The hard part of an AI reorg isn’t the chart. It’s the room.
A lot of the advice on reorganizing around AI starts with the org chart: which roles change, which get created, who reports to whom. After 18 years of running TESMO, I think that’s the easier half.
The harder half is the room: getting a real team to understand the change, trust it, challenge it, and eventually own it. Especially the people most wary of change.
For us, that matters because marketplace management is already a high-complexity operating environment: catalog, content, advertising, demand planning, reporting, brand protection, client communication, and a constant stream of edge cases. AI can absorb a lot of execution inside that system. But it does not replace judgment. If anything, it makes judgment matter more.
Like a lot of companies right now, we’ve been steadily adjusting to a new way of working and evolving the organization as we go. Recently we made the structural changes official and rolled them out to the team. It went well. Here’s exactly what we did, and the one move that made the difference.
The situation
Over the past six months we’d been integrating AI into the company fast. We built an internal library of reusable Skills, rebuilt recurring workflows around AI-assisted execution, and started removing manual steps from work that used to consume days: analysis, reporting, content iteration, process documentation, and internal QA. Underneath it all sits an internal layer we built - our own MCP server - that plugs our AI into nearly every part of the business. None of it was announced as a “transformation.” It just happened, week by week, as the better way to do the work.
By June it was time to make the new shape official: a new accountability chart built around the way we actually operate now - people applying expert judgment across their domain, with AI handling more of the execution underneath them.
The risk was obvious. A new org chart, dropped on a team as an announcement, reads as upheaval. And the people it unsettles most are exactly the ones who are slowest to change - the people whose buy-in you most need and are least likely to win by decree.
The realization that changed the approach
Here’s the thing I almost missed: the change had already mostly happened.
The team had been working in the new way for months. The roles had quietly shifted. The chart wasn’t going to introduce a new reality - it was going to describe one everyone was already living in. The chart had lagged reality; it was never leading it.
That reframed the entire meeting. I wasn’t there to push the team off a ledge. I was there to write down the place we’d already walked to. We were already walking. I wasn’t announcing a change; I was acknowledging one we’d already made.
Recognition, not revolution. That’s a far smaller ask, and - critically - it was true. People can feel the difference between an honest reframe and a spin job. This was honest.
How I actually ran the meeting
The sequence mattered more than the slides. Here’s the order, because the order is the part you can steal.
1. Context before the chart. I didn’t open with the new structure. I opened with the last six months - the Skills we’d built, the workflows we’d changed, the processes we’d streamlined. I walked the team through how far they’d already come. By the time anyone saw a new box on a chart, the room had already agreed on one thing: we don’t work the way we used to.
2. Then the chart, as a conclusion. Only after the context did I present the new accountability chart. Framed that way, it didn’t land as a shock - it landed as the obvious formalization of what we’d just spent ten minutes establishing. The chart answered a question the context had already raised.
3. The “why this shape” slides. A few slides on the thinking behind it. The core idea is simple: when AI can absorb more of the execution layer, you stop organizing around tiers of task-doers and start organizing around domain experts - the subject-matter experts (SMEs) with years of judgment - each using AI to extend their judgment and throughput.
I was careful not to oversell it, and that mattered. AI absorbs most execution, not all of it - some work still needs a human, and the exception rate is real. So the structure isn’t an SME working alone; it’s an expert plus AI plus a thin support layer, sized to how often the genuine exceptions actually come up. And because more AI output means more surface area for quiet errors, someone owns watching the workflows themselves. Showing the team the honest version - including where AI doesn’t reach yet - did more for credibility than a cleaner story would have.
I also used a crawl-walk-run frame: here is what is already working, here is what we are formalizing now, and here is what we are not pretending to have solved yet.
4. Open the floor - everyone. Then we discussed it as a team, and every person had input. This is the step most reorgs skip, and it’s the one that converts compliance into buy-in. You can’t present-and-dismiss your way to ownership. Letting people react, push, and shape the thing is what makes it theirs.
What happened
It went smoothly. Most of the team was genuinely energized by the direction - they’d rather be at a company pushing hard into what’s next than one that isn’t. A few had some trepidation, and said so. Notably, they framed it themselves: they acknowledged they tend to be slower to change.
That last part is a feature, not a bug. Trepidation that gets named in the room is trepidation you can work with. Trepidation buried under nodding compliance is the kind that quietly sinks a change three months later. Because the meeting invited input instead of demanding agreement, the hesitation surfaced where I could address it.
The template you can steal
If you’re about to roll out an AI reorg - or any structural change - to a team that includes people wary of it, here’s the play:
- Audit what’s already changed but unlabeled. Before you announce anything, find the parts of the new structure your team is already living. That’s your opening - and your honesty check. If the answer is “nothing has changed yet,” you’re not acknowledging a change, you’re imposing one, and you should slow down.
- Lead with the distance already traveled, not the destination. Context before chart. Let the team arrive at the conclusion before you show it to them.
- Frame it as recognition, not revolution. “We’re already most of the way there” is a smaller, truer ask than “the ground is shifting under you.”
- Use a crawl-walk-run ladder. It sets the expectation that the end state isn’t due tomorrow, which makes open questions feel normal instead of alarming.
- Put the reframe in the artifact, not just your delivery. I baked the “we’re already most of the way there” framing directly into the deck, not just the way I talked through it. A reframe that only lives in your verbal delivery dies the moment the deck gets forwarded without you. Write it into the thing.
- Be honest about where AI doesn’t reach yet. The credibility of the whole reorg rests on not overselling. Naming the work that still needs a human is what makes the team trust the parts you’re automating.
- Make input the final step, and mean it. The discussion isn’t the formality after the decision. It’s the part that turns your reorg into their reorg.
The chart is the artifact everyone fixates on. But the chart doesn’t change anything by itself. The room does. Get the room right, and the chart is just paperwork catching up to a team that already moved.
Thinking through a change like this in your own org? Let’s talk.