More than one recent report suggests leaders are investing heavily in AI but not seeing the benefits. Grace McNicholas says companies should stop issuing AI mandates and try speaking to employees to find out where it might deliver benefit.
Banks are spending heavily on AI: but they’re struggling to say what they get back.
TXP's AI Value Gap Report surveyed 112 business leaders involved in AI strategy at UK mid-market banks. 52% of leaders say their AI initiatives have failed to deliver value or been underwhelming, despite more than £1m on average spent on pilots – of which just 49% scale into full production. These findings suggest serious amounts of money are being lost on initiatives that fail to deliver. Why?
Mandates produce use, not usefulness.
“Users are silently correcting AI, making it appear more effective than it really is.”
Separate research from Camunda covering financial services organisations and their employees reveals 45% of employees say they have used AI to comply with an organisational mandate – even when it didn’t help them perform a task.
Worse, 68% of employees report that they were not fully consulted on how AI would support their day-to-day work, while 63% say they could use it far more effectively if it were implemented differently. Perhaps most damagingly, Camunda found 49% of employees have manually overridden AI outputs because the underlying process wasn't correctly set up. Some AI pilots may appear successful - but this hides users correcting and overriding incorrect workflows in the background, making AI look better than it really is.
Success remains elusive, despite executive confidence.
Most business leaders appear oblivious to these issues, possibly because they are drowning in the current wave of AI hype and wish to appear current. TXP say 89% of leaders are confident that AI is making their teams more productive, and 96% are confident their organisation has the data needed to support AI at scale. Fully 98% of executives surveyed say they have a clear AI strategy, a confident finding that sits at odds with AI’s present failure to deliver.
A pilot that depends on manual correction doesn’t make a business more efficient.
Blind faith in the benefits of AI won’t deliver results by itself. According to Camunda, almost half of employees report adopting AI because it’s mandated, rather than beneficial, and fewer than half the AI pilots launched end up being used at scale.
Organisations wanting to find the gap between confidence and success should start by asking employees where AI is working, and where it’s not. Right now, productivity gains remain subjective, but failed pilots are all too real – and expensive. If AI is going to live up to the hype, companies need to get better at separating real benefit from the outcomes they wish AI could achieve.