Org design is product design now

Most product organizations are shaped by a constraint that no longer exists: individual output was expensive, so we organized to maximize it. Pods sized for how much a team could build in a quarter. Roadmaps as rationing systems. Design, product, and engineering as separate crafts because each demanded a full-time specialist.

AI broke the physics. A single person with good tools now carries what used to take a pod. And when individual velocity stops being the constraint, the constraint moves up a level into decision latency, review capacity, and coordination cost. Into the org chart itself.

This is why giving everyone Copilot and seeing nothing change is the most common AI story in the enterprise. Tools multiply individual speed; the org then absorbs the gain in the same meetings, handoffs, and approval chains as before. Velocity is an org property, not a tool property.

What actually moves

The teams I've seen genuinely change speed, not demo speed but shipped speed, share a few structural choices:

Leadership goes first. Adoption is top-down. When the executive team uses the tools daily, sets explicit expectations, and inspects usage the way they inspect revenue, behavior changes in weeks. When AI is delegated to an enablement function, it dies in a wiki.

Smaller units, fuller ownership. The pod of eight becomes two or three people who design, build, and ship a whole surface. Not because headcount is the goal. Every seam between people is now more expensive than the work on either side of it.

The EPD merger stops being a metaphor. When designers ship code and engineers generate interfaces, three parallel crafts with three review chains is pure drag. The boundary that matters is judgment domains: who owns quality for what.

Review capacity is engineered, not assumed. If your people produce five times more, someone must judge five times more. Orgs that don't redesign review through smaller units of change, clearer standards, and senior time explicitly reallocated toward evaluation just move the pile from "waiting to build" to "waiting to merge." That's taste as an org property, not a personal one.

Ambition resets. This is the one most leaders miss. If your roadmap was scoped to pre-AI capacity and you 5x the capacity, the honest move isn't finishing the roadmap early. It's asking what you now owe customers that was previously unaffordable. Speed without raised ambition is just cheaper mediocrity.

The design framing

I think about this as a design problem because it is one. An org is an interface between intent and shipped product, and it has all the properties of any interface: latency, affordances, failure states, defaults. The same split shows up in design systems at scale: you can enforce mechanics, but judgment still needs owners. Most orgs are running a UI designed for hardware that's been swapped out underneath them.

Redesigning it is uncomfortable in ways tooling never is because it touches titles, teams, and identities. Which is exactly why it's the moat. Anyone can buy the tools. The willingness to reshape the organization around what the tools make possible is the part that doesn't copy-paste.