Before You Sign the 2027 AI Budget, Somebody Has to Own the Number

Worldwide AI spending is forecast to hit $3.64 trillion in 2027, per Gartner — up from $2.67 trillion in 2026, itself a 49.5% jump from the year before. That number will get quoted in every budget deck this fall. It is the least useful number in the deck. The number that actually decides whether 2027's AI spend returns anything is smaller, less dramatic, and almost never on the slide: who, by name, is accountable for it.

A quarter of $1B+ companies are signing checks with no process at all

Deloitte surveyed 1,434 finance leaders — CFOs and the executives one level below them — at companies with at least $1 billion in annual revenue, across 26 countries. Published September 15, 2026 and reported by CFO Dive, the findings describe an accountability gap sitting directly underneath the spending curve. Sixty percent of respondents expect AI costs and technical complexity to rise substantially through 2027. Asked how they most often approve large AI and technology investments, 66% described an internally driven process with measurement at the center — 27% using a stage-gate model that ties additional funding to milestones met, 25% requiring quantified ROI and a business case before approval.

That leaves roughly a quarter of respondents — at companies with a billion dollars or more in revenue, with finance functions built for exactly this kind of governance — who said company mandates from the C-suite or the board drive most of their AI and technology investment decisions, with no process used to gauge the value of the projects at all.

Read that figure against the size of company it describes. These are organizations with dedicated FP&A functions, controllers, and in most cases a CFO with real authority over capital allocation — and a quarter of them are still approving AI spend on mandate alone. A mid-market company without that infrastructure, where the AI budget conversation is often a founder or CEO weighing a vendor's ROI story against gut feel, is not starting from a better position than Deloitte's respondents. It's starting from a worse one, without the survey to prove it.

Four questions before the number gets signed

The point isn't that AI spend should be smaller. Gartner's growth curve is going to happen regardless of what any one company decides. The point is that a budget without an owner, a gate, and a named use case is not a strategy — it's a subscription with better marketing. Four questions settle whether the number in front of you is ready to sign.

1. Who owns this, by name — not a committee?

Deloitte found 54% of finance leaders now lead enterprise AI and technology capital-allocation decisions, and 48% oversee AI and technology spending and cost controls directly — a sharp expansion of the CFO mandate into territory that used to sit with IT or with whichever function asked loudest. That's the right instinct at large-enterprise scale, where a CFO's office has the staff to actually run the oversight. It doesn't automatically transfer to a mid-market company where the "CFO" is a controller stretched across three other jobs, or where there is no CFO at all and the AI budget lands on a founder's desk between a hiring decision and a lease renewal.

Whether the right owner is a CFO, an AI Operating Partner, or a fractional Chief AI Officer depends on whether the accountability is tied to a specific deal thesis or to the operating model of the company itself — a distinction we've mapped in full elsewhere. What doesn't depend on any of that: the name has to exist, attached to real decision rights, before the budget line does. "The leadership team is aligned on AI" is not an owner. It's the exact structure Deloitte's quarter-with-no-process respondents described right before the follow-up question asked how they measure whether it worked.

2. What's the gate before the next dollar?

Twenty-seven percent of Deloitte's respondents fund AI investment through a stage-gate process — a pilot first, additional funding released only when goals are met at each stage. That is the cheapest insurance policy in the entire budget. A blank-check line item with no stage attached is functionally identical to the quarter of companies approving spend on mandate with no measurement at all; the only difference is the size of the number before anyone notices.

3. Is this budget an addition or a reallocation?

Gartner's growth figures describe global spend, not any one company's headcount plan. Inside individual budgets, finance commentary heading into 2027 points toward reallocation rather than addition — AI tool costs absorbed by trimming elsewhere, not funded as new spend on top of an unchanged base, as actual AI tool bills have in many cases run several multiples higher than what was originally projected against IT budgets growing only in the low single digits. Treat the industry-wide framing directionally: the underlying vendor-level cost data is still noisy and largely self-reported. But the question is worth asking of your own budget regardless of what the industry average turns out to be — if the AI line is new money, name what it's expected to displace; if it's reallocated, name specifically what got cut to fund it, and who signed off on that trade.

This is also where the headcount question gets honest. A tool budget that quietly assumes it will let a team run leaner next year is making a headcount decision without calling it one. If the AI line item's business case depends on not backfilling a role, that should be written down next to the number, not discovered in next year's org chart.

4. What happens if the diagnostic says no?

The honest failure mode isn't overspending on AI that works. It's spending on AI that was never going to work for this company's specific operating constraints, and finding that out after the budget cleared rather than before. That's the actual function of a pre-commitment diagnostic — ours runs $9,500 and exists specifically to answer which use cases justify budget and what kind of ownership they require, detailed in what a fractional CAIO costs, and what you're actually buying, before a company signs a full-time hire, a fractional retainer, or a tool contract it can't yet justify.

We've written before about the gap between AI adoption and AI governance — 82% of mid-market companies running AI in production with only 26% governed enterprise-wide. This is the same gap, priced in dollars instead of risk. Gartner's trillion-dollar figures are a weather report; they describe the industry, not your company's exposure. Deloitte's quarter-with-no-process figure is the number that should change what you do this budget cycle, because it's the one that's still entirely within your control.