A letting agent in the UK typically manages around 100 properties. At Dwelly, which buys letting agencies and rebuilds their operations around AI, one member of staff manages more than 300, by the company's count.
In Hungary, Ominimo started writing motor insurance with models that price each risk on hundreds of variables. Profitable in its first year, 7% of the market in twelve months.
Across professional services as a whole, meanwhile, organisation-wide AI use has nearly doubled in a year... to 40%. Six in ten firms still haven't got the same technology past scattered individual use.
Why does AI transform one services business and stall in another? Look at how each one gets paid. Letting agencies take a percentage of rent collected. Insurers are priced on premium. Neither sells time, so every hour the software removes is pure profit, and nobody holds a partner meeting about why distributions fell.
Whether AI helps or hurts a services firm depends almost entirely on how the firm charges. The adoption curve everyone is arguing about is really a pricing curve.
The two economies
Take one firm billing by the hour and one charging for the outcome, and hand them the same AI.
Under hourly billing, a tool that turns four hours of work into one has just destroyed three hours of revenue. The firm can pass the saving to the client and shrink, or win enough new work to fill the hours... which is possible, cheaper delivery does make more matters worth taking... but that is running faster to stand still, while the outcome-priced firm banks the same gain on the work it already has. Either way, the technology arrives as a cost problem wearing the costume of an efficiency gain.
Under outcome pricing, the same tool turns the same work into margin. The price was agreed for the result, and delivering it in a quarter of the time makes the engagement four times more profitable.
The rule that locks it in
In law, a conduct rule turns the incentive problem into a hard constraint. The American Bar Association said it plainly in 2024... a lawyer billing hourly must bill the time actually spent. If AI compresses six hours of document review into forty minutes, the client is billed for forty minutes.
So an hourly firm cannot quietly keep the gain even if it wanted to. It has to hand the efficiency straight to the client, at exactly the moment it is paying for the technology that produced it. And that, more than culture or inertia, is why huge AI budgets keep producing so little visible change inside hourly firms.
Where AI is already winning
The AI-native companies that have actually taken off in the past two years have one thing in common, and it is not the industry they picked.
Plaintiff law works on contingency, one of the oldest outcome arrangements in the professions, and it is the one corner of legal where AI adoption looks nothing like the rest. Eve, which sells only to plaintiff firms, has signed more than 1,700 of them in about two years, 500 in the last three months, and firms on it settle cases up to 60 days faster... the company's own numbers, though investors valued it at over a billion dollars last autumn, when the firm count was barely a quarter of what it is now.
Customer support is priced per resolution or per seat. Crescendo reports gross margins of 60 to 65%, roughly four times what a traditional call centre makes. Add debt recovery, claims handling, recruitment on placement fees, medical billing. All paid for a result rather than elapsed time, all with an AI-native entrant growing quickly.
None of those companies is cleverer than a mid-sized accounting firm. They just sit on the side of the invoice where efficiency shows up as profit.
The most interesting incumbent has worked this out. About a third of McKinsey's revenue now comes from underwriting client outcomes rather than charging advisory fees, and Bob Sternfels says his hope is that it crosses a majority before his time as global managing partner ends. The same firm counts 25,000 agents alongside 40,000 people. It is tempting to see the pricing shift as a result of all those agents. I think it is the other way round... the pricing shift is what makes the agents possible. You cannot scale 25,000 of them inside a business that sells hours, because every agent you deploy shrinks the invoice.
My guess is that the large firms that pull this off will fix how they charge first and buy the technology second. Do it in the reverse order and you end up running pilots for three years.
What I think happens next
The obvious objection is that people have been declaring the billable hour dead since before the financial crisis, and it has buried every one of them. Fair. It survived because the efficiency gains were marginal and clients could not see them. This time the work is compressing by multiples rather than percentages, and clients are watching it happen on their own screens, because they use the same tools. Around 62% of legal work is still billed by the hour, with accounting and consulting not much different, while 71% of legal clients say they would rather pay a flat fee for the whole matter, and in one survey every single firm said AI is affecting its pricing... yet only a third had changed their pricing model.
So, three predictions. At least one large hourly firm announces a serious move to fixed and outcome pricing and frames it as a technology strategy rather than a discount, and everyone else watches its revenue per partner before following. The middle gets squeezed hardest, as usual - hourly clients, no capital for a proper build, not enough scale to absorb a year of falling revenue while the pricing changes. And the professional bodies eventually have to look at the conduct rules, because a rule that requires firms to bill the time actually spent quietly guarantees that on hourly work, the profession keeps none of the value AI creates.
For anyone running an hourly firm, you do not have to blow anything up to start. Pick the work you do over and over... the matter types you could price in your sleep... and move those to fixed fees. The conduct rules stop biting there, and the hourly book pays the bills while you find out what the new model breaks.
Already tried it? Moved to outcome pricing, or attempted it and watched it go wrong? Reply and tell me what happened. I read everything, and the sharpest replies shape the next edition. Subscribe and every one lands in your inbox.
See you next time.