A Bad System Beats a Good Person: Drive Performance with Org Design

This continues the Friday series — one principle from 10X ORG each week, the ideas that make an organization ten times more effective, not ten percent. Last week was the one everything rests on: own, not rent. This week is the one that makes ownership concrete.

The same launch, built two ways

On October 1, 2013, millions of Americans woke early to sign up for health coverage on the new HealthCare.gov. Most of them met the same blunt banner: Something went wrong. Please try again later.

By 7 a.m., six people had made it through. By the end of day two, 248 enrollments had gone through — roughly one success for every hundred thousand tries. Behind that failing sign-up page sat one of the largest federal IT efforts in history: $1.7 billion in the first months, nearly $210 million of it on the login service alone — the very door nobody could get through.

The build had followed a familiar logic. Break the system into specialized subsystems, write the requirements for each, and hand the pieces to vendors. Fifty-five firms were contracted, each owning its own slice. Every slice may have been reasonable on its own. The whole thing did not work.

Two weeks in, the White House CTO pulled together a small rescue team — ten engineers, cross-functional, put under one roof and given a broad mandate: own the entire problem, work across the whole stack, ship an end-to-end solution. Over six weeks they did what hundreds of specialists hadn't. They fixed some four hundred defects, scaled the site to 25,000 concurrent users, and got page loads under a second. By the end of the first enrollment period, more than eight million people had gotten what they were promised.

Same mission. Same country. Same technology available to both. What changed between the disaster and the recovery wasn't talent or effort — it was how the work was organized. The country had run, without meaning to, a nationwide A/B test on org design.

Design is always happening — the question is whether it's on purpose

Here's the uncomfortable part: org design isn't an occasional event, it's something managers do constantly, usually without noticing. Every recurring coordination meeting, every new policy, every "let's appoint someone to expedite this" is a small act of organizational design. Most of it is unintentional — a local fix that makes today easier and quietly accumulates into tomorrow's complexity.

A VP appoints a project manager to rush one delayed feature for an angry customer. Sensible in the moment. But other work gets interrupted, other clients start escalating, and the people whose priorities just got reshuffled feel it. Do that a hundred times over a few years and you get an organization shaped like a formless creature — five hands, three legs, each pulling a different way, none aware of the others. Not from bad intent. From design that was left unowned and unexamined.

Which is the principle, and it's the one Deming pointed at decades ago: a bad system will beat a good person every time. Or, as the systems thinkers put it, all organizations are perfectly designed to get the results they get. If you want ten times the performance, you don't get it by demanding more from people inside a structure that fights them. You get it by redesigning the structure.

That reframes what "high performance" even means. It isn't a universal score. An organization competing on efficiency needs stability and predictable flow; one competing on speed needs fast learning loops and clear end-to-end ownership; one competing on innovation needs room to cross boundaries without friction. Strategy sets the intent, org design provides the structure, and performance emerges from the fit between them. When leaders feel that fit slipping, they reach first for the familiar dashboards — KPIs, OKRs, RACI charts, stage gates. Every one of those tools was almost certainly in play during that first HealthCare.gov build. None of them addressed the design. Some of them made it worse.

Why this matters more, not less, with AI

It would be convenient if AI let us skip the org design work. It doesn't. Org design drives performance regardless of AI — and AI can accelerate what's happening, but it can't make a misfit structure perform. Someone still has to choose a direction, work with people, and own the design.

Bolt agents onto fragmented ownership and narrow mandates and you don't get a redesigned organization — you get the old traffic jam, faster. Everyone's engine revs louder; the cars still don't move, because the roads are still laid out wrong. It's the Ferrari effect at scale, and the DORA data keeps finding it: hand AI to a struggling system and local speed goes up while delivery gets messier. The design flaw was already there. AI just made it louder.

That's why Paula's line in the book lands where it does — after the leadership team walks the floor, sees the misfit, and realizes AI is not fairy dust: first design, then AI. The move is counterintuitive. To adopt AI strategically, you start by putting AI aside: decide what the organization is actually trying to improve, and what in the current structure makes that unlikely. Only then does AI become a lever worth pulling.

Design it, or it designs itself

Your organization is being designed either way — every meeting you add, every mandate you narrow, every quick fix you bless. The only choice is whether that happens deliberately or by accident. The ten engineers who rescued HealthCare.gov weren't better people than the fifty-five firms. They were placed in a better system.

If you want the whole argument — the stories, the map, and the eight principles that follow — 10X ORG is on Amazon, a bestseller in Management Strategy and several other categories, in Kindle, paperback, and hardcover.

Happy Friday, all.