Steal these goals: four ways to get more fluent with AI this year
Nobody grows fluent in a tool by owning it. You grow fluent by using it badly, then a little less badly, then one day without thinking about it. That is the whole arc. The license is not the finish line.
Copy the objective into your review form. Under each one is a For your form line with a date and one countable thing. Pick one, pick all four, remix them.
Use it on real work, every week
Use Copilot on one real work task each week through Q2, and keep a six-line note of what I asked, what I changed, and what I sent.
Twelve documented work examples by June 30, each naming the task and what I verified before I sent it.
“Just start using it” leaves most people stuck at the same three prompts forever. Done means a real task, not a skimmed lesson — and for each week, one thing you did differently at work.
Prove it with a six-line note naming the real task. “Context: rewrote my status update after feeding the model the last three weeks of project notes.” That note is your evidence.
Start a team shareout, and keep it going
Launch a recurring AI shareout for my team (a 20-minute standing meeting or a dedicated channel) and contribute or facilitate at least twice a month.
Shareout live by end of Q1, sustained for two quarters, with a logged contribution every two weeks.
Fluency spreads sideways, not down. The most useful prompt on your team is sitting in one person’s history. A shareout is a container for that: what saved an hour, what wasted one. Bad outputs are curriculum too.
Done looks like a real place with a rhythm that survives a busy week — not a channel that goes quiet by February. Count the contributions. If people start bringing their own without being asked, you built something that outlives the goal.
Go a week without AI, and write down what hurts
Spend one work week pulling AI out of every task where it is optional, documenting my actual workflow and noting every point where I reach for it out of habit or genuinely miss it.
One documented AI-light week completed in the quarter, with a written friction map submitted after.
This is not a purity test. If the work is load-bearing on AI, keep it there and say so. You are sorting reflex from benefit. You cannot see the water you swim in — pull the optional stuff out and the ache becomes a map: where the tool earns its keep, and where you were reaching out of habit.
The deliverable is the document. A page or two, dated, honest. The person who can name precisely what AI does for them — and what it does not — is more fluent than the person who uses it for everything without a second thought.
Build the thing that makes your work better
Use what I have learned to build one thing that measurably improves my work, my time, or my week, by end of year.
One built thing shipped by December, with a named before-and-after.
Size it to where you are. One person might clear a weekly chore with a saved prompt. Another might build an agent a whole team leans on. Both count. This one changes a Tuesday — better work, freed-up time, or a lighter Thursday. Not a demo. Not something built to look busy in a review.
Prove it with a before and after. “Before, I spent an hour every Monday reformatting the same report. Now it takes five minutes and I use the other 55 on the analysis, which is the part I like.” If you can say it and mean it, you did not just get more fluent. You got more free.
Review before you send, especially when the output sounds confident.
MillerKnoll AI · 2026