TL;DR

Around seven in ten executives call their AI deployments a success. Among the practitioners actually using those tools, it is closer to five in ten. Only 20% of organisations say results have beaten what they expected, and employee distrust now ranks as a top adoption barrier for a third of firms.

Where the gap is widest

The headline split is 69% of executive leaders against 53% of decision-makers and practitioners in professional services and customer success. Within professional services the divide widens by discipline: 55% judged AI successful at consulting firms, falling to 43% across audit, accounting and tax.

That ordering is not surprising and is worth stating plainly. A hallucination in advisory work produces a bad slide. The same error in a regulated audit produces a filing problem. The more consequential the output, the less impressed the people producing it are.

Certinia’s chief executive, DJ Paoni, argued that the gap between firms getting value and firms getting noise “has almost nothing to do with how much AI they bought” — adoption is universal, results are not.

Read the sponsor before the finding

A caveat belongs here rather than in a footnote. The research was commissioned by Certinia, whose conclusion is that “connected systems” separate the winners from the strugglers — and connected business systems are the category Certinia sells. The perception-gap number stands on its own; the prescribed remedy comes from an interested party.

Trust is where the report is more persuasive. It puts employee distrust of AI output as a leading barrier for 33% of firms, and finds skills gaps and distrust weighing more heavily on stalled programmes than budget or leadership backing ever did.

Looking forward

The pattern is consistent across evidence appearing this week. The CBI reports 49% of AI leaders meeting return expectations against 15% of laggards, and Lloyds finds 31% of UK firms short of the skills they need. IBM’s polling, cited in the same report, found 60% of chief executives still experimenting a year after two-thirds expected to be past piloting. For a UK SME, the practical lesson is cheap: ask the people using the tool before reporting the result upward.