Using AI for OKRs? My Dos and Don'ts for 2027 Planning
In short: use AI to prepare, challenge your assumptions and wordsmith your OKRs, but don't let it do the thinking for you. It only knows what you feed it, it never talked to your customers, and the outcomes have to be yours, not the AI's.
It's 2027 planning season, and I've been facilitating annual OKR sessions for the past couple of weeks.
AI is now part of these conversations, whether we plan for it or not. People prepare with it, draft with it, and bring its answers into the room. Used well, it makes the session sharper. Used badly, it gives you a beautiful set of OKRs that nobody really owns.
And because sharing is caring, here's my list of dos and don'ts for using AI for OKRs.
What Shouldn't You Do When Using AI for OKRs?
Don't bring AI to do the work for you.
AI needs your point of view and your critical thinking more than anything. If you ask it to write your objectives, it will, and they'll sound good. But they'll be built on what it can guess about you, not on what you know.
Don't trust everything AI says.
It only knows what you fed it and what's online. It never talked to your customers. It doesn't know what happened in last quarter's review, which team is overloaded, or what your customers told you last month. Treat its answers as a draft, not as the truth.
Don't make the sessions too heavy.
People need time to think, especially with AI involved. AI makes everything faster, and it's tempting to fit more into one day. But a couple of hours a day over a few days beats one long exhausting day. The best thinking often happens between the sessions, not inside them.
Don't force the output you expected.
Sometimes the session goes somewhere you didn't plan. That's usually a good sign. Trust the process, you'll get there.
What Should You Do With AI in OKR Planning?
Prep before the session, and hold people to it.
Prep the AI too, with your real context. The quality of what AI gives you depends on what you give it: your strategy, your data, what you learned this year. Without real context, you get generic answers.
Bubble up learnings from the team.
The people closest to the work know things leadership doesn't. Make sure their learnings reach the leadership conversation, and feed them into the AI as well. This is the context AI can't find online.
Send people off to converse with the AI.
Instead of driving one chat from the front of the room, send people off in small groups to work with the AI, and then bring it back to the room. You get more ideas, more ownership, and much better conversations when they come back.
Let AI challenge your assumptions.
This is where AI is at its best. Ask it what you might be missing, where your plan is weak, what could go wrong. Let it find your blind spots, and then decide as humans what to do about them.
Use AI to wordsmith the OKRs.
Once the thinking is done, AI is great at making the wording clear and simple. This is the part where speed really helps, and where nobody loses anything by handing it over.
Who Owns the Outcomes?
Call it planning, call it goals, call it OKRs. The name matters less than the ownership.
AI can help you prepare, challenge and write. But the outcomes are yours. You're the ones who'll work toward them all year, explain them to your teams, and decide what to do when things don't go as planned.
Use AI to sharpen your thinking, not to replace it. And make sure you own the outcomes, not the AI.
Once your goals are set, the next question is what the product team will actually work on. I wrote about that here.
FAQ
Can AI write our OKRs for us?
It can, but it shouldn't. AI only knows what you feed it and what's online, so OKRs it writes alone won't reflect what you know about your customers and your teams. Do the thinking yourself, then use AI to wordsmith.
How should we prepare AI for an OKR session?
Give it your real context: your strategy, your data, and what you learned this year, including what teams hear from customers. Without that, you get generic answers.
What's the best way to use AI during the session?
Send people off in small groups to work with the AI and bring their thinking back to the room, instead of one person driving one chat from the front.
How long should an OKR planning session be?
Shorter than you think. A couple of hours a day over a few days works better than one long day, because people need time to think between sessions.


Comments