something matters to them more than they're saying directly.
AI reads that layer without the social pressure of being in the room. It's not managing the conversation while it's analyzing it. It surfaces what you were too busy to catch.
The profile is what you do with that. It makes the implicit explicit — and it compounds. The more you use it, the more you update it, the more accurate it becomes.
The Implication
Pick one stakeholder. Someone you communicate with regularly but feel like you're still calibrating.
Find two or three transcripts from meetings with them. Ask your AI tool: *Based on these conversations, what can you tell me about how this person communicates and what they respond to?*
Read what comes back. Save it somewhere you'll actually find it.
The next time you draft something to them, load the profile and ask for a message calibrated to this person. Notice whether it feels different to write — and whether the response feels different to receive.
Update it after the next significant interaction. Then the one after that.
What I'm Testing Next
Individual profiles work. I'm starting to wonder what happens when you apply the same approach to a group.
Most of my high-stakes communication isn't one-on-one. It's a room — a recurring team, a steering committee, a cross-functional group with its own dynamics. Individual profiles tell you how to reach each person. They don't tell you how the room works.
I'm building group profiles now. Same open-ended prompt, applied to the team as a whole: what does this group respond to? Where does alignment typically break down? What framing tends to move them?
It's early. But I suspect the group profile will change how I show up in rooms the same way the individual profile changed how I show up in conversations.
Comments