Five Things I Actually Did with AI & One I Didn't
In post, I wrote about how I started 2025. Curious, under-skilled, and very aware that reading about AI is not the same thing as using it.
Now, let’s talk about what came next.
Not predictions. Not frameworks in the abstract. But five things I actually built or practiced and one line I intentionally didn’t cross.
Each came with its own lessons.
I Built Custom GPTs (and Then Built One to Test the First)
One of the first real experiments I ran was building custom GPTs.
One acts as an advisor for new managers. It is focused on practical guidance, not theory.
The second exists purely to stress-test the first by generating edge cases, uncomfortable scenarios, and “what would you do if…” situations.
What I learned: AI is powerful when you narrow the role. The more specific the context, the more useful and safer the output. I also learned that testing your AI matters as much as building it. If you wouldn’t QA a human advisor, you shouldn’t blindly trust an AI one.
Testing was critical. That thing went off the rails a ton! I couldn't believe the advice it gave. It was willing to make up fake company policies and give legal advice.
I sent it to people I trust and asked them to try to break it. They delivered. I iterated. The tool got better. Most importantly, I learned, relearned and unlearned.
I Built a Small App to Test AI Fluency
Using Replit, I built a simple app to assess AI fluency at three levels:
Individuals
Teams
Companies
The output isn’t a score for scoring’s sake. It generates recommendations based on a framework I’ve been developing—focused on confidence, workflow impact, and decision-making, not tool hoarding.
What I learned: It was way easier to get started than I could imagine.
Being able to share in plain language what I wanted it to go was remarkable.
It still took iteration, sharing with others for feedback and being humble.
Fluency is about how people think, not how many tools they’ve touched. And building even a rough app forces clarity in a way slides never will.
I Automated the Admin, Not the Relationship
I used Zapier to connect my Notion-based networking CRM to my calendar.
The result:
Automated nudges for follow-ups and outreach
Far less time spent remembering who I owed a note to
More energy for real conversations
What I learned: This is where automation shines: removing cognitive load, not replacing intent.
The Thing I Didn’t Do
I did not automate my outreach messages.
I wrote every note myself.
The nudge was automated. The words were not.
Before sending anything, I paused to make sure there was a genuine reason for the connection something human behind it. The nudge was a reminder but I didn’t then reach out just because it told me too.
Efficiency without sincerity isn’t connection. It’s noise.
I Listened to People “In the Arena”
I didn’t learn the most from newsletters or hype threads. I learned from people doing the work. Well, I did learn from newsletters from people actually doing the work. :)
One-on-one catch-ups
Conferences and small events
Slack and Circle communities
Thoughtful voices on LinkedIn who shared what broke, not just what worked
What I learned: The most honest conversations about AI aren’t happening on big stages. They’re happening in small rooms, side chats, and comment threads where people admit uncertainty.
I Wrote Down My Beliefs and Guardrails
As the experiments stacked up, I realized something else was happening.
I wasn’t just learning tools. I was clarifying my personal ethics in the face of new technology.
So I wrote down my beliefs and guardrails. What I will use AI for, where I won’t, and why.
Not as a policy. As a compass.
What I learned: If you don’t articulate your guardrails, the technology will quietly set them for you.
The Throughline
Across all of this, one pattern kept showing up:
The best uses of AI gave me time back, clarity, or better questions. The worst temptations were about speed without thought.
I’m still experimenting. Still changing my mind. Still very aware that the ground keeps shifting.
But these five things and the one I didn’t do helped me stay anchored while moving forward.
Part 3 will be about what I got wrong, what surprised me, and what I’d do differently if I were starting today.
If you’re in the middle of your own experimentation, you’re not behind. You’re just in the work.