Many large companies are spending more on AI than planned, while struggling to tell whether it is paying off. A Capgemini Research Institute survey cited by Diginomica found that 60% of respondents exceeded their AI budgets this year, and 80% of organisations were not calculating AI’s business value against its full cost.
Watch Desk analysis
What happened
The survey covered 500 senior executives at enterprises with more than $1 billion in revenue, across 13 industries and ten countries. Diginomica says the research was conducted in September 2026.
Among respondents whose organisations overspent, the average overrun was 24%; 21% overshot by more than 30%. Two-thirds of those whose budgets were exceeded reported negative financial consequences, including lower free cash flow or reduced discretionary spending.
The report also found that only 12% said they could track most or all of AI’s costs and value across their organisation. Capgemini recommends measuring the full cost and value of AI outcomes, strengthening governance and designing workflows around cost, quality and reliability.
Why it matters
The figures make a useful distinction between buying AI and proving it works economically. Model bills are only part of the cost: the account also includes infrastructure, oversight, governance and rework. If those are missing from the ledger, a cheaper task or faster output may not amount to a better business result.
These are survey findings reported by Diginomica, not an audited measure of every enterprise’s AI spending. Still, the gap between widespread overspending and the small share tracking both cost and value is worth attention, particularly as companies hand more work to agents that can consume substantial computing resources.
Our read
A budget variance is not automatically a failure, and an AI project need not save money to be worthwhile. But spending without a credible way to assess the result is not a strategy; it is an expensive exercise in optimism. Before scaling an AI workflow, organisations should compare its full costs and outcomes with the process it replaces.
What to watch
- Whether companies start reporting the full costs of AI projects, including oversight and rework.
- Whether the reported overspending leads to tighter controls or a shift towards cheaper models.
- Whether organisations publish evidence of business outcomes, not just adoption figures.
Discussion spark: Should companies have to show that an AI system beats the full cost of the process it replaces before expanding its use?
Sources and evidence
- Want to hear more about Agentic AI? (13 November 2026, 08:35 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.