Value, Not Hours: Pricing Services When the Work Takes Two Weeks

Part 2 of 3 · The Time and Tokens series

In Part 1, “Time and Tokens”, I argued that building software has collapsed to two inputs (time and tokens) and that the one thing that still costs what it always did is judgment: knowing what to build, why, and how to keep it running. This is the companion problem, and it’s the one that keeps services-firm founders up at night. If the work got faster, what do you charge for it?

Here’s the paradox in its sharpest form. A project that genuinely delivers five million dollars of value to a client used to take six months and a room full of people, and it sold for two million (a price everyone accepted because everyone could see the effort). Now the same project takes two weeks. The value to the client hasn’t moved and the outcome is identical, but send that same two-million-dollar invoice against two weeks of visible work and the client feels robbed. Bill the two weeks at cost instead, and you’ve torched your business and given away the exact thing that made two weeks possible - your experience. The bind is real, and most firms are trying to escape it in the two worst possible directions.

Time was always a lie we agreed to tell

The billable hour was never what the client valued. Nobody ever wanted hours. They wanted the decision made, the system built, the risk removed. Hours were a proxy, a convenient fiction both sides agreed to pretend in, because effort was roughly correlated with value and effort was easy to count. You couldn’t measure the insight, so you measured the time it took to have it, and everyone nodded. Clients run the same fiction on themselves, by the way; Puneet Badlani calls it token maximizing - AI rollouts scored on prompts sent and sessions opened, while nobody checks whether the work got any better. Activity metrics are hours by another name.

“You’ve built a pricing model that pays people to be worse.”

AI rips the proxy off. When the work that justified the hours compresses from months to days, time stops correlating with value at all, and the fiction becomes visible to the client. This isn’t a fringe view anymore; it’s happening at the top of the market. Roughly a quarter of McKinsey’s fees are now tied to outcomes rather than time, and its peers are moving the same way, not out of virtue, but because clients have started asking why a ten-hour job still costs what forty used to. Once a client can see the proxy, the proxy is dead.

And the center of gravity is moving toward services, not away from them. Of roughly 36 million American companies, about 17,000 are software companies; everything else runs on operations, and more than three-quarters of US GDP sits in services. That is where the labs themselves are now heading. Anthropic has stood up Ode, a $1.5 billion implementation company that embeds engineers inside mid-sized enterprises, and OpenAI has a $4 billion joint venture whose forward-deployed engineers do the same work. The people who built the models have concluded that the money is in delivering them; so the pricing question in front of services firms is not a niche problem, it is the main event.

T&M is doubly dead

For those that know me - customers, colleagues, partners - I have always advocated for deliverables-based SOWs. I would turn down T&M if I could, though sometimes I couldn’t. I came from the PBX world, where TDM is still alive and kicking; I’m used to my team time division multiplexing - working on multiple things at once. If my team can work, say, four projects in a week, sometimes all four in the space of a morning, how do I allocate their time? How much effort do I spend on making sure their timesheets are accurate? In a fixed-fee, deliverables-based world, that is just an internal matter, not a customer-facing matter. And so, T&M deserves its own funeral, for two reasons. The eulogy will be short; nobody ever loved a timesheet.

The first is that it now punishes exactly the behavior you want. Under T&M, the firm that uses AI to deliver in two weeks bills less than the firm that plods along for three months, so the honest, efficient firm is financially penalized for its efficiency, and the slow firm is rewarded for its sloth. You’ve built a pricing model that pays people to be worse. That was always a latent flaw in billing for time; AI just turned the dial to absurd.

The second is darker, and it’s the reason I won’t defend T&M even in its better moments - the model structurally invites fraud. When you bill for effort, the client is buying something they fundamentally cannot verify. They see an invoice for hours; they cannot see whether those hours happened, who worked them, or whether the “senior architect” on the timesheet ever touched the project. The incentive to inflate is baked in, and the ability to detect inflation is not. Padded timesheets, phantom staff, slow-walked deliverables billed as diligence. The model doesn’t just tolerate these, it quietly rewards them. Fraud isn’t just bad business; it’s a crime. This isn’t theoretical for me. I’ve seen it happen.

I raise it not to smear an industry I’ve spent my career in, but because it exposes what T&M actually is - a model that asks the client to trust an input they can’t audit, and asks the provider to resist a temptation that has a financial upside. The answer isn’t better auditing; it’s to stop selling the input.

The two ways firms get it wrong

So if not hours, what? Most firms lurch toward one of two answers, and both fail the same way - each one screams at somebody.

Hold the line at the old two-million-dollar price for two weeks of work and it screams at the customer. It reads as extortion, because the client’s instinct is still to price against effort, and the effort they can see doesn’t remotely justify the number. You’ll win the argument on value and lose the relationship on optics.

Price off your now-tiny cost (cost-plus, or a modest markup on two weeks) and it screams at you. You’ve handed the client the entire benefit of your investment in getting fast and good, kept none of it, and taught them that your expertise is worth roughly two weeks of salary. This is also a race to zero; your competitors will drive yours and their prices lower chasing this model. It’s a race no one wins.

Both errors share a root: they anchor the price on the wrong thing. One anchors on a shock number, the other on cost. Neither anchors on the only defensible thing - the value created.

Price the outcome and the judgment

Outcome-based pricing is the obvious escape, and it’s easy to say and hard to do well, so let me be concrete about the structure rather than waving my arms and preaching.

Anchor the price to the client’s value, not your cost or your hours. The question stops being “how long did this take us?” and becomes “what is a solved version of this problem worth to you?” That reframing is the whole move, because it’s the only frame in which two weeks can legitimately be worth a great deal. It’s also the only frame in which your experience (the thing that compressed six months into two weeks) gets paid for instead of given away.

And the value number is not a figure you spring on the client. You build it with them before any work starts, out of parts they can see and argue with: the revenue the launch unlocks, the in-house build they no longer have to staff, the risk you’re moving off their balance sheet. Add those honestly and the total is rarely small. Count what the system avoids as well as what it earns, and a genuinely useful one clears seven figures.

Then put a number on your share, because a share without a number is where firms lose their nerve. The defensible band is five to ten percent of the first year’s value created. On a five-million-dollar outcome, that’s a fee of roughly $250K to $500K. Look at what that does to the conversation. The client is no longer pricing your time at all. They’re paying a fraction of a win they can measure, and clearing a ten-to-twenty-times return in the first year. No CFO turns down a 10× return. Every CFO balks at a big invoice for two weeks of visible effort. Same money, different frame, and the frame is the whole game.

To keep that share from screaming at either party, the better outcome contracts are built from four parts:

A stable base - a fair fixed fee that keeps your lights on and signals skin in the game, independent of the result.
A clear target - the outcome, defined and measurable, agreed in writing before you start, so “success” is never a post-hoc fight. And I repeat - measurable.
Quality guardrails - standards the work must meet, so nobody games the metric.
Shared upside - that five-to-ten-percent slice, paid when the client wins.

The base covers your existence; the upside pays for your judgment. You are finally being paid for the thing that is scarce, rather than the thing that got cheap.

A word on the tempting third structure, mostly to warn you off it: trading fee for equity or a share of revenue. It’s occasionally right (for the one client a year you’d genuinely want on your wall in five years), but a cap table full of dead startup equity pays no salaries and counts for nothing when someone values your firm. Keep it rare, cap it hard, and never let it replace the cash and recurring revenue that actually build a business. This harks back to judgment though. If your judgment tells you that this customer’s startup equity will be the Facebook painter story of your dreams, listen to that judgment. Then read this paragraph one more time.

Because there’s a second stream hiding in the first piece. If bespoke software is dynamic and needs tending (updated as the business evolves, kept from calcifying into the next unupgradeable monument), then that tending is a recurring, outcome-linked retainer, not a one-off. And this is not a rounding error on the build. In my experience, somewhere north of sixty percent recurring revenue is where a services firm stops being valued like a job and starts being valued like an asset. The build is a fee; the retainer (priced to the value of the thing staying up) is the annuity. Firms that only sell the build are leaving the better business on the table. Or put in simpler terms for a business owner - if you can cover payroll and all your bills with your recurring services revenue every month, life is good; imagine what that unlocks for your business and for you personally.

The honest part

I’d be repeating the sins of every breathless pricing article if I pretended this is easy. Outcome-based pricing has been sold as “the future” of professional services for a decade, and most firms still quietly bill by the hour. Not because they’re stupid, but because the hard parts are genuinely hard.

Attribution is hard: did your work cause the client’s revenue lift, or did the market move? Measurement is hard: many of the most valuable outcomes (a risk avoided, a decision improved) resist a clean number. Some clients won’t share the data you’d need to price a share of it. And the risk transfers to you: outcome pricing means you eat it when the outcome doesn’t land, sometimes for reasons outside your control.

None of that is a reason to keep selling hours; it’s a reason to be disciplined. Define the outcome and the measurement together, in writing, before the work starts. Use guardrails so nobody games the metric. Start with a solid floor and a modest share of upside rather than betting the engagement on a single number. Choose, at least early on, the engagements where the outcome is actually measurable, and build the track record that eventually lets clients pay you for trust. Outcome pricing is a capability you develop, not a switch you flip. One customer at a time, train that muscle memory so that by the time your business becomes more than 50% outcome-based revenue, you’ve learned most of the lessons already.

Two sides of one invoice

The two arguments are the same argument, seen from opposite sides of the invoice. For the buyer, software stopped being something you purchase and bend yourself around, and became something you shape and tend. For the seller, services stopped being something you meter by the hour, and became something you price by the value you create and the judgment only you bring. Code is the commodity now (anyone with an API key can produce it). Certainty isn’t, and won’t be for a long time.

In both, the move is identical - stop charging for the thing that got cheap, and start charging for the thing that stayed scarce. Tokens are cheap. Hours are cheaper than they’ve ever been. Judgment (knowing what’s worth doing, doing it fast because you’ve done it a hundred times and standing behind the result) is the whole business now.

So sell that, and price that judgment. The hours were never the value.

Next in this series: Part 3, “The Whole Pool: Who Actually Builds Software Now” (if code stopped being the qualification, who is now best placed to build?). Further reading: on how the major consultancies are restructuring fees around outcomes, TheStreet’s report on McKinsey, BCG, and Bain; for the same proxy problem seen from inside the client’s own walls (usage dashboards standing in for results, the way hours stood in for value), Puneet Badlani, “AI Tool Overload”; and on why the labs are moving into delivery, Gelila Bekele, “AI Value Creation Moves To Services” (Forbes).