Finance firms seek trust as AI moves into decisions

KPMG's latest global survey found that AI adoption in finance has more than doubled in two years, with the biggest gains in decision-making quality, decision-making speed and forecasting accuracy.

The findings point to a shift in how financial institutions use the technology. AI tools have moved beyond back-office tasks and now play a larger role in analysis, forecasting and operational decisions.

That change is increasing scrutiny of how automated outputs are reviewed inside banks, asset managers and other financial groups. The survey also found that firms able to produce audit evidence efficiently report improvements three to six times greater than those that cannot.

"AI in finance is graduating from doing the admin to influencing decisions, which is a much bigger opportunity, but it also raises the stakes considerably.

"If AI is helping decide where capital goes, how risk is assessed or what happens next in a market, finance professionals need to be able to interrogate the reasoning, understand the evidence behind it and challenge the assumptions before acting.

"KPMG's findings suggest the firms that get the most value from AI will be those that can use it to make better decisions without turning the decision-making process into a black box. In finance, the value of AI will ultimately depend on whether people can trust, test and stand behind the decisions it helps them make," said Varqa Abyaneh, Chief Executive Officer and Founder, Opetek

The issue of oversight is also emerging in debates about AI agents and customer accounts. Large financial institutions have taken a cautious approach to granting third-party AI systems direct access to banking relationships and transaction activity.

Theo Wasserberg, Managing Director for UK and Ireland at Embat, said the sector already has a model for controlled delegation. He pointed to corporate banking mandates, where users can act on accounts within defined limits and with approval steps recorded.

"Caution from the industry's largest institutions shouldn't be read as resistance to AI agents. It is the necessary first chapter of their adoption. We saw the same pattern with embedded finance. Initially, access started narrow and tightly permissioned, and trust was earned step by step. Within a few years, it became infrastructure nobody thinks twice about.

"Corporate banking already offers a blueprint. Businesses have long allowed others to act on their accounts through mandates, with defined limits, dual approvals and a clear audit trail of who authorised what. AI agents need exactly that. They should not hold open-ended credentials. They should hold a digital mandate that sets what an agent can see, what it can do, and when a human must sign off.

"That also brings clarity to the question of responsibility, because accountability follows the mandate. The institutions that design these mandates proactively will be the ones customers trust to put their agents to work," said Wasserberg.

Questions about control and accountability sit alongside growing concern over the economics behind the broader AI build-out. Heavy spending on data centres and computing infrastructure has driven private-market enthusiasm, but some investors are looking more closely at debt, contracted revenues, and customer concentration.

That matters for finance groups as well as technology suppliers. Banks and investors have backed much of the infrastructure expansion through loans, structured finance, and capital markets activity, while also weighing possible AI sector listings.

Varun Datta, Founder and Chief Executive Officer of Truth Ventures, said tighter lending conditions could ripple through private valuations and public market expectations if lenders become less willing to fund expansion based on aggressive assumptions.

"I think the AI investment boom is increasingly facing a financing test. Demand for computing capacity can keep growing while some of the companies building it struggle to generate enough cash to service their debt. As an investor, I'd want to understand both sides before treating that demand as a reason to invest.

"What I'd look at closely is how secure the contracted revenues are, how concentrated the customer base is, and whether the cash comes in before debt repayments fall due. A large order book wouldn't give me much comfort if fulfilling it requires further borrowing or depends on a handful of customers continuing to spend aggressively.

"My concern is that if lenders become more cautious, the effects could reach beyond data centre operators. More expensive capital could slow expansion, put pressure on private valuations and make investors more demanding about upcoming AI listings. I'd have more confidence in businesses that can demonstrate sustainable margins and explain how they'll fund their commitments if growth falls short of expectations," said Datta.

Taken together, the comments show that finance executives and investors are asking a similar question from different angles: whether AI systems and the businesses behind them can be examined closely enough for institutions to rely on them.