Most financial firms expect AI-led operations within years

ActiveOps and HFS Research have published a study on AI adoption in financial services operations. It found that 79% of leaders expect AI to execute or lead operational work within three years.

The research surveyed 101 C-suite and technology leaders at Global 2000 banking, financial services and insurance organisations in the UK and US. It points to a sharp shift in how executives expect work to be divided between staff and AI systems, but suggests many firms still lack the oversight needed to track the results.

At present, 64% of respondents describe their operations as human-led. Within three years, that figure is expected to fall to 21%, while joint human-AI working is forecast to rise from 24% to 40%.

Almost two in five, or 39%, expect their operations to be AI-led within that period, up from 12% today. Yet only 22% said their teams can see the work completed by AI agents, leaving most organisations without a clear view of what those systems are already doing.

The findings suggest firms may be moving faster on deployment than on management controls. That gap is evident in how they measure productivity, track capacity and assess whether time saved by automation has been put to use elsewhere.

More than half of respondents, 52%, said their main goal is to redirect the capacity created by AI to higher-value work or other teams. In practice, only 27% said that is happening.

For 40%, AI has either made no measurable difference to capacity or they could not tell where that capacity had gone. The survey also found that 43% cannot measure productivity before and after AI deployment.

According to the report, the underlying problem is that many institutions do not have a reliable baseline for human work before AI is introduced. Some 63% said they cannot see the human effort required to complete work, and respondents estimated that 39% of operational work is not captured in their core systems.

That creates a broader issue for executives trying to assess financial returns from automation projects. Only 17% said they can see unit cost across people and AI, limiting their ability to compare delivery methods or identify where additional automation might reduce costs.

Visibility gap

The survey focused on leaders in banking, financial services and insurance in the UK and US, where operational processes tend to be large-scale and closely supervised. The results indicate that management information has not kept pace with the spread of AI into day-to-day operations.

Without a clear picture of tasks completed by people, AI, or a combination of the two, firms may struggle to judge whether promised efficiency gains are real. They may also find it harder to show that AI tools are improving customer service, freeing staff for complex cases or helping contain costs.

Mike Hobday, Senior Advisor at HFS Research, said many institutions were looking in the wrong place when they tried to judge returns from AI.

"Work expands to fill the time available, and AI is proving the rule. Our research with ActiveOps shows only a quarter of financial services firms are redirecting the capacity AI creates, while 40% see no measurable change or worse, can't tell where it went. The reason is that most can't see the human effort behind their work today. AI is delivering a return but firms are looking for it in the wrong place. You can't capture time you can't see," Hobday said.

Operational shift

The figures point to a major operational rebalancing over the next three years. If respondents' forecasts prove accurate, human-led models will become a minority approach, replaced by a mix of shared human-AI workflows and AI-led activity.

That transition raises practical questions for managers beyond the initial deployment of software tools. Firms need to know how work is assigned, where delays emerge, what level of human intervention remains necessary and whether staff time released by AI is being redeployed deliberately.

Richard Jeffery, Co-founder and Executive Chairman at ActiveOps, said the issue was not simply adopting AI but understanding what happened after time was saved.

"Regulators on both sides of the Atlantic are right to back responsible AI adoption. But adoption is only the start. The real test is what happens to the time AI saves: more room to resolve complex customer cases, or the capacity to meet growing demand without growing costs. The firms that can see where that time goes will turn it into better outcomes. Everyone else will watch it disappear," Jeffery said.

The study was based on responses from 101 senior executives and technology leaders at Global 2000 financial institutions in the UK and US.