A quarter of planned AI spend is slipping to 2027.
August 15, 2026
Forrester's 2026 predictions expect enterprises to defer roughly 25% of planned AI spend into 2027. The stated cause is not disillusionment with the technology - it is that fewer than a third of decision-makers can tie AI value to their organisation's financial growth, so CEOs are routing AI investment approval through their CFOs on ROI grounds. If that is happening in your company, the question is no longer whether spend gets challenged. It is which line items survive the challenge.
Deferral is reallocation, not contraction
Read the deferral number alongside the growth numbers or you will draw the wrong conclusion. Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, up 47% year over year, with enterprise AI spend specifically at $407 billion, up 34.8% from $302 billion in 2025.
Those two facts are not in tension. Aggregate AI spend is growing fast and a quarter of planned spend is moving out a year, because the money is being redistributed rather than withdrawn. It concentrates on programmes with a demonstrated outcome and drains from programmes without one. In a growing budget, that redistribution is easy to miss until your programme is on the wrong side of it.
What actually gets deferred
Deferral decisions are made quickly and on thin information. In practice, four properties make a line item easy to move:
| Property | Why it makes a line easy to defer |
|---|---|
| No unit economic | Total spend with no denominator can only be argued as a big number. Big numbers move. |
| Unattributed cost | If nobody can say which feature caused the spend, nobody can defend the feature. |
| An unexplained miss | A forecast that missed once without a stated cause reads as a forecast that will miss again. |
| Qualitative benefit only | "Improves developer productivity" competes badly against a line item with a number. |
Note what is absent from that list: size, and whether the programme works. An expensive programme with a credible unit economic is defensible. A cheap one described in adjectives is not.
The overrun problem feeding into this
Most AI programmes arrive at this conversation with a bad forecasting record, which is the third property above. A review of 127 enterprise agentic AI implementations found 73% over budget, some by more than 2.4x, averaging $2.3M in costs that had never been modelled. Uber gave Claude Code to roughly 5,000 engineers in December 2025 and spent its full annual AI budget by April at $500 to $2,000 per engineer per month.
None of those were failures of the technology. They were failures of the estimate. But from the seat approving next year's budget, an unexplained overrun and a failed programme look identical - which is why the explanation matters as much as the number. See 73% of agentic AI projects went over budget.
How to defend a line item
Three things, in this order. None takes a quarter.
- A measured unit economic with a counterfactual. "$0.31 per resolved ticket against $4.10 handled by a person" survives scrutiny. "$180k a month on support AI" does not. One workflow with a real number beats a portfolio-wide ROI estimate nobody believes. See proving AI ROI.
- A full-cost numerator. Inference plus evals, retries, integration, and human review. A unit cost built on the token line alone is provably understated, and being caught understating it costs you the whole argument, not just the delta.
- A variance history with causes. "We were 60% over in Q1; the cause was a 22x call multiplier per task against an assumed 3x; the multiplier is now measured and capped" is a stronger position than having never missed. It shows the estimate is now instrumented. See variance analysis.
Then put it in front of the board in the form they already read. See AI spend board reporting.
The CFO is now the approval path
The structural change worth naming: AI investment decisions are moving from the CTO's discretion to the CFO's approval. That is a different conversation with different evidence standards, and engineering teams generally have not prepared for it - they have optimization stories, which answer a question nobody is asking any more.
It also changes what the finance function needs to be able to do. Ranking AI programmes on cost per outcome requires attribution, full-cost modelling, and a unit metric per workflow as standing capabilities, not as a one-off exercise when a deferral list appears. See how to build an LLM CFO function.
What to do this quarter
- List the AI programmes and mark the ones with no unit economic. That list is your deferral candidate list, whether or not anyone has written it down yet.
- Pick the two or three most strategically important and instrument them properly - tagged costs, full-cost numerator, a counterfactual.
- Write the variance explanation for any programme that has already missed, before someone asks for it.
- Accept that some programmes should be deferred. A finance function that defends everything defends nothing, and volunteering the weak line buys credibility for the strong one.
Related
- 73% of agentic AI projects went over budget
- Proving AI ROI
- AI spend board reporting
- AI spend variance analysis
- State of FinOps 2026 for CFOs
FAQ
How much AI spend is being deferred to 2027?
Forrester's 2026 Technology and Security Predictions expect enterprises to defer roughly 25% of planned AI spend into 2027, as the gap between vendor promises and delivered value narrows appetite and financial scrutiny increases.
Why is AI spend being deferred?
Because fewer than a third of decision-makers can tie AI value to their organisation's financial growth. With no defensible link between spend and outcome, CEOs are leaning on CFOs to approve AI investment on ROI grounds, and undefended line items are the ones that slip.
Does deferral mean AI spend is falling?
No. Gartner still forecasts worldwide AI spending of $2.59 trillion in 2026, up 47%, with enterprise AI spend at $407 billion. Deferral is reallocation, not contraction - the money moves toward programmes with a demonstrated outcome and away from ones without.
What gets deferred first?
Programmes that cannot state a unit economic, whose costs are not attributed to a feature, whose forecast has already missed once without explanation, and whose benefit is described qualitatively. Any one of these makes a line item easy to move; two makes it likely.
How does a CFO decide what to protect?
By ranking programmes on cost per completed outcome against a stated counterfactual, not on total spend or enthusiasm. A programme with a credible unit economic and an honest variance history is defensible even when it is expensive; a cheap one with neither is not.
What should we do this quarter?
Pick the two or three AI programmes most likely to be challenged and give each one a measured unit economic, a full-cost numerator including evals and human review, and a documented forecast variance with a cause. That work takes weeks and is what separates a defended line from a deferred one.