There was no reorg chart. Just a shutdown, and thirty people in it.

Situation

I loved that work. HoloLens was a new category, and nobody had modelled data for one before, so we were inventing how to measure a thing with no prior art to copy. It is the most interesting problem anyone has handed me.

In 2018, at the start of HoloLens V2, a new VP and a new director arrived with a refined charter and pivoted the program toward spatial mapping. The work my group existed to do, data science and analytics and experimentation, was deprioritized in the same decision.

The VP told me at the time that I had an agenda.

I didn't think so. I thought I was making the case for what the data work could contribute and that the case wasn't landing. I said so. I kept making it, in different rooms, for months. Nobody moved and I read that as a communication problem.

He was right and it took me years to hear it. I was arguing for a direction I believed in rather than asking what my team should become now that the program had changed underneath it. I never went back and re-scoped the work against the new charter, which was the actual job.

Not restructured. Deprioritized. There was no new org chart with my boxes moved into it, because there were no boxes. The data function was moving under a different VP entirely, as part of a realignment across the division.

Thirty people, three senior leads, a portfolio of investments, and no landing plan from above. I was the Group Engineering Manager over all of it, which turned out to matter a good deal less than it sounds. There was no seat waiting for me either. My team knew something was happening and didn't know what, and I couldn't tell them how it ended because nobody had decided yet.

Task

Get every person to a strong landing inside about a month. Preserve what we'd built that was worth preserving. And do it without the team spiralling while V2 work was still in flight and still had their names on it.

The reduction was real too. Some roles genuinely didn't map to the new charter, and recommending those cuts was mine to do.

Action

Before the first conversation

Before I proposed a single move, I set transition principles with my director and made them the constraint everything else had to fit.

Teams move intact wherever possible. People come before org charts. Nobody hears about their future in a hallway.

The first one carried the most weight and it's the one that would have been easiest to trade away. Breaking teams up during a reorg is the efficient-looking move on a spreadsheet, and it compounds exactly the thing you're trying to contain. An intact team lands with its trust, its working relationships and its institutional memory still attached. A person landed alone lands with none of it and has to rebuild all three while also being new.

Then I did the homework that made intact moves possible. I met with my director and one of his top reports to map what capabilities the V2 charter actually needed. I met separately with the VP taking on the data function, to understand what her org needed rather than what I wanted to hand her. Those two conversations were the difference between placing people and selling them.

Placing thirty people

I engaged each of my three leads individually before proposing anything about their teams. They had context I didn't, about individual strengths and about who wanted what, and the proposals we ended up taking forward were built with them rather than delivered to them.

I wrote a shutdown exit plan for our current investments and shopped them across the division to see if anyone would take ownership. Nothing landed as-is. So we shut them down cleanly rather than leaving them half-owned. No orphaned projects is its own outcome, and it's one nobody thanks you for.

Two architects were load-bearing in a way the org chart didn't show. One, at L66, had designed the spatial mapping data model. One, at L65, was the backbone of the experimentation stack. For both, I worked with my director to craft an actual role first, with a real impact thesis in it, and then sat down and walked each of them through the work. Not a placement. An offer they could evaluate on its merits. Both stayed and carried the knowledge forward.

I recommended five people for cuts. Roles that didn't map, and performance that had been a question before any of this started. That was a judgment call built on 1:1 data and performance signal rather than a percentage handed down, and I handled every one of those conversations myself.

The parallel track

The structural work was the half I knew how to do. The other half ran alongside it the whole time and was harder.

I brought the full team together early and told them what was happening, what my principles were, and what the plan was. Then I listened, and answered what I could answer. That transparency was about the structure. It was deliberately not about individual personnel decisions, which I held closer. The team needed to know the plan. They didn't need to spend a month wondering whether they were on a list.

The rest of it was unscheduled. When I heard an IC was taking it badly I'd go to their desk and ask what was going on and how I could help, which is a better opening than anything I could have prepared. Sometimes the whole ask was to be heard by someone who knew what was happening.

Going quiet is the failure mode. People read every signal during a transition, and silence doesn't read as neutral. It reads as bad news being assembled somewhere they can't see.

Result

Every team landed inside about a month, with structures kept intact wherever the receiving org would take them that way.

Both architects stayed and stayed productive through the transition. The five cuts were handled directly, with warm handoffs into internal mobility, and without the lingering resentment that comes from people finding out sideways. The investments shut down clean.

Three months of that I spent without a defined seat. I had worked myself out of a role, which was foreseeable from the first week and didn't change the order I did things in. My director gave me the runway to find the right next thing rather than the fast one, and I landed as Group Program Manager for Xbox Data.

The outcome I'd actually point at isn't mine. My three leads are now a Partner Engineering Manager in Gen AI Safety, a Group Data Science Manager in Teams, and a Senior Director in AI Security. Those are their careers and their work, and I'm not going to claim them. What I'll claim is narrower: none of them landed somewhere that stalled them, and in a dissolution that outcome is not automatic.

Learning

The hardest part of shutting an organization down isn't the structure. It's that the structural work and the human work run at the same time, on the same days, and the first one is legible while the second one is where the damage happens.

Putting the team first while I had no landing place myself wasn't martyrdom. It was the job, and it was also the only version that works. A team watching their manager scramble to secure his own seat learns exactly what his principles are worth, and no announcement afterward repairs it.

There is a senior leader I worked with across two orgs who never saw the value in what my team did, and I never changed his mind. I don't have a clean explanation for it, and I have stopped assuming I would win that argument with a better deck. It is still costing me.

I'd carry two things into any organization. Set the principles before the moves, because once the moves start, every principle looks negotiable against a specific person's specific problem. And separate structural transparency from personnel confidentiality on purpose. Those get collapsed together constantly, usually into silence about both.

Two years later, at Xbox, I was handed the same kind of analysis and it pointed at my own function. That one is here.

---

All selected work