For the past two years, most enterprises funded AI largely on the strength of what it might deliver, and few boards pushed hard on the numbers. That patience is wearing thin as technology leaders finalize their 2027 budgets. Finance teams now want to know what AI has actually returned, and at many organizations the spending has grown considerably faster than the evidence supporting it.
A Dataiku and Harris Poll survey of 600 CIOs found that 98% say board pressure to demonstrate measurable AI ROI has increased since 2024. Forrester’s State of AI Survey 2026 found that more than half of CIOs expect to miss their AI ROI targets next year, and only 22% of AI decision-makers reported a positive impact on earnings over the past year.
A Widening Gap Between AI Ambition and IT Budgets
Gartner’s 2027 CIO and Technology Executive Survey projects that IT budgets will grow by an average of just 3.7% next year, which leaves many organizations with little real growth once inflation is factored in. AI spending is on a much steeper path. Gartner separately found that CFOs plan to increase AI budgets by an average of 15% to 16% in 2027, with 92% planning some expansion, and Forrester expects AI to account for roughly a fifth of all business technology spending next year.
With AI budgets growing roughly four times as fast as the IT budgets that contain them, many organizations are already struggling to keep spending in line. A Futurum Group survey of 1,636 technology decision-makers found that 47% of organizations are already over budget on AI, compared with 32% roughly on plan and 6% under. Another 10% have no formal AI budget at all, which makes overruns difficult to measure in the first place.
The way companies are handling those overruns says a good deal about their priorities. Among those over budget, 48% ask for more money and 43% absorb the cost and settle it later, while only 17.2% pause or scale back the AI initiative. When the gap is closed by reallocating funds within IT, the cuts fall first on external contractors and consultants (61%), IT support and help desk (60%) and software development (58%). AI, in other words, is being paid for in part by the teams and services that keep existing systems running.
Why ROI Is So Difficult to Prove
One complication is that many AI providers are moving from seat-based subscriptions to consumption-based models, which means costs rise with usage rather than headcount and are far harder to forecast. Forrester’s 2027 budget planning guides go so far as to describe token economics as a new line item in security budgets, warning that these costs could offset some of the efficiency gains organizations expect from AI. For use cases with thin margins, even a modest increase in per-token pricing can erase the business case.
Expectations about how quickly AI should pay for itself are also working against many projects. According to Forrester, 45% of CIOs want to see a return within one year before they consider an AI rollout a success, a short window for initiatives that often depend on reworked processes and retrained staff. Only about half of technology leaders (53%) expect their overall budgets to grow faster than inflation next year, so there is little slack to carry projects that take longer to pay off.
In many cases, though, the obstacle has less to do with cost than with the condition of the systems AI is meant to improve. Forrester’s planning research, based on a survey of more than 2,600 business and technology decision-makers, concludes that increasing AI spending without addressing data quality, governance and operating models tends to compound fragmented data, duplicated work and technical debt. Under those conditions, the returns projected in an AI business case are difficult to realize regardless of how much is spent.
How Technology Leaders Are Responding
Gartner and Forrester have arrived at similar advice for 2027. Gartner advises CIOs to require 90-day proof points for AI investments, shut down AI projects that consume resources without delivering value and steer teams toward curated, approved AI tools, supported by cost-allocation models that assign AI spending to the business units that benefit from it. Forrester recommends scaling back pilots that lack clear ownership, governance or a path to production, while concentrating investment on data quality, governance and the enterprise knowledge AI depends on. Forrester also expects more organizations to adopt open-weight models for high-volume workloads such as document processing and internal search, where lower-cost models are sufficient.
Enterprises are not pulling back from AI, and spending will continue to rise in 2027, but the criteria for approving that spending are becoming considerably stricter. Projects will increasingly need a defined owner, a measurable outcome and a realistic cost model before they are funded, and those that cannot meet that standard are likely to be the first casualties of next year’s budget cycle.
