共有:
Industry Shift · IT Business Models

Time-and-materials IT
contracting is dying

AI code generation is dismantling the man-month assumptions that have underpinned IT outsourcing for decades. Reports say the collapse became visible in just three months — and it hits both sides of the contract equally.

AI Navigate Editorial·2026.06.22·6 min read
Old model (man-month) 1 feature ≈ 6 months / 3 man-months Cost: high Predictability: high AI AI era Same output days–weeks Cost: low Predictability: volatile This shift became visible in just 3 months
Development timeline compression: old model vs. AI era (conceptual)
01 — The Old World

Why man-month billing worked for so long

The man-month is a unit of measurement that encodes a particular era in computing history — one where the primary constraint on software production was how fast human beings could write and review code. Formalized in the United States in the 1970s and embedded deeply in Japan's systems integrator culture, it became the lingua franca of outsourced IT because it matched the underlying reality of the work.

For decades, the assumption held: a developer working full-time for a month produces roughly the same amount of working code, regardless of who employs them or what tools they use. Variation existed at the margins — a 10x engineer here, a better IDE there — but not enough to shatter the model. A feature that required three people working for six months cost eighteen man-months. This math was stable enough to build contracts around.

Risk allocation was clean. If a project ran long, the contractor absorbed the overrun (or negotiated a change order). If it came in short, the margin widened. Buyers controlled cost by managing scope. Contractors competed on estimation accuracy and delivery reliability. The whole system rested on one unspoken assumption: the labor cost of producing a unit of software changes slowly. That assumption is now false.

02 — What Changed

AI code generation broke the arithmetic of billing by the hour

The problem is velocity. GitHub Copilot, Cursor, and the proliferating fleet of enterprise-deployed AI coding tools complete routine implementation work at ten times the pace of unaided engineers on a growing class of tasks. For certain work — CRUD endpoints, test generation, boilerplate, documentation, refactoring — what previously took a month now takes days.

Industry observers report that since AI coding tools were deployed at scale in Q1 2026, some IT outsourcing engagements saw man-hour requirements drop 60–80% within just three months, collapsing the economic floor under time-and-materials pricing. This is simultaneously a problem for contractors (margins evaporate) and for buyers (the entire basis for price comparison becomes incoherent).

What makes this especially disruptive is that it does not arrive as a gradual trend — it arrives as a conversation. The moment a technically literate procurement manager asks "why does this feature take two months?" and the contractor cannot answer without revealing that AI could do it in a week, the old contract structure is over. Reports from procurement teams at large Japanese enterprises indicate that exactly these conversations are now happening in vendor review meetings.

0 3 6 9 12 Man-months Old: 9 mo. AI: 1.8 mo. Mid feature (same output) Old: 12 mo. AI: 2.3 mo. Large feature (same output)
Estimated man-month requirements for equivalent deliverables: traditional vs. AI-assisted development
03 — Three Camps

Who must act now, who should watch, and who can wait

This disruption does not land equally. Where you sit in the ecosystem determines how urgently you need to respond — and what the right response actually is.

Must act now: contractors who still quote by man-month
The most exposed group is IT service providers still billing clients on a time-and-materials basis for work that AI now completes in a fraction of the time. Reports indicate that technically sophisticated buyers are already renegotiating mid-contract, and that vendors who cannot credibly justify their timelines are losing renewals to competitors who have adopted outcome-based pricing. The window to design a transition to value-based or fixed-deliverable contracts is open now; it may not be open next year. The transition is not optional — it is a survival question.

Should watch closely: enterprise buyers in procurement and IT
Organizations that outsource development on time-and-materials contracts are sitting on an arbitrage opportunity: they are paying for hours that AI has made far cheaper. But the naive response — demanding price cuts — carries its own risks. Contractors squeezed on margins will cut corners on testing, documentation, and security review. The right move is to renegotiate the unit of measurement, not just the price. Define deliverables, acceptance criteria, and quality metrics before reopening rate conversations.

Largely unaffected for now: internal teams on outcome-based metrics
In-house development organizations that are already evaluated on shipped features, cycle time, or business impact — rather than hours logged — are positioned to absorb AI tooling as a pure productivity gain. No contract structure needs to change; the AI just makes the team faster. That said, "unaffected" is not the same as "nothing to do." These teams face a different challenge: as throughput rises, stakeholders will raise expectations, and maintaining quality discipline at higher velocity requires deliberate process investment.

The man-month will not disappear overnight — too much contract law and procurement process is built around it. But its role will shift from a pricing mechanism to a rough planning heuristic, used internally and never surfaced to clients. The deeper question — how do you price software creation when the cost of creation approaches zero? — is one the industry has not yet answered.

AI Navigate — Daily Update · 2026.06.22