The category few are managing

There is a category taking shape inside most large companies that few are managing. It is already substantial, growing at close to half a year, and in many organisations it has no owner, no strategy and no figure that anyone can state with confidence. The category is AI, and the research on its returns has grown noticeably more sober.

The spending is not in question: Gartner places worldwide AI spending on course for roughly $2.59tn this year, an increase of about 47%. What has changed through 2026 is the candour about what it returns. MIT found that 95% of enterprise AI pilots produced no measurable effect on profit or loss. RAND concluded that more than 80% of enterprise AI projects fail to deliver their intended value, roughly twice the failure rate of ordinary technology work. S&P Global reported that 42% of companies had abandoned most of their AI projects during 2025, and IBM put the proportion delivering their expected return at around a quarter. Four independent bodies, using different methods, arrive at a broadly similar conclusion.

The reasons individual initiatives disappoint are, by now, well rehearsed — unclear ownership, poor data, a solution built before the problem was properly understood. Less examined is the aggregate consequence: an organisation can be simultaneously disappointed by every individual AI project and unable to say what it is spending on AI in total. The two failures are related, for where nothing is owned centrally, neither the value nor the cost becomes anyone's responsibility to account for.

This is not an argument against AI. The gains where it works are real, and the pressure to adopt is genuine. The difficulty lies not in the technology but in how it is being bought. Were any other line of cost — freight, contingent labour, professional services — to be rising at close to half a year, it would prompt the ordinary apparatus of category management. AI spending instead arrives as uncontrolled spending has always arrived: a feature enabled, a usage bill growing, a trial becoming a production system without a decision ever being taken to make it so.

The symptoms will be familiar to anyone who has managed indirect categories. Costs surface as usage charges that were not forecast — Suplari found 78% of IT leaders reporting exactly this. They sit within tools that were never competed. And they increasingly sit outside sanctioned procurement altogether: unsanctioned AI use is estimated to affect around two-thirds of enterprises, a governance and security exposure most IT functions have yet to address. We saw the same pattern produce maverick spend across indirect categories two decades ago; what differs now is the speed, since a consumption meter compounds a good deal faster than a signed contract.

How MomentumX solves this

---

This is the discipline MomentumX brings, and it starts where the difficulty starts: with the total. Through its Value Engine™, the firm assembles a first, complete view of AI spend across every tool, team and usage line — the number most organisations cannot yet state. Until that figure exists, none of the discipline that follows is possible, and assembling it for the first time is itself real work.

With the total in view, the firm builds a Rate-Anchored Should-Cost™ for the category: a considered sense of what a given AI capability ought to cost for the value it returns, so that a usage bill can be judged rather than simply paid. The recoverable value is sized on the Four-Lever Framework™ — price, specification, demand and cadence — and anchored to the MomentumX Benchmark Basis™, so that each figure rests on a cited public reference rather than on assertion.

Around this sits a light layer of governance over how AI is bought and scaled: metering, sensible caps, and a clear path from trial to funded production. The intention is not a committee that slows teams down, but a small set of rules that keeps a fast-growing category legible — the same discipline that any other line of this size would attract as a matter of course.

The work is delivered through the Operating-Advisor Model™, with senior judgement held in one place and execution drawn from a specialist bench. Most engagements begin with a Spend-Leakage Audit™, or a Confidential Spend Review™ of a defined slice of AI spend, and where a client prefers, the savings that follow are tracked to the profit and loss on the Value Realisation Tracker™. The outcome is not less AI, but AI that a board can see, weigh and defend, with the spending directed toward the minority of uses that genuinely return something.