Banking software supplier SBS says AI agents capable of making live decisions could reach its UK building-society clients by next summer. The technology is moving faster than the controls and institutional willingness needed to let agents act in a regulated business.
Watch Desk analysis
What happened
SBS chief executive Eric Bierry told Scottish Financial News that the company needs another six to eight months to build technical controls for agentic AI. Scottish Building Society, its first client for the capability, currently uses generative AI rather than live agents.
SBS is building a control plane to define what agents can do and a sandbox for testing them, alongside an AI gateway developed with Axway. Bierry said every decision in a financial institution needs to be explainable and repeatable. He also said two other building societies were close to signing up for the capability, with as many as 30 institutions potentially able to use it over time.
Bierry described a separate trial at a French bank, where agents are repairing about 90 per cent of failed payments previously fixed by staff. He said people are checking the agents’ work for now, with the longer-term aim of retaining an audit trail even if human checks are phased out.
Why it matters
This is a concrete view of where AI agents might enter banking first: not by replacing a whole bank’s core system, but by handling bounded work such as repairing failed payments. SBS’s pitch is that the ledger remains deterministic while agents take on more tasks around it.
The gap between a working tool and permission to use it is the important part. Banks must be able to explain what an agent did, reproduce its decision and show that guardrails catch behaviour outside the rules. Bierry says those controls are harder than building the agent itself. That is a rather less glamorous challenge than the demo, but it is the one that decides whether the demo gets near a real payment queue.
Our read
There is meaningful progress here, but distinguish the stages: the first building-society client is using generative AI, not live agents, and the payment-repair example is a trial described by SBS’s chief executive. The useful signal is that providers are working on governance and auditability alongside automation, rather than treating them as finishing touches.
What to watch
- Whether SBS announces the two prospective building-society clients.
- Whether the French bank’s payment-repair trial moves beyond human-supervised testing.
- What evidence SBS provides that its control plane makes agent decisions explainable and repeatable.
Discussion spark: For bank tasks such as repairing failed payments, should human review remain compulsory, or can a reliable audit trail justify letting agents act on their own?
Sources and evidence
- Source update (12 November 2026, 10:15 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.