Welcome to Prompt, Tinker, Innovate—my AI playground. Each edition usually gives you a hands-on prompt to test how AI can sharpen your thinking, streamline your process, and improve your creative work.
But this week is a little different.
There's no prompt to copy or tool to try. Instead, the experiment is a simple behavior change—and it may be one of the most effective ways to help your team actually adopt AI in their day-to-day work.
This week's playground: Share the chat, not just the finished work
This one is for managers and AI champions.
You may already understand how much AI can change the way you work. But you're watching smart colleagues stay stuck at the starting line, unsure how to begin.
Here's the setup: You spent 40 minutes working with an AI tool to build something good—a deck, customer email, analysis, or plan. Then you sent your team the finished file.
They learned nothing from it.
The deliverable shows the destination. The thread shows the route: the two attempts that went nowhere, the moment you realized you were asking the wrong question, and the correction that finally worked.
Your team will hit some of those same walls next week. You already have the map. You just didn't send it.
The finished work hides the lesson
Most people on your team who aren't using AI aren't stuck because they lack technical skills. They're stuck on two questions nobody says out loud:
Am I allowed to use this?
Am I doing it wrong?
A shared thread answers both in about ten seconds. It signals that AI is part of normal, accepted work. It also shows what a real session looks like—which is rarely as clean as the polished demos people see online.
That gap between the demo and the real experience is where many people quietly decide AI isn't for them. It's also why manager behavior can influence adoption more than access to another tool. I wrote about the data behind that here.
Watching someone recover builds confidence
I teach AI sessions constantly, including plenty of vibe coding courses, where things have a habit of going sideways.
I plan the build. I test it multiple times. Still, something occasionally breaks during the session, and I have to troubleshoot it live in front of everyone.
The surprising part is that participant feedback is rarely about what I built.
People remember the moment something broke and they watched me work through it. They mention the unexpected error, the approach I abandoned, and the debugging I did out loud. Across very different audiences, that's the part people ask to see more of.
It's also the part I would normally edit out of a recording, demo, or deck.
A polished demo shows the result. A broken one shows the recovery—and recovery is the skill people need when they try AI on their own.
A shared thread gives them that lesson without requiring you to be in the room.
Share the process without sharing sensitive information
ChatGPT, Claude, and Gemini can create shareable versions of conversations. In most cases, the shared page captures the conversation as it appeared when you created or updated the link.
But do not assume that sending a link only to your team makes it private.
Treat any share link as public unless the product explicitly confirms that access is limited to your organization. Before you send it, review the entire conversation for anything you would not want forwarded: client names, personnel issues, internal documents, private links, or numbers that have not been released.
When the thread contains sensitive information, copy the useful exchanges into an approved internal document instead.
Your AI experiment: Try this instead
👉 Time to tinker: Pick one thing you worked on with AI this week that required a few attempts. Share the conversation with your team and add a note such as:
"Here's the thread behind Friday's deck. The first two attempts missed the mark. The fix starts at message four."
This tells people where to look and gives them permission to have an imperfect process of their own.
If you manage a team: Share one thread a week in your team channel, then invite other people to share theirs. A manager showing imperfect work answers the permission question in a way no policy memo can.
If you don't manage anyone: Send a thread to one colleague who has already asked how you created something. It's a smaller step, and it works just as well.
What did you discover?
Send one thread this week and pay attention to what comes back.
Does someone reply with a conversation of their own? Does a colleague admit they've been stuck on the same issue but never said so? Does the thread lead to a useful discussion about how your team works with AI?
Give the experiment a couple of weeks. The response may come more slowly than with other experiments in this newsletter, and it may look more like a conversation than a measurable result.
📩 Not subscribed yet? Hit the button at the top. You won't want to miss what's next.
