CEOs stung by surprise costs begin scaling back AI initiatives
A new study of corporate leaders has found that 42% regard the costs of operating AI as ‘largely invisible’. As many encounter rising prices from the technology they seemingly expected to be cheap forever – even when AI companies had a monopoly – half have scaled back or paused further role out of AI initiatives.
Talk of innovative potential seldom translates to anything more than PR for UK businesses. One study by PA Consulting recently found that of 4,595 businesses, 47% ranked themselves between seven and 10 on an ‘innovation scale’, in spite of having failed to adopt any of 20 critical technologies researchers defined as “critical for the future”. Meanwhile, just 24% of businesses were able to say that they had sold at least one technology as part of their products or services – and even when they had, 97% of UK patents fail to generate commercial returns.
This tendency has been a core part of the persistent, triumphant announcements in every sector about becoming ‘AI-ready’. Roughly one-fifth of firms investing in the technology account for three-quarters of all the returns on investment from the technology – as per PwC estimates. Yet thanks to their lack of understanding how to measure such impact (Cynozure finds under two-fifths of AI leaders currently visualise the impacts of the technology on their firms in terms of currency), and their general tendency to over-estimate the effectiveness of a firm’s innovation, many bosses have taken a long time to come round to the idea the technology might not be yielding the kinds of benefits they thought.
However, a new report from KPMG suggests that a sudden rise in costs may have poured cold water on the hopes and expectations of the C-suite regarding AI. A new survey of 2,145 senior leaders across 20 countries found that 29% struggled to understand their operating costs as they scale their enterprise AI deployments.
This left them vulnerable to a short, sharp shock in 2026, as earlier in the year, AI giants such as Anthropic and OpenAI shifted some services away from flat-rate subscriptions toward usage-based billing. With some firms blowing through their annual budget for research in a matter of weeks, thanks to this move, many firms have quietly begun re-evaluating the situation.
KPMG found that a third of senior corporate leaders identified limited understanding of AI costs and economics as a challenge to deploying AI agents. And in response, the researchers also found nearly half of organisations had rephased AI deployments when they found costs actually outweighed the expected value. Lower-cost, high-fidelity models are the fastest-growing influence on AI strategy, up 7% from the first quarter of 2026.
"As usage-based pricing models become more common, many organisations are still building the capabilities required to forecast, monitor, and manage AI spending effectively," KPMG said. “Organisations appear increasingly focused on concentrating investment where expected returns are strongest.”
The finding underscores what many workers forced to use AI tools on the job have come to suspect. Many corporate leaders treat AI as a plug-and-play solution for lowering overheard – and disciplining labour by weakening its position in wage negotiations – without understanding the technology. And as their plans meet a cold, hard reality, the AI industry faces a reckoning – with Forbes reporting that more than half of bosses who attempt to replace workers with bots regretted it, and a third already reversing the process. As the realisation of what AI’s ROI actually is continues to hit home, that may still rise.
