A services firm that bills by the hour has a structural problem with AI, and it is not a skills problem. If AI lets the firm deliver an engagement in a tenth of the hours, the firm just destroyed ninety percent of its own revenue on that deal. No partner gets promoted for that. So the incumbents will adopt AI exactly as far as it protects utilization and not one step further.
You can see this in how quotes behave. When AI enters the conversation, hourly estimates from traditional firms shrink at the margin, a modest trim that acknowledges the tools exist. Meanwhile the actual frontier of what one engineer with agents can do has moved by an order of magnitude. That gap between the quoted improvement and the possible improvement is not ignorance. It is the business model doing exactly what it is designed to do.
Enterprises feel this most acutely on legacy work. Rewrites and replatforming projects still get scoped in thousands of hours, priced as if code is produced the way it was in 2020. The buyer suspects the number is wrong, the vendor knows the number is wrong, and neither side's contract has a way to say so. Fixed-bid helps a little, but the deeper fix is pricing outcomes, where the vendor keeps the efficiency as margin instead of surrendering it as lost billings.
That is the opening for a new kind of firm, AI-first from the ground up, built by people who trust agents with real work because they build with agents every day. Services are not a detour for agent companies, they are how the category gets built, which is why the go-to-market for agent platforms is services-shaped and why the forward-deployed model keeps winning in enterprise AI. The firm sitting inside the customer's hardest problems, unburdened by an hourly P&L, will run circles around the incumbent whose margins depend on the work staying slow.
The hourly model survived every previous wave of tooling because tools made engineers somewhat faster. Agents do not make the work somewhat faster. They change what a unit of work costs, and pricing models built on the old unit do not get to vote on that.
Key takeaways
- An hourly services firm that uses AI to finish work ten times faster loses ninety percent of its revenue on that engagement, so its incentives run directly against the technology.
- This is why AI-assisted quotes from incumbent firms shrink at the margin instead of by the order of magnitude the tools now allow.
- The winners will be AI-first services firms that price outcomes instead of hours and treat efficiency as margin rather than lost billings.
FAQ
Why don't large consulting firms just adopt AI aggressively themselves?
Their revenue model punishes them for it. When you bill by the hour, every efficiency gain is billable time you erased. Firms will adopt AI in the ways that protect utilization, which means incremental gains, not the structural rethink the technology enables.
What should enterprises do differently when buying development services?
Push for outcome-based pricing and ask hard questions about how the vendor's engineers actually work with AI. A quote denominated in thousands of hours is increasingly a signal that the vendor's process, not the problem, is what you are paying for.
Sources
Related Essays
From Agent to Platform: Why the GTM Is Services-Shaped
There is no winner-take-all platform in agents. The GTM is services-shaped, the harness is commoditized, and coding-agent companies have a narrow window to pivot before the floor falls out.
The Forward-Deployed Model Is the Only One That Actually Works
Every agent needs to be personalized and context engineering is hands-on, so the vendor who embeds engineers inside the customer has a structural advantage. Self-serve ships shelfware.
Follow No Goose Absolutely
Design partner programs prevent enterprise AI companies from building three products when they should have built one.
Drowning in pull requests that need your review? Try Tembo Review, a beautiful AI-assisted PR review tool unlike anything you’ve used.