Almost every financial services firm is now using AI. No surprise there. What might raise an eyebrow is that very few can show the value of the AI they’re using. Why is ROI proving so hard to pin down, and what are the firms seeing value doing differently from the rest?
Where are we now?
When we polled senior financial services leaders at a recent industry event, not one person described their organisation as operating AI at full scale. More than half (53%) placed themselves at 'scaling attempts' where AI is established in one or two areas, but not spreading for reasons that aren’t clear.
The rest were still experimenting, or committed but stuck. Spending seriously while the board asks about returns.
When asked for their single biggest AI priority, 40% chose proving value and ROI, twice the share of any other option.
This is not an isolated finding. McKinsey's latest State of AI research found that most organisations say AI has made their people more productive, yet only around a third can point to any impact on profit.
Why are firms stuck in the middle?
Many financial services firms are in the middle of the maturity curve where the easy wins run out. First-wave use cases (a drafting copilot, contact-centre summarisation or coding assistant for example) sit inside single teams, roll out with little disruption and make individuals faster.
The next wave cuts across functions, depends on shared data, needs risk and compliance from the start and asks people to do things differently. What worked for the pilots doesn't carry over.
Firms admit they only partly understand the AI they already run. Scaling that is a hard sell to a risk committee, and rightly so.
Why is AI ROI so hard to prove in financial services?
Most organisations measure the wrong thing. Licences, active users and prompts per week are easy to report, but they describe activity, not whether the business is performing better.
Productivity is the next trap. Saving ten minutes a day for thousands of employees sounds significant. But that time is spread thinly and rarely shows up in a P&L unless someone decides what the freed-up capacity is for.
In financial services, much of the value in risk, compliance and operations is avoided cost or reduced exposure, which is real, but harder to count. And with AI spend split across technology, operations and business budgets, nobody owns the full picture.
There are signs of progress. Evident's research suggests the fastest returns come from redesigning repeatable, high-volume workflows and validating the outcomes directly with finance and business leaders.
What are successful firms doing differently?
McKinsey's research shows that the key to successful AI is mostly organisational. High performers are almost three times as likely to have redesigned their workflows around AI. They set goals for growth and innovation as well as efficiency. Senior leaders are visibly committed, they have defined ways of measuring impact, and they invest with more conviction. They treat AI as a change to how the business operates.
We see the same in our own financial services work. The shift that matters most is usually from a handful of enthusiastic early adopters to organisation-wide, governed usage, with shared platforms, clear ownership and measures agreed in advance.
What does agentic AI change for financial services?
Back in 2024, the Bank of England and FCA found very few fully autonomous AI use cases.
That's changing quickly. The FCA's AI Live Testing programme now includes major banks testing agentic AI and agentic payments. Its Mills Review uncovered real consumer appetite for AI that acts on people's behalf, alongside concerns about trust and control.
Firms need to know who is answerable for an agent's decision, what it can do without a person in the loop, and how to evidence the reasoning afterwards.
Governance designed for predictable systems and slow change wasn't built for that, and tightening the existing controls won't close the gap.
There's a commercial angle too. If customers delegate tasks to their own AI agents, firms need to be easy for agents to work with.
What does all this mean for financial services leaders?
What matters now is the ability to say, with evidence, what AI investment is worth, and a willingness to change how the business works to capture it.
That starts with an honest view of where value needs to show up, measures agreed before anything is built, and the patience to prove it in one place before scaling it everywhere. It also means treating AI as an ongoing commitment. People, governance and operating models must move alongside the technology.
If your organisation is somewhere in the middle and wants a clearer route to value, we'd welcome the chance to talk it through.
FAQs
Meet the author
Simon Hull is Head of Financial Services at CreateFuture. With over 20 years in banking and wealth, including UBS, Barclays, BlackRock and Deutsche Bank, he helps firms turn AI strategy into practical, accountable change.
Industry Insights
Explore the latest thinking from our industry and tech experts.
What is AI native?
Why do AI experiments fail and how do you scale AI-native delivery?
CreateFuture expands its Anthropic AI capabilities as Version 1 Group achieves Preferred Services Partner status