The Efficiency You Can’t Keep
AI, the prisoner’s dilemma, and the fiction that a project team has one goal.
Isn’t this just normal competition with a fancy name?
Pretty much — and that’s the point. An economist would call it price competition passing a productivity gain to buyers, and they’d be right. The prisoner’s dilemma frame is useful because it lets you feel the mechanism instead of nodding at a definition. It shows you exactly why the outcome everyone would prefer — holding the gain together — is the one the group can’t reach. Calling it competition tells you it happens. Calling it a prisoner’s dilemma tells you why you can’t stop it by wishing.
If the gains just get competed away, why adopt AI at all?
Because sitting it out is the one move that guarantees you lose. The gain gets competed away whether or not you personally adopt — the firm across town adopts, drops its price, and now you’re matching a number you have no efficiency behind. Adopting doesn’t hand you a durable edge; it keeps you in the game at the new cost floor. Refusing just means you’re the last one paying full freight for work a machine now does in an afternoon.
Doesn’t the firm that cuts price first come out ahead?
For a while, yes — which is exactly why somebody always moves first. The first firm to pass the savings along wins work and takes share before the rest react. But “for a while” is the whole story. The others match, the floor resets, and the early lead dissolves into the new baseline. First-mover advantage on price is real and temporary. If your entire strategy is being first to the bottom, you’ve picked a race whose only prize is getting there first.
Can’t I just keep my AI advantage quiet and hold it longer?
You can try, and you’ll buy yourself some time. But the tool you’re using is the same tool your competitor can buy tomorrow, so the secret has a short shelf life. And the market reads price, not press releases — the moment anyone uses the efficiency to win on cost, the signal is out whether you announced it or not. Quiet buys months. It doesn’t buy a moat.
Some firms are building their own tools. Are they the exception?
A handful of very large firms will build real capability, and a few of those will build something good. But for the industry as a whole it isn’t realistic. The same competency-and-capacity problem that makes you hire out structural engineering makes standing up a serious AI shop between change orders a losing proposition. Most firms that say they’re “building their own” are configuring someone else’s platform — which is fine, but it’s renting with extra steps. The tool layer is where that value concentrates, and it mostly sits outside the firm.
You’re telling me not to lower my price. Clients will walk. How is that realistic?
I’m not telling you to refuse to move on price — the market will move with or without your permission. I’m telling you not to build your firm’s identity around a price you can’t defend, because a number anyone can match is not a reason to hire you. Compete on price where you have to, and win on the things that don’t show up on a rate sheet. If the only answer to “why you?” is “we’re cheaper,” you’ve given the client permission to leave the second someone’s cheaper still.
If everyone rents the same tools, where does any real advantage come from?
From the part of the work AI can’t commoditize. Judgment about which risk actually matters. The trust that makes an owner believe your bad news. The willingness to own an outcome instead of documenting your way out of it. The tool makes everyone faster at producing the work; it does nothing for the thing the owner is actually buying, which is a building, a night’s sleep, and someone accountable when those two are in tension. That’s where the durable advantage lives, and it always has. AI just clears away the busywork that used to hide it.
Does this kill the subconsultant model? Should I pull structural — or anything — in-house?
It loosens the model more than it kills it. The reason you hired out was competency and capacity, and when a small team plus a good tool can supply both, the make-versus-buy line moves and more work comes inside. But “can” isn’t “should.” Pulling a discipline in-house means owning its risk, its liability, and its worst day — not just its good output. Some firms will integrate and win. Others will find they bought a headache they used to rent. Run the math on the whole thing, not the efficiency alone.
Where do owners land in all this?
Mostly better off, and slightly more exposed. The competed-away efficiency flows to them as lower cost — that’s the whole mechanism. But a race to the bottom on price is not the same as a race to the top on outcome, and owners who chase only the lowest number will find that the value that got competed away included things they actually wanted. The sharp ones use the savings to buy more judgment and accountability, not less.
What do I actually do Monday?
Adopt the tools — being the last holdout is the only guaranteed loss. Then stop treating AI as a margin play and start treating it as a way to point your best people at the work that still has a price. And get honest about your answer to “why you?” If it’s a number, it’s borrowed. If it’s judgment, trust, and accountability, it’s yours to keep.
