Essay
Your Team Isn't Bad at AI. They're Grieving.
Most teams treat AI adoption as a training problem, so they buy tools and run sessions and wonder why nothing moves. The thing actually in the way is harder to schedule around.
"I've got a great idea," I told my manager. "I'm going to build our design system, turn it into a plugin, push it to our org repo, and train everyone to run it themselves." I paused. "I guess then I won't have a job."
I said it as a joke, but there was fear under it. Not that I was replaceable, but that I wasn't as smart or as creative as I'd thought, and AI was about to prove it. I was proud and terrified at the same time. And I realized that if that was how I felt, my teammates had to be feeling it too.
The thing that stalls a team is grief
Every team hits roadblocks adopting AI, and the technical ones are easy to see. Mine was leaning into my infrastructure engineer husband's office to ask what a grep was, and why you push and commit to a repo when those seem like the same thing. Those are yield signs. They slow you down, they don't stop you.
What stops a team is the emotional part. People can't learn while they're working out whether the thing they were good at still matters. AI is making everyone redefine what they believe about their own ability, and when that redefinition lands on "then I was never as good as I thought," you don't have a training problem. You have a team of people grieving, trying to learn from the tool that caused it.
You have a team of people grieving, trying to learn from the tool that caused it.
Proof reaches people when reassurance can't
Being told I wouldn't lose my job didn't help. It wasn't true, anyone can lose their job, and I knew it while he was saying it.
What would have reached me was evidence: the three headlines on our homepage that AI couldn't have written, the campaign idea that was all mine, the product logo I'd drawn by hand. Show someone proof of the ideas only they could have had, and you can redraw the definition of their job with them. "Am I going to be okay" has no answer. "What am I the only one here who can do" does.
Show someone proof of the ideas only they could have had, and you can redraw the definition of their job with them.
What stays human when AI arrives
My team got there on our own. We're capable and we landed on our feet, but that's a circumstance, not a system. The next team might not be so lucky.
Our growth marketer knows which campaign is worth building and which one to kill. AI can write the report. It can't make the call, because the call runs through market, personas, data, internal politics, budget, and instinct all at once, and then someone has to be wrong in public if it goes badly.
Our producer will get knee-deep in an argument with the CEO over a single word, because he knows it's the wrong one and he's right. He's spending real social capital on a word. A model will produce the approximately-right word every time and never once feel the difference.
Our webinar series came to me framed as a teacher in a classroom. I changed it to a conversation between a leader, a researcher, and a teacher, because our audience doesn't want to watch someone teach. They want to see research become a change in a classroom.
None of that came from a tool, and none of it was going to. Once a team is solid on what makes their work good, alone and together, AI finally becomes the thing it was sold as. It saves time.
Someone has to go first
What would make this easier for any team is a ritual for the grief. (I don't mean a funeral for my design briefs.) I mean that as the work speeds up, you stop and mark what's changing and what's being lost, and you let someone say out loud that this is hard and disorienting and a little frightening.
That's the leader's job, and it starts with the leader going first. Someone has to say "I don't know where this is headed or what's going to be valuable, and that's scary" before anyone else will. It doesn't change the plan. You still ship. It just means the fear has been said, and nobody's carrying it alone for a while.
Someone has to say "I don't know where this is headed or what's going to be valuable, and that's scary" before anyone else will.
I've rarely seen a team actually do this, including my own, and I understand why. It asks a lot to stop and name fear when the work is piling up, and capable people would rather just push through. But pushing through alone is the expensive version.
You can't solve the grief. But once it's out in the open it becomes less oppressive, and that's what clears room for the technical work everyone wanted to start with. You can hand out real ownership, so everyone knows who holds the standard for which skill. You can automate the thing that makes you go "ugh, not this hour of my week again." And the team stops fighting the change, because you're not asking them to pretend it costs nothing.
The invitation
I was afraid AI would prove I wasn't as good as I thought, and it did, in a way. It's very competent at the parts of my job I'd pinned most of my worth to. I'm efficient, I'm organized, I'm excellent at pattern recognition. What it can't do is the part I had the hardest time naming, which turns out to be the part that was only ever mine.
Every time the model gets better, it asks the same question again: what's left that only I could have made?
Every time the model gets better, it asks the question again. What's left that only you could have made?
You can hear that two ways.
One is a threat, evidence that the ground keeps shrinking until one day there's nothing left that's yours. The other is an invitation, because there's always more to find, in the work and in yourself, and a better tool is only what sends you looking.
I hear it as an invitation. I've never found the bottom of what a person can come up with, and I don't expect to. Leading through this is mostly choosing to hear it that way out loud, so your team can hear it that way too.