Starling adds AI ‘skills’ to its assistant chatbot

Starling Bank has introduced AI-based “skills” to its Starling Assistant chatbot and plans to add more on a regular basis, deputy CIO Frédéric Laurent described in a recent interview. The bank defines skills as encoded sequences of steps that tell an AI model how to carry out specific tasks.

Starling began layering AI into customer services about 18 months ago. The bank launched Spending Insights, which uses natural language processing and large language models to analyse customer spending. As models gained the ability to handle images, Starling deployed Scam Intelligence to flag potential fraud from photos. The Starling Assistant added reasoning features, and the new skills formalise repeatable tasks into discrete tools.

The idea for skill-enabled chatbots came from internal use of coding agents. Engineers identified daily, repeatable activities and converted those workflows into skills so automation could be scaled across teams. Laurent said the same approach can be applied beyond software development to help customers and non-technical staff with routine problems.

He gave the example of a student needing a budget calculated from their income, a recent loan and expected expenses as a case the assistant’s tools could handle. Laurent has led the bank’s AI, machine learning and data work since joining in 2019 and described the decision process for adopting new AI capabilities as a team effort. He follows industry newsletters and podcasts, shares ideas with data scientists and engineers, and tasks teams to test usefulness before wider rollout.

Starling has not published a full roadmap for upcoming skills. Laurent noted the bank intends to add new skills based on customer feedback and to consider where an effective capability can be applied across engineers, non-tech staff and customers. “We’ve deployed AI for the benefit of our engineers; we’ve deployed AI for the benefit of our non-tech staff, and we’ve deployed AI for the benefit of our customers,” Laurent added.

The bank’s skills work builds on earlier customer-facing AI features and the internal automation patterns used by its engineers. Starling characterised the development as an extension of the technical capabilities that became available in recent model updates and the bank’s own efforts to operationalise those capabilities.