Is time-based pricing dead across professional services when the work no longer takes time?

The shift

Time-on-task for the drafting, analysis, and production layer — the contract, the deck, the reconciliation, the comp — collapses from scarce billable hours toward near-zero, while the value delivered stays flat or rises. The hourly meter worked as a proxy because effort and value moved together; AI severs the two, so the input a firm has always sold no longer tracks the thing a client actually pays for.

The axioms

1. Expert time is the scarce input, so the unit of sale is time: rate × hours, or a day rate, or a fixed fee back-solved from an hours estimate.
2. Time-on-task is a fair proxy for value — more hours means more work done means more worth delivered, so billing the input is close enough to billing the outcome.
3. The hourly meter is legitimate because it's auditable: the client pays for effort actually spent, and both sides can point at a timesheet.
4. Firm economics run on utilization and leverage — bill junior hours at a markup, keep everyone's calendar full, and margin comes from the spread between salary-cost-per-hour and billed-rate-per-hour.
5. A faster firm earns less on the same job, so there is no incentive to compress the hours — efficiency is quietly penalized by the model.
6. Someone must bear the risk and be accountable if the work is wrong, and part of the fee is really buying that cover — but it's bundled into the hours rather than priced on its own.
7. The client hands over the work and accepts the bill because they trust the specific people, not because the timesheet convinced them.

Invalid axioms

1. The unit of sale is time, because expert time is the scarce input. Drafting, first-pass analysis, and production are exactly the plausible-first-draft work AI is strongest at, and they were the bulk of the billed hours. The habit-trap: firms still quote and invoice by the hour or the day for work whose hour count has cratered, which means the more AI they adopt internally the less they can bill for the same delivered result — margin inverts precisely as capability rises.
2. Time-on-task is a fair proxy for value. The proxy held only while effort and value were coupled. Once a contract or a model that took twenty hours takes two, billing two hours undercharges for identical value, and billing twenty is a fiction the client can increasingly smell. The habit-trap: firms keep defending the meter as "fair" while it now systematically misprices — under on high-value fast work, and over on padded slow work.
3. The hourly meter is legitimate because it's auditable. Auditability assumed the timesheet described real, necessary effort. When the effort is a prompt and a review pass, the timesheet either shrinks to something that looks unserious next to the fee, or it gets inflated to protect the fee — and clients now have their own AI to estimate what the job "should" take. The habit-trap: leaning on transparency of hours as the thing that makes the bill defensible, when the hours no longer explain the value.
4. A faster firm earns less, so don't compress the hours. The perverse incentive only exists while revenue is tied to hours. It was always a drag; AI turns it into an existential one, because a competitor who reprices off the meter can deliver the same outcome faster and capture the value the hourly firm is now leaving on the table. The habit-trap: treating efficiency as a threat to revenue rather than repricing so that speed is something the firm gets paid for.

Unchanged axioms

1. Someone must bear the risk and be accountable when the work is wrong. A signed opinion, an audited statement, a filed contract, a launched campaign that must not defame — these carry liability that attaches to a licensed, insured, reputationally exposed human or firm. A model can't be sued, struck off, or fired. This didn't get cheaper when the drafting did; if anything it got more valuable, because confidently-wrong output is now cheap to generate and someone still has to stand behind the version that ships.
2. Judgment on the specific, high-stakes, non-repeating situation stays scarce. Which clause will actually get litigated, which number management wants told versus the real story, whether this brand can survive this campaign — these don't match a prior pattern cleanly, and they're the part clients were really buying under cover of the hours. The commodity was the draft; the judgment about whether the draft is right for this client was never the thing time measured well.
3. Trust and the relationship gate whether the work happens at all. Clients hand over sensitive numbers, admissions of exposure, and the standing to advise them to specific people built over years — not to whichever tool is cheapest this quarter. The fee has always partly priced this, invisibly. It doesn't move to a model, and it doesn't reset with the tooling.

New axioms

1. When the artifact is near-free, you have to price the outcome, the liability, or the trust directly — and there's no established mechanism for pricing judgment detached from time. The profession spent a century building rate cards, utilization targets, and realization metrics around the hour. Value-based and fixed-fee pricing exist but were mostly back-solved from hours estimates; nobody has a defensible, repeatable way to price "we stand behind this and we're right" without a timesheet underneath it. This is the core unsolved problem, and it's the same one everywhere: the input got cheap, so the price has to attach to something that didn't.
2. When hours no longer justify the fee, the client asks what they're actually paying for — and the honest answer (accountability, judgment, cover) has never been itemized. Firms have hidden risk-bearing and relationship value inside the hours for so long that they can't cleanly quote them. If you can't say what the fee is for once "time spent" is off the table, you can't defend it against a client who now has an AI estimate of the job and a competitor who'll do it for a fraction.
3. Because clients now have the same generation capability, the price ceiling is set by what they can't replicate internally — and that boundary is moving. In-house teams can draft the contract or build the model themselves; what they can't self-supply is the accountable sign-off and the judgment on their specific stakes. How much of professional-services work sits above that line versus below it is a fast-moving call — as models get more reliable and more agentic through mid-2026 and beyond, the line keeps rising, and where it settles determines how much time-billed work survives at all.
4. When the meter comes off, firm economics built on leverage and utilization have no obvious replacement. The whole pyramid — juniors billed at a markup, everyone kept booked — was a machine for converting hours into margin. If the hour isn't the unit, utilization isn't the lever, and nobody has redesigned what "productive capacity" or a "good month" even means when a firm's output is decoupled from its headcount's hours.

Where it breaks

Firms are adopting AI internally to deliver faster (rational) while still billing that work by the hour (invalid) — which means every efficiency gain directly cuts their own revenue, and the better they get, the worse the meter pays. The model now punishes exactly the capability the market is forcing on them, and the collision with the new problem (no mechanism to price the value detached from time) means there's no built landing spot: they can see the meter failing but have nothing repeatable to move the fee onto.

A second collision: clients are being asked to accept fees whose stated justification (hours spent) has quietly become fiction, at the same moment they've acquired their own AI to both generate a first draft and estimate what the job should cost (new). The thing the fee is really buying — accountability, judgment on their specific stakes, trust — was never itemized because it never had to be, so the firm is defending a number it can no longer explain to a client who can now price the parts the firm made cheap.

ai-axioms.vercel.app/axioms/society/time-based-pricing-ai-axioms