Where are you with AI?

• You’ve tried ChatGPT, Gemini, or Claude once or twice in your personal life.
• At work, you mean to use AI more, but the moment never quite arrives.
• You can’t name a task where reaching for AI feels obvious yet.
• The cost is invisible. You don’t know what you’re missing because you haven’t tried.

Pick one task you do every week (drafting an email, summarizing a meeting, looking something up) and try AI on it three times this week. Don’t aim for transformation. Aim to find one task where AI saves you ten minutes. That’s the start. Most people get stuck here because they’re looking for a flagship use case. There isn’t one. There are small ones, and you find them by doing.

• You’ve tested AI on a handful of tasks; some clicked, some didn’t.
• You bounce between tools (ChatGPT one day, Claude the next, Copilot in your inbox).
• You don’t have a system. You’re collecting data.
• You’d say AI is useful, but you couldn’t draw a line around exactly when.

Stop collecting; start narrowing. Pick the two tasks where AI clearly worked, and commit to using it the same way every time for two weeks. Save the prompt that worked. Show one colleague. Experiments tell you what’s possible. Repetition is what makes it stick.

• You’ve named the tasks where AI is now your default. You don’t reconsider it each time.
• You probably keep prompts saved somewhere (a notes app, a Doc, browser bookmarks).
• A few of these workflows are starting to matter. If the tool changed under you, you’d feel it.
• You can teach a colleague what you do, even if the prompt is mostly in your head.

Treat your adopted workflows like the tools they’ve become. Where do your prompts actually live, and what happens when the model updates? Do you have a way to notice when output quality drifts before it costs you something? Adoption stalls when the workflows are unsupervised. The move from this stage to the next is mostly about making them survivable.

• You’re not just using AI; you’ve changed your habits, your tools, your week, because of what AI enables.
• You’ve stopped doing some things you used to do. You’ve started doing others you couldn’t before.
• You have opinions about which tools fit which jobs.
• You’re starting to wonder whether the changes are actually paying off (in hours, in quality, in the scope of what you can deliver).

Measure three of your changed workflows honestly. Hours saved, quality shift, scope expanded. Without numbers, every “should I keep this, expand this, drop this” call is taste. With numbers, the next round of adaptation has a baseline. One more thing: notice what you’ve stopped doing. Sometimes the most valuable adaptation is subtraction.

• You can’t separate “AI work” from “your work.”
• You design what gets done; you rarely do all the steps yourself.
• The output of your work depends on AI being part of it.
• You help others get more out of it, sometimes formally, sometimes by accident.

The question stops being “how do I use AI” and becomes “what should I be doing with the time AI gives me back?” That’s not a tools question. It’s about judgment, taste, the parts only you can do, and whether you’re actually spending your hours on those. The work at this stage is mostly about you. The AI is settled.