How Wells Fargo is building agentic AI into its workflows

• Key insight: Wells Fargo is re-architecting its systems to embed agentic AI into core workflows.
• What's at stake: Competitive advantage and reduced operational risk hinge on secure, governed agent deployment across firms.
• Forward look: The bank plans to deploy interoperable agent-to-agent banking and customer AI "buddies" that help conduct transactions.

Source: Bullets generated by AI with editorial review

Wells Fargo isn't treating AI as just another IT upgrade. It's re-architecting the entire bank to embed the technology into core workflows, including through agentic capabilities.

"We've broken our strategy into a couple key focus areas," Tracy Kerrins, Wells Fargo's head of consumer technology and generative AI, said during a panel at Money 20/20 last week. "We've had a developer platform that we've been working on and building and enabling and rolling out a lot of use cases. The portion we're partnering with Google on is the employee capability." 

To be sure, Wells Fargo is not the only bank redesigning technology infrastructure and processes to prep for agentic AI. JPMorganChase and Citi have been rolling out agentic AI to developers and Capital One and Goldman Sachs have also been preparing to use AI agents across their organizations.

Wells Fargo will soon be rolling out an enterprise-wide generative AI platform for all employees — part of a broader effort to redesign processes so humans and AI agents can work alongside one another. 

Wells Fargo began working with Google Cloud and Microsoft Azure in 2021 to support advanced analytics. The following year, it launched a digital assistant called Fargo, built on Google's conversational AI platform. By 2025, the bank had adopted Google's Agentspace to develop custom AI agents and introduced Deep Research and NotebookLM to about 2,000 employees, with plans to expand adoption across its workforce. 

These tools form the backbone of Wells Fargo's enterprise-wide generative AI platform, which will eventually be rolled out to all employees. The system is designed to boost productivity and give staff faster access to the information they need to do their jobs.

Agentic AI flips incremental change on its head

While generative AI produces text or answers based on prompts, agentic AI goes a step further by using those outputs to execute, or complete tasks on a user's behalf. The bank's platform-based approach to AI adoption across business units marks a shift from earlier tech modernization efforts that were based on incremental upgrades.

"We've even gone back and looked at some of our existing modernization efforts that were underway and said, do we need to regroup and change those up a little bit based on this new technology?" Ms. Kerrins said.

Rohit Bhatt, general manager and managing director at Google Cloud, said agentic AI has the potential to disrupt traditional technology upgrade processes.

"Is it that we're trying to automate the fifteen steps in an onboarding process, or are we trying to reimagine the onboarding process entirely?" he said at Money 20/20.

Agents that embed themselves into workflows

A key shift in moving from generative AI to agentic AI is the transition from answering questions to executing tasks, a change Bhatt described as a move from "ask me a thing" to "getting something done." Successful implementation, he said, depends on trust in how the system operates, along with credentialed, authenticated transactions performed by agents under human supervision.

In August, Kerrins told American Banker that early use cases include reviewing and summarizing lengthy vendor contracts, identifying know-your-customer requirements, and assisting with post-trade inquiries in the corporate and investment bank. Other agents are being deployed to help branch and service staff handle straightforward customer requests, such as debit card replacements, under human supervision.

Adding guardrails and driving adoption

To deploy AI agents successfully, banks need to ensure regulatory, data-privacy and risk-management protocols are followed. To that end, Google Cloud and Wells Fargo's engineering teams are working together to design the AI platform so it conforms with the bank's internal governance and risk framework.

"It's not just about the fun and cool stuff that we want to do out in the market, but it is also making sure that it is adoptable in financial services with the right security," Bhatt said.

Kerrins added the bank is pairing the technology rollout with employee education and literacy, ensuring staff members across business units understand how to use AI tools responsibly.

"We have a whole literacy program … to try and educate and inform everyone, whether it's from our engineers to our employees to our executives," Kerrins said. "You have to come into this and think, 'forget about everything you know about being a banker right now. Take everything you just learned about this technology and reimagine — how would you do it?'"

Preparing for agent-to-agent banking

Wells Fargo and Google Cloud say they are preparing for a future in which the bank's AI agents no longer work in isolation but can interact directly with one another across systems.

"A lot of the technologies that we've brought to bear in the AI space are focused on making sure [agents are] interoperable and integratable with a very large open ecosystem," Mr. Bhatt said.

"We're already thinking about this whole AI-buddy, AI-assistant, this hot new trend," including how a customer's agents interact with the bank, said Kerrins. "Pretty soon I think we'll all have our [AI] buddy there doing business for us, and part of that business isn't just ordering DoorDash or booking travel — it's banking," Kerrins said.