Your AI Wrote Your Goals. That's Why Nobody's Hitting Them.

It took four minutes. You pasted the quarter's context into ChatGPT, asked it to draft your team's OKRs, and what came back was genuinely good — specific, measurable, sensibly scoped, better structured than anything you'd have written by hand at 4:45 on a Thursday. You cleaned up a couple of words and sent it. Everyone nodded in the meeting. The goals were clear.

Then the quarter went slack. Not dramatically. People just didn't move on them the way the goals promised they would. By week six, you were the only person who could remember what objective three even was, and you were the one who wrote the prompt.

It's easy to read that as an execution or motivation problem with your team. It's neither. The goals were fine. That was the problem.

Good goals and owned goals are not the same thing

There's a study worth knowing about here, because it isolates exactly this. Researchers at the University of Pennsylvania and Bern ran an experiment comparing goals that people wrote themselves with goals that an AI wrote for them, each drawn from the same personal reflection the participant had just completed.

The AI goals were better. Measurably. They scored far higher on all the qualities we're taught to want — specific, measurable, achievable, relevant, time-bound. If you graded the two sets on paper, the AI would win handily.

Yet, the people who got the AI goals owned them less. They felt less commitment to them, rated them as less important, and — the part that should stop you — two weeks later, far fewer of them had actually done anything. Roughly three in four of the people who wrote their own goals had acted on at least two of them. Among the people handed the better, AI-written goals, fewer than half had. Same reflection going in, higher quality on the page, and nearly thirty points less follow-through.

When the researchers traced why, it wasn't quality doing the work in either direction. It was ownership. A sense that this goal is mine, that I authored it, predicted every downstream thing that mattered — commitment, importance, whether you got off the couch. Objective quality predicted none of it. The goal being good didn't make people chase it. The goal being theirs did.

That word — theirs — is the same thing running underneath so much of how teams actually work. It's the autonomy need, the first of the three things every team needs from a leader, showing up at the scale of a single goal instead of a whole job: the sense that your judgment is yours. Hand the authoring to something outside you, and that's the exact thing that never gets to form.

Why the efficiency is the trap

This runs directly counter to what AI is supposed to be for. The entire pitch is that it removes the slow, effortful, kind-of-annoying part of the work so you can skip to the output. And for a lot of tasks, that's exactly right — nobody needs to feel ownership over a reformatted spreadsheet.

But a goal isn't an output. The slow, effortful, kind-of-annoying part — sitting with what actually matters this quarter, arguing with yourself about what to cut, deciding what you're willing to be measured on — isn't overhead getting in the way of the goal. It is the goal. It's the part that turns a sentence on a slide into something you'll actually feel on a Tuesday when it would be easier not to. Let the AI do that part for you, and you get the sentence without the thing that makes the sentence move you.

There's a sharper edge in the study, too. The people who lost the most ownership were the ones lowest in self-efficacy — the people least sure of their own judgment, who are precisely the people most tempted to let the confident machine take the wheel. The tool erodes ownership most in exactly the people reaching for it hardest. If you've ever handed a decision to AI because you weren't sure of your own call, that's the mechanism, and it's pointed right at you.

Keep the judgment, lose the drudgery

None of this means close the tab. It means being precise about which part of the work you're handing over.

Use AI to pressure-test a goal you wrote — make it sharper, catch the vague verb, ask what you're missing. That keeps you the author and makes the output better. What costs you is the other direction: letting it author the thing and casting yourself as the editor. The moment the first draft of what matters comes from outside you, the ownership that would have carried it never forms, no matter how clean the sentence is.

The practical version is almost embarrassingly simple. Write the ugly first pass yourself — badly, in your own words, before you open any tool. Then bring the AI in to challenge it. You'll end up with a goal that's both good and yours, which the study suggests is the only combination that actually gets done.

This is the whole design idea behind the Manager's AI Prompt Kit. It doesn't write your hard conversations, your feedback, or your goals for you — it puts you through the reps, so your own judgment is sharper when it counts. You stay the author. The AI is the sparring partner, not the ghostwriter.

See how the AI Prompt Kit works →

Before you automate the next decision that's actually yours to make, it's worth knowing how you tend to lead when the pressure's on — where your judgment is sharp and where you're quickest to hand it off.

Find your leadership pattern →

Already know your style? Grab your free cheat sheet →

Sources: Chi, Rietsche, Göldi, Ungar & Guntuku, “Optimized but Unowned: How AI-Authored Goals Undermine the Motivation They Are Meant to Drive” (2026) | Self-determination theory (Deci & Ryan) on autonomy and psychological ownership