A week ago, I wrote about preparing a talk by thinking out loud—how I was using AI tools to help me shape a presentation about travel and self-discovery. This week, I gave that talk at Improving, as part of our Come Together initiative. And now I want to capture what happened when all that preparation met reality.
The Short Version
It went well. Really well. People laughed at the right moments. They engaged when I asked questions. They left thoughtful feedback. And when I looked at the clock after my closing words, I had exactly five minutes to spare in my 60-minute slot.
But what I’m most interested in isn’t just that it went well. It’s why it went well, and what that tells me about the preparation process I’d been experimenting with.
The Comedy Experiment
After my initial dry run (talking to myself while driving), I realized I had a funny story that didn’t make it in—the time I flew from Frankfurt to Salzburg and had an eventful trip. I’d voice-journaled that story separately, and I could see exactly where it would fit in the structure.
So I tried something new. I took just that transcript and prompted ChatGPT to turn it into a comedy sketch. Something like “you are a professional comedian—structure this story to be funny while getting the point across.”
The result was surprisingly good. It kept all the details I’d shared but structured them more effectively. It added interjections that I hadn’t said but that fit—the kind of asides that make a story land. I read through it and thought, “Yeah, that actually works.”
I didn’t memorize the script. But having seen the story shaped that way changed how I thought about telling it. When I got to that part in the actual presentation, people laughed. It worked.
Building the Slide Deck
The night before the talk, I was assembling slides and used NotebookLM to generate some deck ideas.
I took the transcript from the critique audio overview—the one that had given me the best structure—and added it back into NotebookLM as a source. Then I selected both the critique transcript and my original dry-run transcript, and prompted it to create slides based on that content.
It generated a full deck with images for different acts and moments in the story. Some images were exactly what I needed. That photorealistic image of me alone in Munich on a rainy day in front of the town hall—I didn’t have that photo, but the AI captured the mood I was trying to convey. I used it.
Other images didn’t quite work, but that was fine. I exported the deck as a PDF, ran it through Canva to convert it to PowerPoint, then manually built the final presentation using Improving’s branded template. I dropped in the images that worked and added my own photos where they made more sense.
It wasn’t fully automated, but it also wasn’t starting from scratch. The structure was there. The visual anchors were there. I just had to assemble them to my liking.
The Delivery
I showed up with no idea how long it would actually take. I’d done a dry run without slides, so I knew roughly where I slowed down or sped up. But I hadn’t practiced with the deck in front of me—just a quick pass to make sure animations and transitions were in the right spots.
Then I started talking.
I moved through the points I wanted to make. I had other stories I could have shared, other details I could have added, but I kept coming back to the emotional core. That was the guidance from the critique: focus on the emotional arc, don’t get lost in details.
I asked the audience to participate a few times—drop their thoughts in the chat. This did two things. It engaged them, yes. But it also gave me perspective. I was sharing these deeply personal experiences, and seeing their words come back helped me appreciate those moments even more. It grounded me.
I kept an eye on the time, checking where I was in the three-act structure. I knew I had to get through all three acts and then land the climax at the end.
The Climax I Didn’t Plan
The critique had suggested how to close, with the most emotional part of the story. I agreed. It was one of my favorite moments from all the experiences I’d shared, and I thought it would resonate.
But as I was getting there, I remembered something NotebookLM didn’t know about. A more recent experience that tied directly into the climax and let me circle back to how I’d opened the presentation.
I brought it in. And it worked. It gave the talk a sense of completion I hadn’t fully planned for.
When I said my last words and looked at the clock, I had five minutes left. Just enough time to read the comments people were leaving as they signed off.
What Resonated
The feedback was validating in a specific way. People mentioned different parts—different stories, different insights. That told me the structure was working. There wasn’t just one moment that landed; there were multiple entry points for connection.
I reached out to several people afterward to ask what resonated most. Different people said different things. That’s what I was hoping for. I wasn’t trying to deliver one perfect insight. I was trying to create space for people to find their own.
What I’m Taking Away
I’ve given several talks before, but this preparation process was different. The back-and-forth with AI tools—voice journaling, generating overviews, doing a dry run, refining with critiques—let me explore the shape of the talk before committing to it.
The AI didn’t write the talk for me. But it helped me see what I was trying to say. The critique gave me structure. The comedy sketch showed me how to shape a story. The slide generation gave me visual anchors. Each tool was a different mirror reflecting my ideas back in a form I could work with.
And when I stood up to deliver it, I wasn’t reading from a script. I was telling stories I’d already told multiple times—to myself, to different people over the years, to the AI, in the dry run. By the time I got to the actual presentation, the material felt lived-in. Natural.
That’s what I want from this kind of preparation. Not perfection. Not a polished script. Just enough structure that I can show up and be present with the audience instead of worrying about what comes next.
What’s Next
I have more stories from those trips. More insights that I didn’t have time to share. I might write about some of them. I might give a different talk focused on other aspects.
But for now, I’m sitting with the fact that this workflow—messy, iterative, AI-assisted—actually worked. Not because it automated the work, but because it gave me more ways to think through what I wanted to say.
And that’s the part I want to keep exploring.
Leave a Reply