Ask almost anyone in a data team what governance feels like and you'll get some version of the same answer: it's the brake. It's the committee that has to approve things, the classification exercise nobody wanted, the catalogue nobody updates, the department of no. Governance is the friction you tolerate, the tax on doing the interesting work, the reason the exciting project takes three months longer. I believed a milder version of this myself for years — that governance was necessary, the way flossing is necessary, and about as inspiring. I've started this year working in an environment where that framing isn't just wrong, it's dangerous — and it has permanently changed how I think about what governance is for. Governance isn't the brake. It's the steering wheel. And a vehicle with no steering wheel isn't safe because it's slow; it's terrifying at any speed at all.

The metaphor, and why it matters

Let me make the metaphor precise, because the whole argument lives in it. If governance were really a brake, then the trade-off everyone assumes would be true: more governance, less speed; less governance, more speed. Loosen it and you go faster. That model is why governance is resented — it's cast as the enemy of velocity, and in a world that rewards shipping, the enemy of velocity is the enemy.

But a brake isn't what governance actually does. A brake only ever slows you down; its entire purpose is subtraction. Governance's purpose is direction — knowing what data you have, what it means, who owns it, where it came from, who's allowed to see it, and whether you can trust it. That's not subtraction from speed. That's the thing that lets speed be useful rather than merely fast. A car without a steering wheel doesn't go faster for the lack of one. It goes into a wall. Remove governance and you don't get a quicker data organisation; you get one moving at speed with no idea where it's pointed, which is not the same as progress even though it feels like motion.

The organisations that resent governance as a brake have, I think, simply never experienced good governance. They've experienced bad governance — which really is a brake, and I'll come to why — and reasonably concluded that the whole category slows you down. But that's like driving a car whose steering wheel is seized and concluding that steering wheels are what stop cars moving. The problem was never the wheel. It was that theirs didn't work.

What safety-critical work teaches you fast

I've spent the start of this year around genuinely safety-critical infrastructure — the kind of environment where a data error isn't an awkward number in a report, it's a consequence in the physical world that people's safety depends on. And in that environment, the brake framing collapses in about a day, because the relationship between governance and speed is visibly, undeniably the opposite of what the brake model claims.

Here, you cannot move fast without governance, because moving fast on data you don't understand, can't trace, and can't trust isn't fast — it's reckless, and reckless gets stopped hard the moment anyone responsible notices. The teams that ship quickly in this environment are precisely the ones with excellent governance: they know exactly what their data means and where it came from, so they can build on it with confidence, at pace, without stopping every ten minutes to check whether the ground under them is solid. Their governance is what lets them run. The teams with poor governance don't move fast and loose; they move slowly and fearfully, because every step is over ground they can't trust, and fear is slow.

That's the insight I wish I could transplant into every organisation that thinks governance is the enemy of speed: in any environment where being wrong has real consequences, governance is what makes speed possible, not what prevents it. Safety-critical work just makes visible, at high contrast, a truth that's equally real in a marketing database — the consequences there are simply quieter, so the lesson takes longer to arrive.

Why bad governance earned the bad reputation

I promised I'd defend the resentment, because it isn't stupid — it's a correct response to a real thing. Bad governance genuinely is a brake, and most people have only ever met the bad kind. So it's worth being precise about what makes governance the seized-up wheel rather than the working one, because the difference is the whole game.

Bad governance is governance as gatekeeping — a committee whose job is to say no, a process that exists to slow requests down, a set of controls imposed by people far from the work onto people doing it, justified by risk and measured by how much it prevents. That kind of governance really does subtract from speed and add nothing anyone can feel, which is exactly why it's hated. It's a brake wearing a steering wheel's clothes.

Good governance is governance as enablement — infrastructure that makes it easier to do the right thing than the wrong thing. A catalogue that means you can find and trust data instead of rebuilding it from scratch. Lineage that means when a number looks wrong you can trace it in minutes instead of days. Classification that means you can confidently open data up because you know what's sensitive and what isn't. Ownership that means every dataset has someone who answers for it. None of that slows the work down. All of it speeds the work up, by removing the fear and the rework and the endless re-litigation of "can we trust this?" Good governance is the steering wheel: the thing that lets you commit to a direction at speed because you can actually control where you're going.

The distinction in one line: bad governance asks "how do we stop people doing the wrong thing?" and good governance asks "how do we make the right thing the easy thing?" The first is a brake. The second is a steering wheel. And organisations that have only met the first can be forgiven for not believing the second exists.

The same organisation, both kinds

Let me make this concrete, because "good versus bad governance" stays abstract until you've watched the same decision go both ways. Picture an analyst who needs customer data for a new piece of work — an ordinary Tuesday request, the sort that happens a hundred times a week across any data organisation.

Under the brake version, here's her Tuesday. She doesn't know if the data she needs exists, so she asks around. Someone thinks there's a customer table somewhere but isn't sure it's current. To get access she files a request with a governance committee that meets fortnightly, so she waits. When access finally comes, she can't tell which of the four customer datasets is the authoritative one, or what "customer" even means in each, so she picks one and hopes. Two weeks later her analysis is questioned because she used the wrong source, and the whole thing goes back to the start. At no point did governance help her; at every point it either blocked her or failed to inform her. That's a brake, and she's right to resent it — it added friction and subtracted nothing but her time.

Under the steering-wheel version, the same Tuesday: she opens the catalogue, finds the certified customer dataset in a minute, sees its plain-language definition and who owns it, notes it's classified in a way that permits her use, and starts work that afternoon on data she can trust — because the lineage tells her where it came from and the certification tells her someone stands behind it. Same analyst, same request, same organisation even. The difference is entirely whether the governance was built to gate her or to equip her. The steering-wheel version didn't slow her down for safety's sake; it made her both faster and safer, which the brake model insists is impossible and which good governance does routinely. Once you've watched that contrast play out, the brake framing becomes impossible to take seriously.

When governance genuinely does slow you — and why that's still right

I don't want to oversell, because there's an honest exception the steering-wheel enthusiasts skate past, and skating past it is how you lose the credibility of the whole argument. Sometimes good governance does slow a specific thing down, on purpose, and rightly. When someone wants to use highly sensitive data in a novel way, or push a consequential change to a critical system, a good governance process makes them stop and think, get a second set of eyes, and proceed deliberately. That is slower than just doing it. And it's correct that it's slower, because the thing being slowed is precisely the thing that should be approached with care.

But notice this isn't the brake model — it's the steering wheel working exactly as designed. A steering wheel also slows you down for the hairpin bend; that's not a failure of steering, it's the entire point of having control. The distinction is that good governance slows you selectively, at the genuinely risky moments, so that everywhere else you can move faster with confidence — whereas bad governance slows you uniformly, at every moment, risky or trivial alike, out of a blanket institutional nervousness. Good governance is a scalpel: friction applied precisely where consequence lives and nowhere else. Bad governance is a blanket thrown over everything. Both involve some friction; only one of them aims it. And a board that understands the difference stops asking "how do we reduce governance friction" and starts asking the better question: "is our friction aimed — landing on the risky things and lifting off the safe ones — or is it just a blanket?"

The tooling is the easy part

A word on tools, because this is where organisations reliably start — and reliably start wrong. Modern platforms give you real governance capability: a catalogue, lineage, classification, sensitivity labelling, a place to record ownership. On the Microsoft stack I work in, Purview is the home for a lot of that, and with Fabric having just gone GA, the pressure to get governance right is only climbing, because a platform that makes it that easy to create data makes ungoverned sprawl just as easy. The tools are good and getting better.

But the tools are the easy 20%. I've watched organisations buy the governance platform, switch it on, and discover a year later that it changed nothing — because governance isn't a tool you install, it's a set of human agreements the tool merely records. The catalogue is empty because nobody agreed to fill it. The classifications are wrong because nobody agreed what "confidential" means here. The ownership is unassigned because nobody agreed to own anything. The tool faithfully reflects the absence of the agreements, and everyone blames the tool.

Governance is not a product you buy; it's a set of agreements you reach, that a product then helps you keep. Buy the tool without reaching the agreements and you've installed a very expensive record of the conversations you never had.

Which means — and by now you'll have seen this coming — that governance is fundamentally a communication problem wearing a technical costume, the same shape I keep finding under everything in this field. The hard part of governance isn't the lineage graph; it's getting two departments to agree what "customer" means so the lineage graph describes something true. It's a negotiation, often a mildly political one, about meaning and ownership and access — and the tool can only ever hold the agreement, never make it. This is why governance projects led purely by technologists so often stall: they treat as a tooling problem what is really a problem of getting humans to agree, out loud, on things they've been comfortably disagreeing about for years.

What changes when you believe this

So what actually changes if you take the steering-wheel view seriously rather than the brake view? Quite a lot, and all of it practical.

  • You stop pitching governance as risk reduction and start pitching it as speed. The brake framing sells governance on fear — do this or bad things happen. The steering-wheel framing sells it on capability — do this and you can move faster, open data up more, trust your numbers, build without rework. The second pitch is both more honest and vastly more fundable, because it promises something people want rather than merely threatening something they dread.
  • You design governance to enable, not to gate. Every governance decision gets tested against one question: does this make the right thing easier, or does it just make the wrong thing harder? The first is a steering wheel; the second is a brake. Aim relentlessly for the first, and the resentment mostly evaporates, because people don't resent the thing that speeds them up.
  • You put the human agreements first and the tools second. You spend the early effort on the conversations — what do our words mean, who owns what, what's actually sensitive — and only then wire the tool to hold what you agreed. Backwards is the default and backwards is why so many governance programmes quietly die.
  • You measure governance by velocity, not by control. A governance function is succeeding if teams are shipping faster and more confidently because the ground under them is trustworthy — not if the committee blocked a satisfying number of requests. Measure the enablement, not the prevention, and you'll build the enabling kind.

The steering wheel, one last time

I'll close where I started, because the reframe is the whole point and it's worth landing cleanly. The reason governance has such a miserable reputation is that we've described it, sold it, and too often built it as a brake — a thing whose job is to slow the organisation down for its own protection. And a brake is genuinely no fun; nobody was ever inspired by a brake. But that was always the wrong picture. Governance, done well, is the steering wheel: the unglamorous, absolutely essential thing that lets an organisation go fast and end up somewhere it meant to go, instead of fast into a wall.

Working where being wrong has real consequences has made this vivid for me in a way a slide deck never could. But the truth generalises all the way down: no organisation is actually made faster by not knowing what its data means, who owns it, or whether it can be trusted. That's not speed. It's just velocity without direction, which is the most dangerous thing on any road. The steering wheel was never slowing you down. It was the only reason it was ever safe to press the accelerator at all.