Starling Bank has added new AI-based “skills” to its Starling Assistant chatbot, inspired by advances in the tech sector and the way its internal software development team has used AI to support coding tasks.
The bank plans to add these so-called “smart tools” on a regular basis, based on customer feedback. Frédéric Laurent, deputy CIO at Starling Bank sees this latest development at the bank as a culmination of the work it has done on customer-facing AI. He says: “It’s the fruit of all of the customer-facing AI developments we’ve done over the last 18 months and is essentially the journey following what the AI industry at large has been able to develop from a tech point of view.”
As AI technology improves and the AI industry identifies how best to use it, Laurent says Starling Bank has tried to make sure those technological advancements benefit both the bank and also its customers.
“We started with Spending Insights, which uses natural language processing based on LLMs (large language models), then the AI companies added multi-modal capabilities, so we created Scam Intelligence, which can take a picture and decide whether there are some red flags relating to fraud. Then they added reasoning, so we created the Starling Assistant, and then the industry seems to have coalesced around the concept of skills, which is a way of telling an AI model how to perform a task through a succession of steps that you’ve essentially encoded.”
As Laurent explains, the smart tools project was inspired by the way in-house software development has used AI. “In order for our engineers to get the most out of coding agents, we realised it would be useful to codify our understanding of our code bases into skills which are certain activities that an engineer would do probably once a day,” he says.
By codifying daily tasks means this repetitive work can be scaled out and it is something that Laurent and the team identified as having application areas outside of core software development. He says: “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. Every time there’s something that seems to work in one area, in my role I try with my teams to think about where else it can be applied.”
Looking specifically at the bank’s latest development with smart tools, he says: “We are essentially applying AI to the day-to-day problems that customers might face, like if they are starting university tomorrow and need a budget based on the income they have, the loan they have just taken out, and their expenditures.”
Laurent has worked at Starling Bank since 2019 and one of his primary responsibilities as a deputy CIO is AI, machine learning and data. He says that keeping up with the pace of AI development is “a team effort”.
“At a personal level, I follow a number of newsletter, and podcasts. It’s overwhelming, continuously consuming sources of information in order to stay on top,” he says.
When asked how he decides what AI innovations are relevant and also what to ignore, he says: “There are things that catch my eye, which I share with my teams. My lead data scientist might find this incredibly useful, or conversely, tell me, ‘Fred, you’re an idiot’, or it is not useful.”
Other AI innovations may not have an immediate use case at Starling Bank, but they may become more relevant at some point in the future.
Laurent believes all organisations need to decide where to focus in terms of the innovation coming being developed by AI companies. He says: “I’m sure there may have been things – opportunities – we let slip, but conversely, I think we’re quite happy and proud of the decisions we’ve made over the years, including getting into machine learning and the deep learning space in 2019.”
S 004