TL;DR

Research from engineering and consulting firm Expleo finds UK business leaders happy to let AI handle reversible admin and markedly less happy once it decides anything expensive. Trust runs at 73% for summarising documents and meeting notes, drops to 54% for decisions inside preset guardrails, and reaches 43% where financial or budget approvals are concerned. Overall sentiment towards AI, meanwhile, improved.

The shape of the drop

The gradient is consistent rather than erratic, which is what makes it useful. Scheduling and calendar management sits at 72% comfort. Signing off low-risk administrative actions runs at 65%, and letting AI prioritise work and suggest next steps at 64%.

Then the floor gives way. Where consequences are lasting — money, employment, direction — leaders pull back sharply. Some 43% are actively uncomfortable with AI making hiring or HR calls; on strategic direction the discomfort runs higher still, at 46%.

Jeff Hoyle, who runs Expleo in the UK, put the logic plainly: a bad summary is caught and fixed in seconds, whereas a flawed decision about someone’s job or a company’s finances may not be reversible at all. His prescription is unglamorous and specific — decide in advance where the machine may act alone, fix approval points where it may not, and keep an audit trail showing the reasoning behind each one.

Confidence rising, comfort not

The caution coexists with warming sentiment. Expleo’s overall score recovered two points to 65 after the previous month’s low. Belief that employers can deploy AI successfully climbed five points to 67%. Excitement nudged up two points to 65%. Trust in employers to use the technology ethically reached 74%, having sat at 69%.

Two counter-currents run alongside. Worry about how AI is reshaping the workplace climbed four points to 39%, and cybersecurity remains the standing concern, cited by 59%. Optimism about the technology and unease about autonomy are evidently not the same variable.

Looking forward

The finding lands the same day as a survey of 178 financial advisers by GBST and the lang cat, which found the identical boundary drawn by a different profession: comfortable with AI on reconciliation and data collation, unwilling to hand it client money. It also sharpens yesterday’s Certinia research showing executives rating their AI rollouts more highly than their own staff do. Leaders are optimistic in aggregate and conservative in specifics — which is a defensible position, but only if the guardrails Hoyle describes actually exist rather than being assumed.