A few years is a short runway — but that’s just how long we may have before AI agents are doing our banking for us.According to the Cambridge Centre for Alternative Finance’s 2026 Global AI in Financial Services Report, 81% of financial institutions expect agentic AI — systems that plan, decide and execute without human supervision — to be banking’s new normal by 2030. The report surfaces an important gap, though. Despite the industry’s overwhelming expectation for continued AI growth, just 14% of institutions view AI as transformational to their strategy and competitive position today.Put differently, it’s fun to look forward to what’s next, but most banks are still waiting for the AI they’ve already deployed to start working. The ambition and the reality are running on different tracks. Now, before the industry races further ahead, this gap warrants inspection.A cautious postureFinancial services has long recognized technology adoption as a decision of risk management — not just competitiveness. And, of course, it makes sense. The stakes are high in banking. The consequences of a security failure or bad output at a bank or credit union are far-reaching. And that’s why this caution is a strength. It’s the same instinct that built the foundation of trust banking depends on.The breakneck pace of AI, however, is now creating friction with this caution. Banks that move too slowly are ceding ground to faster-moving peers. Meanwhile, banks that move too quickly could make a costly mistake. AI’s speed changes nothing about the underlying stakes of a wrong move. Abandoning the governance discipline that defines responsible tech adoption in financial services would be a dangerous mistake.Two worrying findingsCurrent AI deployments suggest governance is already under strain. Between 55% and 76% of industry respondents report difficulty measuring AI’s business value — a large accountability gap. Meanwhile, 70% of surveyed industry firms rank hallucinations and unreliable outputs among their top concerns.With AI pushing rapidly toward greater autonomy, these two findings should give us pause. An industry that can’t fully measure AI’s value or trust its outputs is being asked to stop supervising it.Governance gaps grow with autonomyMost AI deployed today assists human decisions — no matter how great the output, a person stays in the loop to apply judgment and catch errors. As AI systems take on more autonomous roles and act across chains of decisions without human review, this check will slowly disappear.The Cambridge report identifies loss of human oversight as the third-highest AI risk overall, cited by 51% of respondents. As AI-driven activity grows in volume and speed, meaningful human review becomes harder to maintain. The capacity to explain AI decisions — to account for why a system reached a given conclusion — is equally underdeveloped: 79% of regulators rate this capability as critical or important, yet only 50% of industry firms have adopted it in practice. An autonomous system that can’t explain its decisions can’t be held accountable for them.Governance can’t wait for regulationRegulatory frameworks are evolving, but AI is advancing at an extraordinary pace. Policymakers face the difficult task of creating oversight for a technology that changes faster than traditional governance models were designed to accommodate. While regulation will remain an essential part of the solution, organizations deploying AI today need practical safeguards and accountability measures that operate in real time — not years down the road.Demand more from vendorsClosing gaps will likely fall to banks and credit unions themselves. This means building strong governance structures before AI deployment, not tacking them on after something goes wrong.It’s a lot for a bank or credit union to take on, but a core part of the task can be delegated to AI vendors by requiring verifiable standards. Proving business value and security (such as safety from hallucinations) must be a hard procurement standard.As AI becomes more agentic, banks and credit unions deserve clarity from vendors on how much autonomy a system has — and who holds accountability in the event of a mishap. This clarity should be codified in vendor contracts before any system goes live.Keeping up with AI is critical for financial institutions to stay competitive. But speed to innovate should never compromise governance. As AI systems grow more capable and autonomous, this principle only becomes more crucial. No#BankingAI #GovernanceJustin DiPietroCo-founder & Chief strategy officerGlia28 May, 2026