AI

Two-year roadmaps ask for faith. 12-week cycles produce evidence.

Matt Leta
Matt LetaCEO, Future Works
September 26, 20266 min read
A finance leader and a delivery lead reviewing quarterly results on a laptop at a conference table beside tall office windows, daylight mixing with warm interior light.

Two-year roadmaps ask for faith. 12-week cycles produce evidence.

Buyers are done funding an eighteen-month AI transformation roadmap on faith; they want proof on a shorter clock.

The scene has repeated in boardrooms all year: a CIO presents an eighteen-month AI transformation roadmap, walks through a mobilization plan, and promises the first hard number sometime in year two. The board approves it, because that is how large technology programs get funded: on faith in the plan, not evidence of the result. Nobody in the room can name a figure that would prove the bet wrong before the budget is spent. That gap, between commitment and evidence, is the real failure mode in enterprise AI transformation, and it has little to do with whether the underlying models work. The alternative is already showing up in board decks: outcome-staked fees and evidence on a 12-week clock.

The 18-month AI transformation roadmap is an expensive mobilization habit

Most enterprise AI transformation still runs on a mobilization model built for ERP rollouts two decades ago. Three to six months stand up a program office, a year or more goes to data integration and stakeholder alignment, while the go-live date slips past when value was due. Seventy-four percent of enterprise AI projects still show no measurable value, according to BCG research, and eighty-eight percent of AI pilots never reach production, according to IDC research. Neither number is about the technology; both describe a program whose only checkpoint sits at the end.

Long timelines diffuse accountability without anyone intending it. A dozen teams touch the program before year one is out: data engineering, security review, change management, a steering committee that rotates as executives move on. When something breaks at month fourteen, it is hard to say whose assumption failed, since most were locked in long before anyone could test them against reality. The executive who signed the roadmap rarely still owns it by the time anyone can grade the result.

A program built as a sequence of 12-week cycles, not a multi-year bet

A different structure is showing up in board decks: a program run as a sequence of 12-week AI delivery cycles, not one multi-year commitment, with the baseline locked alongside the client's finance team in the first two weeks, not the first two quarters. Build and deployment run weeks two through nine; by week ten finance checks whether the value materialized, and by week 12 the result funds the next cycle or it does not. None of this requires better AI. It changes how often the client checks the math, the same discipline behind booking transformation quick wins in the first 90 days.

That cadence looks more like how a public company runs its own quarters than how it runs a technology program. A steering committee that has not met in two months has nowhere to hide inside a 12-week cycle: week two forces a baseline into daylight, week 12 forces a verified answer. "A two-year roadmap asks a board for faith. A 12-week cycle asks finance for a check," says Matt Leta, founder of Future Works. "When you get a verified answer every quarter, the board stops debating the plan and starts deciding what to fund next." A weak cycle shows up in 12 weeks, not 24 months, so the program can be redirected, or stopped, before a second large commitment stacks on the first.

The self-funding AI transformation roadmap replacing the multi-year budget ask

The clearest sign of the shift is the fee itself: outcome-staked, not billed by the hour regardless of outcome. A meaningful share of every fee, commonly a third, rides on a value target the client's own finance team validates. Miss it, and the firm eats a service credit; beat it, and the bonus runs the other way. Either way, the client stops carrying the entire risk of a bet it cannot verify until the invoice arrives.

Future Works, built around this cadence, saw the pattern in its first cycle with a Fortune 100 healthcare manufacturer: measurable working-capital and freight outcomes, verified by the client's own finance team, built the case for the next cycle before any multi-year contract existed. The delivery model behind that result is AI agents plus named experts, a small pod of senior people, each orchestrating a fleet of agents, accountable inside a single cycle rather than rotated out mid-program. That structure earns its own name: an AI-native transformation engine, one that behaves like an operating partner rather than a contractor, producing a finance-verified result every 12 weeks instead of a promise redeemable only at a multi-year contract's end.

What cycles cannot do, and what to ask for at the next renewal

The honest complication is that cycles force scope discipline, and discipline has a cost. A 12-week cycle cannot absorb an open-ended ambition like reimagining an entire supply chain in one pass; it has to break into a sequence of provable moves, each with its own baseline and verification. Some ambitions are too large for that treatment: a multi-year ERP migration, a multi-country regulatory overhaul, an acquisition spanning several business units. Those still need longer arcs, and pretending otherwise trades one kind of dishonesty for another.

The cadence argument does not ask a five-year vision to shrink into 12 weeks; it asks for the value delivered along the way to be checked that often. A five-year ambition can still be real; it just does not need a five-year invoice before anyone learns whether cycle one worked. That is the distinction boards are starting to draw, and it changes the next renewal conversation: less "what is the multi-year plan," more "what did the last cycle prove, and what is staked on the next one."

Boards have noticed the shift. Few still open the year asking for a five-year AI strategy deck; most now ask a blunter question: when does the next cycle end, and what number lands on that date. An eighteen-month roadmap fails for a specific, structural reason: it gives accountability eighteen months to diffuse before anyone can prove the bet was sound.

For a board that has stopped funding multi-year promises and started asking for dates, the 12-week version of the answer is at Future Works.

Matt Leta, Founder and CEO, Future Works.

Keep reading

Related Articles

Real S&OP optimization with AI starts with routing rules nobody has re-examined in years

AI

88% of AI pilots never reach production, and the blockers are fixable

AI

Human middleware is the biggest unpriced risk in agentic enterprise workflows

AI