A machine can write better feedback about you and your work than your boss can.
In one study, people who got feedback from AI improved twelve percent more than those who got feedback from a real human.
Then to complicate things, the researchers told them that a machine wrote it.
As a result, the people's performance didn't just dip. Oh no. It fell way below where they started in the first place.
Same words, same advice.
The only thing that changed was who they thought it came from.
That's the whole story this week. Let me explain.
I'm Alexey, I co-run Polar Bear, a people ops company, and this is People Ops News — the weekly show on what's really happening in the world of people and work.
Today we're going deep on feedback, because it seems everything we know about it is a bit backwards. Let's go.
So right now it's mid-year review season. And guess what? Managers are using AI to write their performance reviews. And the pitch from basically every HR tool in 2026 is: AI writes better feedback, in seconds.
And on paper, that pitch is true.
Think back to the study I just mentioned. AI feedback generally worked better. It was more consistent, more specific, less biased, than what the managers wrote.
So if feedback was just about the quality of the words, AI would win. Easily.
But hold that thought. The study told us that the words were never the point.
Quick one before we go deeper. On Thursday, July 2nd, I'm running a free 60-minute workshop, in collaboration with Agency Folk, where you will build your own career framework, live and in sixty minutes. The link is in the comments — and please follow me so People Ops News shows up every week.
Okay, back to feedback.
The Science — 30 Years of Feedback Research
Last year three researchers did something that no one has done for thirty years. They went back through the entire science of performance feedback. One hundred and seventy-three studies. All to settle one question: when does feedback actually make people better?
And the answer splits in two.
Positive feedback — telling someone what they did well — almost always works.
Negative feedback — the criticism, the "here's what you need to fix" — is a bit of a coin flip. Sometimes it helps, sometimes it doesn't, and sometimes it makes things worse.
And what decides how it goes is not how well it's written. It's whether there's a trusting relationship between the person giving feedback and the person receiving it.
If your people trust you, if you've got real credibility with them, criticism lands as coaching and your people get better.
If the foundation isn't there, your criticism lands as a threat, and performance dips.
Now go back to the AI. A machine has no relationship with you. No trust. No history. No skin in the game.
So the second people found out the feedback was written by a machine, it backfired.
The research and the experiment land in the same place. Feedback isn't information, it's a relationship.
The words just sit on top of it.
Which makes what's actually happening out there kind of tragic.
McKinsey's big 2026 HR survey found out that one in five employees had no career or feedback conversations within the past twelve months. Not even a bad one. Zero.
And the rest, once, maybe twice a year. Far too rare to actually help anyone grow.
So we've got two failures running at the same time.
In a lot of places, feedback is just absent.
And where it's not absent, we're rushing to hand it to a machine.
Both of these skip the one thing that makes feedback work: the human relationship.
What Actually Works
So if you run a team, here's what the evidence supports — and none of it is a new tool.
One, lead with positive. It's the one kind of feedback that almost always works.
Two, earn the right to criticize before you do it. The hard conversation only works if the relationship is already there. So invest in it before you need it. And when you do criticize, tie it to a specific goal, not a vague "you need to get better." That's one of the few things that help criticism land, even if the relationship is quite thin.
Three, if you use AI — and you should — use it in the open. The second your team realizes that the feedback you said came from you actually came from AI, you've traded a few saved hours for the exact trust that makes feedback work.
The whole industry is trying to fix feedback by making the words better. Faster, smarter, AI-polished. But thirty years of feedback and one awkward experiment both say the same thing: the words were never a problem, it's the human relationship underneath them that needs attention.
There's a second half to this.
Once the relationship is there, it still helps if you have something concrete to point to. And this is where a career framework really earns its keep. It's one of the best tools out there for giving specific feedback.
Because a good career framework really spells it all out. So instead of telling someone "be more strategic" and watching them nod and have no idea what you mean, you can just point to the career framework.
This is what good looks like at your level. This is where you are, this is what you need to do. Now that's a real feedback conversation.
And here's one concrete thing you can do.
On Thursday, July 2nd, my co-founder Pauline and I are running a free 60-minute workshop with Agency Folk, called "Build your career framework in 60 minutes."
Live, with you, we go from no levels to a complete career framework: your agency's levels and roles, your expectations, and in your own language.
And you will walk away with a draft you can actually use. Link is in the comments.
And that's People Ops News. I'm Alexey. Please follow me to catch the next one, and I hope to see you on the second of July.