How to Prepare Your Bank Now for Customer Relationships Led by AI Agents

That possibility strikes at something banks have traditionally treated as a core asset: the customer relationship.

“If that is depreciated or demised,” Cowell says, and an agent begins choosing on the customer’s behalf, loyalty and brand equity become less influential. The customer may remain with a bank for years, but the agent can still move money or shift products whenever another institution offers a better outcome.

Cowell acknowledges why bankers are skeptical. For an AI agent to move money autonomously requires a significant degree of consumer trust, and the technology would need strong guardrails. But that skepticism is exactly why the scenario deserves consideration.

Key insight: The strategic question is not whether every customer will hand financial control to an AI agent. It is whether banks are prepared for a world in which some portion of financial decision-making moves outside the traditional customer-bank relationship.

Make Customer Data Work Harder

If AI agents eventually mediate more financial decisions, banks will need another way to maintain relevance. Cowell sees proprietary customer data as one of their strongest potential advantages.

He uses a simple observation to describe the industry’s current problem: “Your bank knows a lot about you, but yet tells you very little.”

Banks have access to years of transactional and non-transactional information, yet often fail to translate those signals into useful customer engagement. A customer regularly transferring money to another institution, for example, can provide an obvious indication that the relationship may be changing. That signal could trigger a proactive conversation, a relevant offer, or another intervention designed to preserve or deepen the relationship.

Cowell believes banks should invest in cleaning up their data, strengthening the analytics layer around it, and identifying the specific customer behaviors that should trigger action. The goal is to turn data into contextual insight rather than simply accumulate more of it.

Key insight: Data capabilities could become even more valuable as AI becomes embedded in financial decision-making. A bank with billions of transactions and a rich understanding of an individual customer’s history may be able to provide recommendations that a general-purpose AI cannot easily replicate. Cowell even sees potential for proprietary data and contextual recommendations to become a revenue stream through subscription or consumption-based models.

Rethink the Value of Human Interaction

As digital self-service becomes more capable, Cowell sees an important role emerging for human interaction — but only when the context makes that interaction valuable.

That could mean a banker reaching out after a fraud event, during a major life event, or when data suggests a customer may be considering a major financial decision. Cowell gives the example of recognizing that a customer has recently married and determining that a conversation about buying a home may be more appropriate than another generic digital prompt.

This requires banks to rethink how they decide who gets access to a human.

Cowell argues that relying primarily on cost-to-serve can lead banks to make the wrong decisions. Instead, banks should consider the value of the interaction and the potential value of the broader relationship.

Key insight: As the market becomes more fragmented, customers will have more choices about where they keep deposits, obtain advice, and purchase financial products. If the products themselves become easier to compare and more commoditized, the quality and context of the relationship can become a meaningful differentiator.

The same principle applies to distribution. Cowell expects banks that thoughtfully combine digital and physical channels, self-service, and banker interaction to have an advantage. Physical banking remains significant in the U.S., even as digital engagement expands.

Build the Foundation for Multiple Futures

The practical challenge for executives is finding room for long-term investment while still meeting near-term financial commitments. Cowell’s answer is to ring-fence a discrete pool of capital and talent that operates on a different time horizon and uses different measures of success.

That investment does not need to begin with speculative technology. In fact, Cowell points to several foundational capabilities that can produce value on both the near-term and long-term clocks: modern core and banking architecture, clean APIs, rich data, and analytics, greater AI fluency and a culture of product and service innovation.

He returns to those foundations when asked what every financial institution should prioritize today.

What should you do? Modernize the core and infrastructure. Make APIs real-time, clean, and easily exposed. Build AI fluency throughout the organization. Then take a hard look at customer segmentation and engagement models, including how digital and physical distribution should work together.

Those investments also give banks greater flexibility to respond as the market evolves.

Flexibility matters because Cowell does not expect industry success to be determined simply by scale. Larger institutions may have more discretionary capital available for long-term bets, but size does not necessarily make an organization faster at changing its business model, culture, or customer experience. Smaller institutions and neobanks have already demonstrated that they can identify areas of customer friction and compete effectively by borrowing experience-design principles from other industries.

Bottom line: The banks best positioned to remain relevant in 2050 may be those that preserve ownership of the customer relationship while combining their balance sheets, data, AI capabilities, and distribution networks in ways that create value at the moments customers actually need it.