Saul Van Beurden has a folder on his phone labeled “productivity.” He’s constantly consulting the many AI apps that live there, whether sitting in his 60th-floor office in Hudson Yards or in front of his broken laundry machine.
Loading audio narration…
Van Beurden, who was named Wells Fargo’s head of AI in November, is betting the technology will catapult the major retail bank forward after the Federal Reserve lifted its $1.95 trillion asset cap last summer. Some growth is already starting: the bank opened almost three million new credit card accounts and refurbished 700 branches, with plans to upgrade more. And Wells Fargo has big ambitions, including becoming one of the industry’s “top five” M&A advisors, partly by hiring top talent.
To Van Beurden, AI is crucial to attaining those goals — it’s going to touch every line of business and every workflow, from call centers to the investment bank. His job is to keep a bird’s-eye view of it all, making sure to prioritize the most urgent, innovative initiatives.
“It creates a complete new window of opportunity,” Van Beurden, a Netherlands native, told Business Insider about AI. “If you look at our strategy, it’s pretty simple: to fundamentally transform the way the bank operates.”
“Air traffic control”
The bank’s AI infrastructure runs through a “hub and spoke” model: Van Beurden’s small, central team is the hub, while each business line, from commercial banking to human resources, has a designated generative AI lead as a spoke. Those leads are responsible for identifying, scaling, funding, and executing use cases. Alongside his AI duties, Van Beurden is also co-CEO of consumer banking and lending. He suspects that his dual roles make it a bit easier for leaders on both the business and tech sides of the bank to trust him.
Part of his job is identifying which tools can be used firm-wide and spotting potential conflicts before they happen.
“We will do air traffic control when we start to see there’s a collision,” he said. “So if two use cases need something at the same time — ROI, the return on investment, will determine which one will go first. It’s very clear. The highest impact use case will get the priority.”
AI agents are becoming central to some of the high-impact projects, especially those that rely on manual workflows, but they aren’t yet used in typical human-to-human interactions.
Wells Fargo has relationships with several providers, including Microsoft and Google Cloud. The bank ranks sixth on Evident AI’s most recent tracker of AI maturity in the banking industry, above rivals Goldman Sachs, Bank of America, and Citigroup.
Adoption is key
On any given day, Van Beurden likely knows how Wells Fargo is faring on its AI journey. His team tracks AI tools and usage against eight other large banks, measuring daily, weekly, and monthly adoption rates. Wells Fargo declined to share the data.
The retail banking giant doesn’t mandate AI use, partly because Van Beurden thinks forcing adoption can feel like simply “complying to my boss.” He instead wants to build “AI literacy” through efforts such as training, internal demos, and regular communication.
Leaders, he said, must be able to answer a basic question for employees: “Why should I use it? Why is it better? If you cannot answer that question, you fail.”
Three key principles
Van Beurden believes a few key principles will help Wells Fargo differentiate itself: continuous learning, simple messaging, and relentless execution.
A small team is always exploring new LLMs, since models and their capabilities evolve by the day, or even by the hour.
“This is a horse race with constantly different horses leading the race,” Van Beurden said about the AI vendors. Wells Fargo is building platforms that can accommodate new models as they come out, making decisions based on capabilities and costs.
Despite operating in the world of technical jargon, Van Beurden believes in explaining AI as simply as possible to employees. A company, he said, loses people as soon as it starts “hiding behind” technical language.
Still, strategy and messaging mean little without execution.
“It takes a village to get an AI program off the ground and at speed,” Van Beurden said. “Execution is really the thing that matters the most.”
