The billable hour was always a proxy for value. AI just broke it.

You quote a contaminated land assessment at forty thousand dollars. You build the number the way you always have: roughly how many hours, at roughly which charge-out rates, plus a margin you do not say out loud. The team gets it done in nine days instead of three weeks, partly because one of your seniors is now quietly running half the desktop review through an AI tool that reads the historical records faster than any graduate ever could. The work is good. The margin is excellent. And somewhere underneath the relief there is a flicker of worry you would not raise at the partners’ meeting: if the client knew how quickly that came together, would they feel they had been gouged?

Hold onto that worry. It tells you something true about how your firm actually makes money, and it points at a problem that is going to arrive a great deal faster than most firm owners expect.

Why the hour became the price

The billable hour survives because it solves a problem the client cannot solve any other way. A client cannot directly assess the quality of your senior contaminated land specialist, and they cannot judge whether your assessment is any good until long after they have paid for it, if ever. So they reach for the one signal they can see: how much expert time went in, at what rate.

There is an old parable about an engineer called in to fix a generator that had defeated everyone else. He listened to it for a while, chalked a single mark on its casing, told them which windings to replace, and sent a bill for ten thousand dollars. Asked to itemise it, he wrote: one dollar for the chalk mark, nine thousand nine hundred and ninety-nine for knowing where to put it. It is usually told as a joke about expertise. It is really a joke about pricing. The client pays for the knowing. The invoice has to be denominated in something countable, so we count the chalk mark.

Two surrogates are hiding inside the hour. The hourly rate stands in for the quality of the person: the principal at four hundred dollars an hour must be worth more than the graduate at one fifty. The number of hours stands in for the substance of the work: a report that took two hundred hours must hold more value than a memo that took ten. Both are proxies. Neither measures the thing it claims to measure. We tolerated the gap because, for a long time, the proxies tracked closely enough to the truth, and because nobody had a better unit.

What AI actually breaks

When a senior plus a capable model does in two days what used to take a small team three weeks, the value of the output is unchanged. The client still gets an assessment they can lodge, a liability they can quantify, an approval they can bank on. But the hours have collapsed. The proxy has detached from the thing it was standing in for, and it has detached in the worst possible direction for a firm that prices on time: downward. Every efficiency you find now reads, on your own invoice, as a reason to charge less.

This is not a forecast. It is already the live problem at the top of the market, and the firms with the most invested in the billable hour are the first to admit it. Industry reporting puts the Big Four and the major strategy houses well past ten billion US dollars of AI investment since 2023. PwC alone announced a billion-dollar, three-year generative AI programme and became OpenAI’s largest enterprise customer. KPMG signed a roughly two-billion-dollar alliance with Microsoft. EY’s global managing partner for growth, Raj Sharma, has told Business Insider that AI agents are pushing the firm toward what he calls a “service-as-a-software” model, where clients pay on outcome rather than hours. PwC’s US leadership has openly signalled a shift away from hourly billing.

When the firms whose entire economics rest on selling leveraged teams of human hours start talking about pricing on outcomes, that is not marketing. It is the people with the most to lose reading the same gauge you are. The change is slow even for them: one late-2025 analysis found that even at McKinsey, only about a quarter of fees are tied to outcomes. The model is breaking faster than anyone has worked out how to replace it.

PwC’s own 2026 AI performance study, which surveyed more than twelve hundred senior executives, sharpens the point. It found that around three-quarters of the measurable economic value from AI is being captured by roughly a fifth of organisations, and that the firms pulling ahead are pointing AI at growth, not at cost-cutting. The ones using AI only to shave delivery cost are quietly handing the savings to their clients. The ones winning are using it to change what they sell.

Adjacent industries solved this decades ago

Software never made the cost-plus mistake, because it never could. The marginal cost of serving one more user of a SaaS product is close to nothing. Cost of goods sold often sits under twenty per cent of revenue, sometimes in the low single digits. A software firm that priced on what it cost to serve the next customer would price at almost zero and die. So software priced on value from the start. The working rule of thumb is to capture something like ten per cent of the value you create for the customer and let them keep the rest. The cost of production never enters the conversation.

Hardware does the same in plain sight. Nobody buys an iPhone by asking what the bill of materials cost to assemble. That number is unknowable to the buyer and irrelevant to the price. You pay for what the thing is worth to you. Apple’s margin is not a secret the customer resents; it is the point.

The contrast with your firm is exact. Software and hardware decided what their work was worth to the buyer and priced to that. Consulting, for all its talk of value and partnership, still mostly prices cost-plus: count the hours, apply the rate, add a margin. We have spent decades describing ourselves as advisors who sell judgement while invoicing as contractors who sell time. AI is the event that finally makes the contradiction expensive.

The trap waiting for you: a standardised output has no floor

Here is where the optimistic version of this argument breaks down, and where you need to be honest with yourself.

Value-based pricing implies a band. The floor is your cost, the ceiling is the value you create, and you capture a slice somewhere in between. But where you land in that band is not decided by the value. It is decided by how substitutable you are. Software sits near its ceiling because the solution is genuinely differentiated, and because the buyer can perceive the difference before paying: they demo it, trial it, compare the integrations. Perceived differentiation is what holds price off the floor.

An environmental deliverable fails on both counts. The output is normalised to documented process. The sections of a contaminated land report are effectively prescribed by the guidelines, the NEPM and the state EPA framework, so a compliant report from one competent firm looks like a compliant report from any other. And you cannot demo it. The client cannot assess quality before engaging, and after the fact the only quality signal they get is that the regulator accepted it. A credence good, with a compliance-shaped ceiling, sold by suppliers the buyer treats as interchangeable, prices toward the floor.

AI makes this worse, not better. Process-normalised, document-heavy work is exactly what the models do cheaply, so the floor drops sharply. And because the offer is undifferentiated, price follows the floor down. A competitor who adopts AI and stays on hourly billing can underbid you on the same standardised report and still make margin, because their cost base just fell. That is a race to the bottom, and it happens to you whether or not you choose to enter it. Productising the report does not save you here. It just makes you a cheaper commodity.

The way out is to sell a different thing

You escape the floor by abstracting one level up. Stop selling a compliant report and start selling the lowest total cost of carrying the contamination constraint.

For the developer, the constraint cost was never your fee. It is investigation, plus laboratory work, plus the holding cost of delay, plus the big one: excavation and offsite disposal. Dig-and-dump is where the real money sits, and it is driven by waste classification. Whether material leaves the site as General Solid, Restricted Solid or Hazardous waste swings the disposal bill by multiples, and conservative classification combined with over-excavation can blow a budget by hundreds of thousands of dollars.

Now look at what your pricing model rewards. A commodity consultant minimises their own fee and, in doing so, often maximises the developer’s total cost: standard sampling, conservative classification, recommend digging the lot out. A consultant selling on value minimises the developer’s total cost: delineate tightly so you do not over-excavate, design the sampling so you can defensibly down-classify the waste, weigh on-site treatment, validation, or leaving material in place under a long-term management plan instead of carting it to landfill. That judgement genuinely varies between firms. It is worth real money. And the client can measure it.

The abstraction fixes all three of the commodity problems at once. It raises the value number, because you are now pricing against a six or seven figure disposal bill rather than against the cost of writing a report. It restores differentiation, because two firms produce wildly different total-cost outcomes from the same site. And it makes the differentiation perceptible, because there is a measurable delta against the obvious default. The compliance report had none of those qualities. The constraint strategy has all of them.

Two things to be honest about

The value is hard to quantify going in. This is the real one, and you should not let anyone sell you value pricing as though every job arrives with a clean dollar figure attached. You do not know the full extent of the contamination until you have characterised the site, which is the same uncertainty the developer is hiring you to resolve. Pricing a total-cost outcome before you have seen the ground is pricing blind.

And some of the value you create goes to people who will never pay you for it. When you abstract up to the environmental condition you are protecting, you create real value for the aquifer, the community, the next owner. None of them holds the chequebook. The developer pays you to reduce their cost, their risk and their delay. The developer does not pay you to protect the groundwater better than the regulation requires. That is, in fact, why the regulator exists: it sets a floor of protection precisely because the party paying the bill will not buy more of it voluntarily. So be clear-eyed about which value you can actually invoice. Price on the value the payer captures. The wider public good your work produces is real, and it is part of why the profession is worth being in, but it is not a line item, and a firm that prices as though it were will do excellent work and still get competed to the floor.

The risk you inherit the moment you price the outcome

The instant you price on the developer’s total cost, you have taken on the risk that used to be theirs. And contamination is the canonical “you do not know what is down there until you dig” problem.

This failure is not hypothetical. It is the standard one in any firm that prices physical uncertainty at a fixed number. RCR Tomlinson and Forge Group were both large listed engineering firms with West Australian roots that grew fast and then collapsed after cost overruns on a handful of fixed-price construction contracts. RCR went into administration in late 2018, with around three and a half thousand staff, after a write-down of roughly fifty-seven million dollars on Queensland solar work. The mechanism in both cases was the same: they fixed a price across a physical uncertainty before that uncertainty had resolved, hit conditions worse than they had tendered for, and a couple of bad jobs took down a company turning over a billion dollars or more.

The contaminated land version of this failure is structurally identical, and the exposure is worse. A construction firm that misprices ground conditions loses money on earthworks. An environmental consultant who fixes a remediation price before site characterisation is complete has done the same thing, on a smaller balance sheet, with no subcontractors to share the pain. RCR was turning over a billion dollars and still went under on a handful of bad jobs. For a firm of 10 to 50 people, one fixed-fee remediation where you guaranteed the disposal volume and then found three times the contamination is not a bad year. It is the end of the firm. The abstraction that creates the value and the trap that destroys firms are the same move seen from two sides. The value lives one level up, but so does the irreducible uncertainty. You cannot capture the upside without deciding, deliberately, who carries the downside.

A seven-point plan to get there

None of this resolves with a memo announcing that the firm now prices on value. It resolves as a sequence, and the order matters.

1. Build the write-up bank first, before you touch a price. Have people log the hours they actually worked, AI included, and bank the gap between that and the budget as a personal credit. It counts at their review as efficiency they created, and it cushions the next job that overruns. You reward speed instead of punishing it, you kill hour-padding, and you finally learn what each deliverable really costs. Keep two ledgers and never let the bank rewrite the actuals: the job is costed at hours genuinely worked, the bank only governs how the person is judged. Guard that baseline harder than anything else.
2. Split your book into commodity and constraint. Not every job has value worth pricing. A site and soil evaluation for effluent disposal is a genuine compliance check, and installers often give the work away. Productise it, price it low, let AI do most of it. Save value pricing for the jobs where a real constraint is moving real money. Running one pricing logic across both is the mistake that sinks the transition.
3. Carry the baseline yourself. The client cannot give it to you. A developer cannot tell you what best case looks like, because resolving that is what they are paying for. So you bring the reference point: what a commodity firm would do with this site, and the number the developer would face without your thinking. None of it has to be said plainly. It shows in how you talk about your method and in the delta a client can expect from your approach. Price against that delta, not the hours.
4. Separate priced value from priced risk, and stage-gate. Price the strategy and judgement on value, because that is what is differentiated and worth paying for. Do not fold the physical unknown into a fixed number. Price the investigation phase first. Once the uncertainty has collapsed and you can see what you are pricing, put an outcome price on the works. Never fix a price across an unresolved uncertainty.
5. Cap the downside with shared savings. Borrow the structure construction already uses. Set a floor on your fee, take a share of the savings you create against the baseline, and put a hard cap on your exposure. The developer keeps most of the upside. You capture a slice you can demonstrate, without betting the firm on what turns up under the slab.
6. Turn the cost data into a pricing asset. After a few months the write-up bank has told you two things you never reliably knew: what each deliverable costs to produce now, and how far that sits below what it is worth. That is your floor and your ceiling, measured rather than guessed. Use it to set fixed and value prices service line by service line, as the data earns your confidence. The firm with the honest dataset prices accurately while its competitors are still guessing.
7. Retool the measurement underneath. Stop running the firm on utilisation alone. Report margin per engagement and value delivered alongside it, and let those numbers carry more weight over time. The metric you manage to is the metric that changes behaviour. The endgame is a price the client pays with no hours in sight, and a scoreboard that no longer rewards anyone for burning time.

The cost of doing nothing

The comfortable assumption is that you can wait and watch. You cannot, and it helps to be precise about why.

If a competitor adopts AI and stays on hourly billing, they underbid you on the standardised work and still make margin. The race to the bottom is run with or without your consent, and standing still simply means losing the commodity work to someone faster and cheaper.

If you adopt AI but keep billing by the hour, you hand the entire efficiency gain to your clients as a lower bill. You will have done the genuinely hard work of getting faster, purely to cut your own day rate. You take the disruption and forgo the reward.

If you do neither, you become the most expensive supplier of a commodity in a market whose floor is dropping every quarter, with nothing differentiated to retreat to, because you never built the constraint-optimisation capability that would have justified a higher price.

And the gap compounds. The firms pulling ahead are moving now, building the cost data and the client relationships and the pricing confidence that take years to accumulate. PwC’s own research shows the leaders are already capturing the overwhelming majority of the value. Your good people feel it too. The ones worth keeping do not want to pad timesheets and bill hours for work a machine now does. The firm that rewards judgement and efficiency keeps them. The firm that rewards time loses them.

What changes for you

The firm you are running today has its revenue mechanically tied to how long its people take. Every hour saved is an hour you cannot bill, so efficiency is quietly your enemy, and AI, which is nothing but a very large efficiency, is therefore a threat. That is a miserable position to defend, and it is precisely the position the billable hour now puts you in.

The firm on the other side of this prices on what its work is worth. Efficiency becomes margin instead of lost revenue. The better and faster your people and your tools get, the more you keep. AI stops being the thing that erodes your day rate and becomes the thing that widens your margin.

The billable hour was always a proxy for value. For decades it was a good enough one, so we stopped noticing it was a proxy at all. AI has now pulled the proxy and the value far enough apart that you can no longer pretend they are the same number. The firms that recognise the hour for what it was, a convenient stand-in we mistook for the real thing, will be the ones still setting their own prices in five years. The rest will be explaining to clients why the work that used to take three weeks still costs three weeks.