Most of the AI-disruption conversation points inward: which enterprise workflows get automated, which functions get leaner, which roles get redesigned. The most under-covered disruption points the other way — at the firms that sell advice to the enterprise. The billable hour, the seat license, and the retainer are all the same instrument: a proxy that meters human time because, until now, human time was where the value came from. Agents break that instrument. When a piece of analysis that took 60 hours takes six, the hour collapses as a measure of worth — and every business built on billing time faces one choice. Reprice, or shrink.
This is no longer a forecast. It is a pricing shift already visible in the numbers.
The proxy is failing in public
The professional-services industry is repricing itself in real time. McKinsey reportedly now derives more than 30% of global fees from outcome-tied arrangements rather than time-and-materials. BCG's leadership has said AI-related work would grow from roughly 20% of 2024 revenue toward 40%. Bain puts AI- and tech-enabled work near a third of its business and climbing. EY leaders describe, openly, a "service-as-software" future — advice delivered as a product, not a timesheet.
Then the platform vendors walked onto the same field. On July 2, Microsoft launched Frontier Company — $2.5 billion and roughly 6,000 embedded experts, described not as a consultancy but as an "outcome-driven engineering organization." Two days earlier, AWS stood up a $1 billion embedded-engineer unit on the same premise. When the hyperscalers start selling outcomes rather than licenses, they are not entering the software market. They are entering the advisory market — and they are pricing it the way the incumbents are only beginning to.
The direction is unmistakable, and it runs one way. Time is deflating as a unit of value at exactly the moment AI is inflating the value a single expert can produce. Those two curves cross, and where they cross, the hours-based model stops being a business and starts being a liability.
Why AI is fatal to time-based pricing specifically
Every prior productivity wave made professionals faster, and the hourly model absorbed it — faster work simply meant more matters billed. AI is different in a way that matters at the unit level: it decouples output from hours rather than compressing hours per unit of output.
A tax memo, a market model, a due-diligence pack, a controls assessment — each once required a human to spend the hours, and the client paid for the hours as a stand-in for the judgment embedded in them. An agent that produces a defensible first draft in minutes does not make the judgment less valuable; it makes the hour a lie about where the value sits. Continuing to bill by time in that world means one of two confessions: either the firm is billing for hours it no longer spends, or it is passing the entire productivity gain to the client for free. The first is untenable. The second is a slow liquidation.
From billable hours to billable assets
The way out is not to defend the hour. It is to change what gets sold. The unit of value moves from effort to asset — from time spent to governed, reusable intellectual property that compounds across clients and quarters.
The shape of that shift is a stack, and finance is the clearest place to see it:
- Reason. Frontier models supply judgment on demand — bought, interchangeable, cheapening every quarter. This is exactly where not to seek advantage; it is the commodity layer, available to every competitor on identical terms.
- Delegate. Agents execute multi-step workflows under human oversight. Advantage accrues to whoever encodes the process well — but a well-run agent is still, on its own, a cost.
- Manufacture. Each engagement becomes governed, reusable IP — an ontology, a diagnostic, an agent suite, a benchmark — that is owned and priced as an asset. This is where hours become assets, and where the margin stops leaking to the client.
Spend that expires is a cost. Spend that becomes an owned, auditable capability is equity. A firm that runs a diagnostic 200 times and sells the 201st in an afternoon is no longer billing time; it is amortizing an asset. That is the whole game.
What it means for whoever owns the P&L
For the firms that sell advice, the reallocation is blunt: stop measuring utilization and start measuring reuse. The metric that predicts survival is not hours billed per head; it is the share of revenue carried by reusable assets versus bespoke effort — and the rate at which the former is growing. A partnership that cannot answer "what did we manufacture this quarter that we can sell again next quarter" is optimizing a melting asset.
For the enterprises that buy advice, the mirror move is to stop paying for hours and start contracting for outcomes and reusable artifacts — the diagnostic that stays, the agent that keeps running, the benchmark that updates — so the firm's productivity dividend accrues to the buyer instead of evaporating into a timesheet.
And for everyone, the warning in Microsoft's $2.5 billion is the same: the platform vendors will sell outcomes over the heads of anyone still selling hours. The productivity gain from AI is real and large — and most of it currently leaks: 85% of employees already save one to seven hours a week with AI, but only 14% consistently turn that into positive net outcomes, and roughly 40% of the time saved is lost reworking the output (Workday, Jan 2026). The only open question is who captures it — the firm that turns expertise into an owned, governed, priceable asset, or the client and the platform that route around the one still counting time. The model was never the moat. Neither was the hour. What you manufacture and own is.