My team is using AI to collaborate. So why does something feel off?

Sara and Tom were working on a project together. Design on one side, technical build on the other. Dispatched across the globe, different time zones, different professional languages. The deadline was real. AI made it possible to move fast across all of that distance and difference.

It also made it possible for them to never actually meet.

Not in the way that matters.

They were working across a language gap, not just geography. Sara’s world and Tom’s world didn’t naturally translate. So they each used AI to bridge it: Sara to turn her design thinking into technically legible briefs, Tom to turn his builds into something she could actually read and respond to. Things that would have taken hours of back-and-forth got handled in minutes.

What Sara would send Tom wasn’t four bullet points. It was the output of a full conversation with AI — all the thinking, the reflection, the decisions already made — distilled into a single clean message. Tell Tom everything we just worked through. Make it legible for him.

Tom would do the same back.

And at the beginning, that felt like collaboration.

The messages got longer. More thorough. More polished. Sara looked at one of Tom’s updates and thought: do I even need to read this? His AI understood her AI. The results seemed in place. She moved on.

He probably did the same with hers.

What neither of them noticed was that the judgment had left the building. Tom was executing from her prompts without really thinking about the product. Not out of carelessness, but because the relay had already removed the expectation of thinking. Things got changed that shouldn’t have been changed. Things didn’t get caught that a human paying attention would have caught.

When Sara reviewed the work, she couldn’t find Tom in it. No hesitation, no particular way of seeing, no friction that would tell her a specific person had wrestled with this.

She wondered if he felt the same thing reading hers.

They did have a call. Once. He didn’t turn his camera on. The conversation was flat — information exchanged, questions answered. Professional. Efficient. He didn’t ask about the vision behind the project. She didn’t push. Neither of them knew how to, by then. The temperature had already been set.

What disappeared wasn’t the collaboration. The collaboration functioned. The project moved. Things got built.

What disappeared was the shared pride of building something together.

That specific feeling, when you and another person have gone through something hard and come out the other side with something real, and you can look at each other and know you both did that. That feeling requires struggle. It requires the friction of two humans actually trying to understand each other, failing a little, trying again.

AI removed the struggle. And without the struggle, there was nothing to be proud of together.

Sara was complicit in this. She chose speed. She stopped reading his messages. She trusted the relay — her AI to his AI — and let the relationship become a protocol.

It’s a strange thing to grieve. Nothing broke. No one was unkind. The work got done.

But she kept thinking about Tom. About whether he existed somewhere behind those messages, curious about something, frustrated by something, seeing the project in a way she never got to know.

She thinks he was there. She just never created the conditions to find out.

AI didn’t make this collaboration worse.

It made non-collaboration feel like collaboration. And that’s a different problem entirely.

When collaboration breaks down the old way — misunderstandings, missed messages, communication that fails — you feel it. You course-correct. You pick up the phone. When AI smooths it over, you feel nothing. You keep moving. And somewhere along the way, without deciding to, you stop showing up as a person and start functioning as a filter.

We are optimizing for smoothness. In most workflows right now, smoothness is the highest value — the metric that tells us things are working. And some of that friction we’re removing is waste. Genuinely. Time lost to misalignment, to messages that never needed to be sent.

But some of it is the relationship that makes the work real.

We’re removing both. Without knowing the difference.

The question worth sitting with — for anyone building teams inside an AI transformation — is whether the people on your projects can still find each other in the work. Whether there’s enough friction left for someone to know a specific human wrestled with this. Whether the pride of building something together is still possible, or whether the relay got there first.

Sara and Tom finished the project. On time. Within scope. Neither of them would call it their best work. And they both know why, even if they haven’t said it to each other.

What they lost doesn’t show up in any project management tool. It’s not a metric. It’s the thing that makes people want to build something hard together again.

That’s what we’re optimizing away. And we haven’t noticed yet, because the work keeps getting done.