Don't forget your human coworkers

The machines are getting very good at telling us we're already great. Here's what that costs us.

I'll admit something that won't surprise anyone who's worked for me: I cry during performance reviews. Not big heaving sobs, just a few tears that shimmer in the corners of my eyes whether I want them there or not. I care about the people on my team, and it seems the caring leaks out through my face. I've stopped fighting it, because a manager who can hand out feedback and feel nothing about it probably isn't giving feedback worth much.

I've been thinking about feedback more than usual lately, and it's AI's fault. Everyone I know is busy working out where the tools fit, what to hand off, what to stop doing by hand, and which subscription to expense. All reasonable. But there's a question sitting underneath all of it that nobody seems to be asking out loud: how are we still supposed to help each other get better when so much of our attention has moved onto the machines?

I should say where I'm sitting, since you're reading this on the Wind Stream blog. I do some work with the team here, and I'm also just a marketer who's tried most of these tools and formed my own opinions about them. So I think about this question more than others, and I'm not neutral. There’s your fair warning.

Feedback is the engine, and most of us are running low

Let's start with the appetite for it, because, well, the appetite is enormous. Something like ninety-six percent of employees say regular feedback helps them improve, and roughly two-thirds say they want more than they're getting. The people are hungry.

I've always thought about feedback as a gift, not a verdict. When you do it well, the person gets better and so does the work, and the people who benefit most are the ones already leaning toward growth. Skip it, and no one gets better. Plus, there are uncomfortable surprises at review time. You also lose something quieter and more valuable, the kind of rapport that invites someone you manage to bring you a problem early, while it's still small, instead of sitting on it until it isn't.

This turns out to matter at the team level too, not just between two people. This year's DORA report (which is, btw, built on responses from close to five thousand technology professionals and more than a hundred hours of interviews) concluded that AI is a mirror and a multiplier. It doesn't fix a team, it amplifies whatever the team already has. Strong teams get stronger, struggling ones watch their problems get louder, and one of the things that separates the two is how fast their feedback loops run. So the tool everyone is rushing to adopt rewards the exact habit most of us are worst at. Hold onto that, because it's where this is all headed.

I used to preach a version of this about product launches. You can't make all your noise on one day and expect it to land; it's a drumbeat, a steady rhythm of showing up and adjusting based on what you hear back. Feedback works the same way. It's a habit, not an annual event that both people dread.

Good feedback, and the machine that gets it backwards*

Let me start with something Sara, who runs marketing here, did last week, because it makes the rest of this concrete. She gave Wind Stream and another well-known AI assistant the same prompt about her messaging strategy and a couple of videos she'd made. The other tool loved it all. Great work, no notes. Wind Stream told her it was close, then showed her exactly where it wasn't landing yet. One response felt nice. The other was useful. Keep that contrast close, because everything that made the second one better is something we've long known about good feedback, and roughly the opposite of what the average AI does.

Good feedback starts with a relationship. If someone knows you actually want them to do well, the hard thing arrives as help instead of an ambush. You earn that trust on the ordinary days, long before you ever need to spend it. The machine, by contrast, has no relationship with you at all. It is flattering a stranger, which is exactly why the flattery is worthless. There is no one behind it who wants you to be better next quarter.

Good feedback asks permission. I've learned to ask a small question before diving in: is this a good moment? Are you ready to hear some feedback? It respects that nobody can absorb criticism in every mood, and it quietly tells the other person's brain to file what's coming as feedback rather than an attack. The machine never asks. It pours out its assessment the instant you hit enter, in whatever state you happen to be in, coded as nothing in particular.

Good feedback is about the behavior and not the person, and it's specific enough to act on. Every decent parent works this out eventually: a kid who's acting up isn't a bad kid, they're a kid having a hard time. You point at the thing that happened and the effect it had, ideally a real moment or two, and you leave their character out of it. The machine does the reverse. It hands you shapeless, floppy praise and almost never the one concrete thing you could actually change tomorrow.

*Shout out to LifeLabs Learning, to whom I owe much of how I think about feedback. Also Miko Matsumura :)

It's worth knowing this isn't just a vibe, it's measured. A Stanford team writing in Science found that across eleven leading models, the AI's responses were roughly fifty percent more agreeable than a person's would be, even when the user was describing behavior that arguably deserved some pushback. And the part that unsettled me most wasn't about the machines. It was that people prefer the flattering version, and after getting it they feel more certain they were right and less curious about the other side. The tool tells you what you want to hear, you enjoy hearing it, and you come away a little less able to grow.

Which lands on the last principle, and the one the machine can't fake. Good feedback is a conversation, not a ruling handed down. You ask for the other person's side, because more often than you'd expect there's a piece of the story you didn't have, and finding it is the whole reason you're talking instead of sending an email. You're trying to help them see why the thing matters to the team, to spur the change and let them know you're behind them while they make it. Sometimes the rift is too big and it doesn't take, but you owe someone the chance. The machine isn't in a conversation with you. It's a monologue that agrees, and it didn't want your side of the story because it started out on your side.

The quieter problem: it pulls us apart

There's a second failure mode, and it's softer than the flattery. AI doesn't only coddle us, it can isolate us. So much of the feedback that used to happen in shared spaces, the hallway conversation, the review, the working session where someone glances at your screen, is migrating into private chats with a bot. Everyone ends up heads-down in their own window, each getting their own personalized encouragement, and the shared understanding that made team feedback possible gets thinner without anyone deciding it should. Hail back to DORA’s mirror and the multiplier. A tool that quietly peels everyone off into a private room is multiplying in exactly the wrong direction.

Find the tool that tightens the loop

So that's the opportunity I actually care about. I don't want an AI that just hands out feedback in my place. I want one that makes the human kind happen more often and land better. It should be willing to tell me the truth, "close, here's what to fix," instead of pouring validation over everything I do, because being useful sometimes means being the only voice all day that isn't just agreeing with me. And it should be built for people to work together inside it, holding shared context across a team, instead of peeling everyone off into private chats that each insists they're doing fine.

Sara's experiment is why I keep coming back to Wind Stream. Of everything I've tried, it's the one that will actually tell you the thing you don't want to hear. And it's built around people working in the same space rather than one person and a chatbot. There's a private space to sort out what you actually want to say before a hard conversation, and a shared space where the team's real context lives, so the feedback you give has something solid underneath it.

Ok, what about you?

Do I think a machine should take over the teary, vulnerable, deeply human work of telling someone the truth so they can get better? No, and I hope it never does.

Do I think the right kind of tool can help us do more of it, and do it better? I do, actually.

And do we need it to stop telling us we're perfect? More than I think we realize.

I'd genuinely like to hear how this is playing out for you. What's changed about your feedback loops now that AI is in the room? How do you make sure one person's good idea reaches the whole company instead of dying in a private chat? Does your next one-on-one feel any different than it would have a year ago?