TL;DR

Starling has released a set of AI-built “smart tools” that customers can shape themselves. The opening set covers fraud controls, tax analysis, savings worked out automatically and a budgeting planner aimed at students. Ideas submitted through the bank’s agentic assistant can be published for other customers to use, and the bank says new capabilities will ship every week.

Shipping, not piloting

Most UK banking AI announcements this year have described trials, internal pilots or partnerships. This one describes a product in customers’ hands with a weekly release cadence — a meaningful distinction when the sector’s default posture is still a controlled experiment with a fixed end date.

The crowdsourcing element is the genuinely unusual part. Customers propose tools through the assistant, and some of those become publicly available to others. Group CIO Harriet Rees framed the reasoning around a split in the customer base: plenty use the assistant daily and have worked out how to bend it to their own needs, while others remain unsure what it can do for them. Publishing what the first group builds is a way of showing the second.

The context Starling would rather you skipped

In July the bank said increased AI investment would mean restructuring its banking and technology operations, with roughly 130 roles going to remove duplication and accelerate product launches. That was seven weeks ago.

The connection is not hidden — a weekly shipping cadence is exactly what “accelerate product launches” was meant to buy. But it does put a number on the trade. This is what a bank looks like when it converts headcount into release velocity, and Starling has now demonstrated both halves within two months.

Looking forward

Starling is not new to this. It rolled out an agentic money manager in March and romance scam detection in June, so the assistant has an established install base rather than a launch-day one.

For UK financial services the question this raises is about governance rather than novelty. Customer-suggested tools published to other customers is a distribution model with no obvious analogue in retail banking, and the consumer duty questions around a tax analysis feature designed by a member of the public and used by strangers have not obviously been answered yet.