The Next Chapter of AI in Banking Will Be Defined by Accountability
As banks move AI from pilots into everyday operations, the harder question is no longer what a model can do, but whether accountability holds as decisions move across people and systems.
A few days ago, I came across an article by Ben Saunders, Co-Founder of WeBuild-AI, titled “Banking Doesn’t Need More AI Pilots, It Needs the Confidence to Scale Them.” It was an interesting read, particularly because it touched on something many of us in the industry have probably observed over the last few years.
Banks are not lacking AI initiatives. If anything, there are plenty of pilots. Transaction monitoring, onboarding, document processing, summarisation, customer service, fraud detection, the list goes on.
Yet, very few institutions would comfortably say that AI has become a natural part of their operating model.
The question seems to have shifted.
It is no longer, “Can AI do this?”
It is increasingly becoming, “Are we comfortable allowing AI to do this in production?”
That is a very different discussion.
In banking, confidence is rarely built on capability alone. It is built on understanding who is responsible when something goes wrong.
Who approved the action?
Was the right person involved?
Was the decision reviewed?
Can we explain what happened six months later?
Can we reconstruct the sequence of events across multiple systems?
These are not AI questions. They are governance questions.
Interestingly, regulators around the world appear to be moving in a similar direction. Whether it is operational resilience, internal controls, human oversight, or independent review, the common theme is accountability. Vietnam’s SBV Circular 83/2025 is one recent example that reinforces the importance of governance and operational discipline within financial institutions.
AI has simply made these questions more visible.
An investigator assisted by AI, a recommendation generated by a model, or an action initiated in one system and completed in another all introduce an additional layer of complexity. The technology itself is often the easy part. Ensuring accountability survives the journey is considerably harder.
My personal view is that the industry will eventually stop measuring AI maturity by the number of pilots conducted.
Instead, it will ask a much simpler question:
Can this institution scale AI with confidence?
The answer, in many cases, will have less to do with the quality of the model and more to do with the quality of the governance surrounding it.
Perhaps the next chapter of AI in banking will not be defined by intelligence.
Perhaps it will be defined by accountability.
Inspired by “Banking Doesn’t Need More AI Pilots, It Needs the Confidence to Scale Them” by Ben Saunders, Co-Founder of WeBuild-AI (17 July 2026).