UK business leaders are using one AI to fact check another despite being aware of AI's propensity to make mistakes.
Research from GWI has revealed that AI is now an everyday tool for 81% of UK business leaders, but 56% have made a major decision based on an AI-generated insight that later turned out to be wrong or misleading.
As a result, just under four-in-ten reported needing internal rework, 36% said it had led to delayed decisions and 35% said they'd had customer complaints. For some, the repercussions were more severe.
More than half (26%) reported financial losses as a result, with 24% experiencing reputational damage and 20% pointing to weaker campaign performance.
Leaders are aware of the risks, the study found, with 84% saying they'd seen at least one negative consequence of inaccurate AI output within their business. The same number rated the risk of decisions based on inaccurate AI insight as significant.
“AI is only as reliable as the data behind it, and most leadership teams already know that,” said Tom Smith, founder and CEO of GWI.
“The businesses feeling the impact aren’t behind on AI adoption. They’ve moved faster than most and while they embrace artificial intelligence alongside their own business intelligence, the shortfall tends to be a lack of human intelligence entwined into both. What they’re missing is the discipline to verify real human truths, because AI sounds just as certain when it’s wrong as when it’s right.”
Sign up today and you will receive a free copy of our Future Focus 2026 report - the leading resource for IT decision-maker insight on priorities and investment areas in AI, security and more.
Taking AI at face value
According to the study, two-thirds of leaders said they’ve unknowingly acted on data or insights that were out of date. Worse still, they believed colleagues were doing the same.
Verifying the authenticity or accuracy of AI outputs has become a recurring challenge for many respondents. Only 29% say they verify AI outputs “most of the time” while 31% only check occasionally.
Shockingly, another 36% cross-check using a different AI tool. While this may appear to be sound advice to verify outputs, the study warned users risk further muddying data and insights.
With this in mind, GWI said that more robust governance is vital, as is sticking to best practices when using the technology. However, few organizations are making solid progress on either of these fronts.
When asked who carries the can when an AI-informed decision goes wrong, 34% said responsibility is shared, 23% cited the organization, 22% the AI provider, and 20% the employee using the output.
Only 27% said that ensuring their data is methodologically defensible and traceable is a top priority.
"A framework can set the rules, but it can’t take responsibility for a decision,” said Smith. “The organizations that win will know what their AI decisions are built on, know when to question them, and to ensure they’re grounded in fresh, representative evidence from real people at scale.”
AI-related accuracy concerns have become a recurring issue for workers across a range of professions over the last two years. A study from ISC2 earlier this, for example, found 63% of cybersecurity professionals often find themselves reviewing and validating AI outputs due to accuracy concerns.
Nine-in-ten, meanwhile, said that AI-recommended actions had led to incorrect outcomes.