AI in 2027: bigger budgets, tougher questions

Most organisations are setting their 2027 AI budgets now, and the forecasts point one way. Gartner expects worldwide IT spending to pass $7 trillion for the first time in 2027, reaching $7.35 trillion, up 15% on 2026, with AI infrastructure and AI-enabled cloud platforms among the main drivers (Gartner, October 2026).

The evidence of return has not kept pace. In McKinsey’s 2026 State of AI survey, 44% of respondents said AI was scaling across their enterprise, up from 38% a year earlier, yet the share attributing any EBIT impact to AI stayed at about 37%. The high performers who credit AI with at least 5% of EBIT were still about 6% of the sample (McKinsey, The state of AI in 2026). Next year’s budgets will be judged on that gap.

Agents make each task more expensive

The price of running a model keeps falling, but agents use far more of it. Gartner expects the inference cost of each agentic workflow to rise more than fivefold through 2028 (Gartner, August 2026), and more than 40% of agentic AI projects to be cancelled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls (Gartner, June 2025). In the FinOps Foundation’s 2026 survey, 98% of practitioners said they now manage AI spend, against 31% two years earlier. Our BIAAS analysis looks at what this means for a data team’s 2027 budget.

The rules now have dates

In Europe the deadlines are set. Under the Digital Omnibus on AI, approved in June 2026, obligations for stand-alone high-risk AI systems, in areas such as hiring and credit scoring, apply from 2 December 2027. Those for high-risk AI built into regulated products follow on 2 August 2028. The AI Act’s transparency rules, including telling people when they are talking to a chatbot, have applied since 2 August 2026 (Jones Walker).

Manufacturers have an earlier date. The EU Machinery Regulation applies to machinery placed on the market from 20 January 2027, with no transition period (BAuA), and machinery whose safety functions rely on machine learning cannot be certified through internal checks alone. Our manufacturing report sets out what else plant leaders should plan for.

Fewer job cuts than planned

McKinsey found 39% of respondents expecting AI to reduce their organisation’s headcount in the coming year, up from 32%. Yet the share who expected cuts in last year’s survey was about double the share now reporting that cuts happened. In customer service, Gartner expects half of the organisations that planned to cut their service workforce significantly because of AI to drop those plans by 2027 (Gartner, June 2025). That is the case for giving AI to agents first.

What to take into 2027 planning

Tie each AI line in the budget to a process or a profit-and-loss line. Personal productivity gains are real, but on their own they rarely show up in EBIT.

Budget by cost per completed task, and assume that cost rises when a feature moves from an assistant to an agent.

Put the regulatory dates in the plan now: 20 January 2027 for machinery placed on the EU market, and 2 December 2027 for high-risk systems such as those used in hiring and credit scoring.

Expect a dip. Research on US manufacturers found that productivity fell after AI adoption before adopters pulled ahead, with older plants hit hardest (our manufacturing report).

These are the questions we will put to speakers at BIAAS 2027 in Amsterdam and CAIS 2027 in Singapore.