A year of building with AI put me on three stages. I liked the message more than the microphone.
TL;DR: In July 2025 I started building with AI t...
For further actions, you may consider blocking this person and/or reporting abuse
The table about AI's role seems worth moving near the start of the talk, before the pilot results. 'AI helped write the script' and 'documents went to an external service' answer different questions for that audience, and the table makes both visible. I'd add one plain-language data-flow arrow per pilot so listeners needn't infer where a document went. Did the questions in Riverside change once you explained that distinction, or did the concern remain?
That's a good idea. I think I unintentionally left out so more technical stops because I was concerned the audience would think I had too much tech in it. lol The concern did remain because I didn't explain it clearly enough. I only really thought about what they were asking after and how I could have explained it better. Plus since our courts are different sizes, mine is much smaller, I am able to have more flexibility to experiment. Because their environment is more restrictive they couldn't reconcile that.
The difference in court size sounds like a useful part of the explanation itself: your room to experiment is a condition of the project, and their restrictions are conditions of adopting it. A short 'what would need approval in a larger court?' slide might help them separate an interesting prototype from something they could deploy tomorrow. That also gives their technical questions a practical destination instead of treating them as a request for more implementation detail. Were they mainly asking where the data went, or who could authorize using it?
This would absolutely terrify me! 😂 I can happily spend hours building something, experimenting, figuring out what works and what doesn't, but the thought of standing in front of a room full of people talking about it? That's a whole different challenge.
And I love what you said about the message mattering more than your comfort. That's a pretty powerful reason to keep showing up. Also, I appreciate your approach to the AI pilots, particularly recognizing when the technology wasn't actually the right solution. Sometimes figuring out what not to automate is just as valuable as building something that works.
Really enjoyed reading this. Especially the part about walking off stage and feeling relief rather than accomplishment. I think there's something really refreshing about admitting that even after doing something successfully three times, you still don't particularly enjoy it. I admire your willingness to keep putting yourself out there.
@sizzlebop Thanks so much for reading! And for the record I was terrified. lol I think the biggest lesson I learned was that it was okay to not move forward if the pilot didn't work. But that was scary in itself since I was given the flexibility to try something new. That was difficult. And now thinking back worth it because sometimes the failure gave us a new solution we hadn't considered before.
I really like that perspective!.I think it's easy to get so invested in making something work that we forget stepping away from an idea can be just as valuable. Especially with AI, where there's sometimes this pressure to find a way to make it fit even when it isn't the best solution.
And I love that the failed pilots sometimes led to entirely new ideas. That's one of my favorite things about experimenting with technology. Even when something doesn't work out the way you hoped, you can still come away with something useful.
Also, I'm glad to hear you were terrified too. 😂 Makes me feel a little better about the fact that public speaking scares the hell out of me!
I think there's a difference between getting comfortable with the material and getting comfortable being the person presenting it. You can know your subject inside out and still hate having everyone's attention on you.
One thing I'd find useful is using AI to rehearse the parts of a technical talk that are hardest to predict. Say you're presenting an AI-assisted workflow to an engineering team, and someone asks what happens when the model produces an invalid migration or silently drops data. You can practice answering those questions without relying on a perfectly rehearsed script, then try again with a skeptical audience or a live demo that doesn't go as planned.
There's research on virtual audiences and exposure-based practice for public speaking, too, so the idea of rehearsing in increasingly challenging conditions isn't just a productivity trick.
But I like that your story doesn't end with you suddenly loving public speaking. Sometimes the win is being able to do something worthwhile without needing to enjoy every part of it.
Hi @sinarezaei Yes, I love this observation! "I think there's a difference between getting comfortable with the material and getting comfortable being the person presenting it." And it does feel like a win because I was able to overcome my nervousness and feeling of being an imposter in this space. We'll see if I get asked back to present again.
That’s a solid win, especially because the confidence came from actually doing the work, not just preparing for it. I think the next interesting step is getting comfortable with the questions you can’t script. If someone asks why a pilot stopped at 80% accuracy, or how you know a model didn’t miss something important in a court record, being able to explain the failure criteria and why you decided not to proceed can be more valuable than delivering the slides perfectly. That kind of technical honesty builds trust, even when the answer is that the tool wasn’t good enough. Hope you get asked back!
"I trusted the work, even when I didn't trust myself to present it." That line is the whole post for me. And the second talk being harder makes total sense once you see how the brain handles this. Your subconscious decides what you notice through a filter called the RAS. After Albuquerque it didn't file the stage under "safe." It filed it under "now there's a bar." So in LA it went hunting for every reason the stakes were higher. Biggest court in the country, your court's name on the badge, hosts who were excited to have you. Same talk, way more threat data.
The relief instead of accomplishment is the tell. It means the alarm ran the whole time and you got through on willpower, which honestly is impressive. But you don't have to love speaking to stop dreading it. The ease you have when you write is already yours. It just hasn't been allowed into the room yet. When you get to the compliments on your list, notice the second right before you open your mouth. That's the same alarm, just smaller, and every time you go anyway the filter gets new data. Does not matter how it lands. The action is the win. Also, jokes landing with a room full of finance officers is a real win.
Great insight on why the 2nd talk was more difficult. At first I didn't understand why, I just felt significantly more pressure. There was definitely more threat data!
I think noticing more, being present more and adding complementing a stranger will help my nervous system not be on red alert as much. Here's to hoping!
That pressure you felt was the threat data, so now you've got a name for it. Noticing and being present is a big part of it. The alarm usually gets loudest when you're stuck in your head, so getting back into the room calms it down. And the compliments give it proof from somewhere new. Let me know how the first one goes.
cool!
Thanks!
np!
I like the contrast here: AI opened doors you didn't expect, but it didn't magically make public speaking enjoyable. Sometimes growth means doing something you're still uncomfortable with, not necessarily learning to love it.
This is such an honest and useful write-up, and the line that stays with me is that you liked the message more than the microphone. Separating “enjoying something” from “finding it worth doing” is a distinction more people should make, and it explains why you’d still say yes to a fourth invitation. I loved that both court pilots stopped by design: setting a 90% accuracy bar before starting, landing around 80%, and concluding the answer was a form with less handwriting rather than better technology is exactly the judgment courts want from frontline staff. Hung Jury is my favorite, since discovering over one weekend that the existing Excel workbook was already the best tool is the kind of dead end worth finding early. The “what I’d change” section is a model of good reflection, and the table separating generative AI, AI-assisted code, and machine learning reading redacted sample forms is the clarity nervous audiences need. Also, listening to your own talk as an MP3 in the car is a great rehearsal trick. The approval-first, sample-data-only, human-reviewed, stopping-point-written-down framework makes “frontline staff can lead AI innovation” feel like something others can copy. Congratulations on three stages, and on the laughs from a room full of finance officers.