When Your Customer Sends a Bot

The first piece of evidence for this shift comes from the concept of the machine customer, a non-human economic actor that independently evaluates and transacts. Gartner projects that by 2028, machine customers will drive a significant portion of revenue in financial services. Consider a customer named Maya who delegates her mortgage renewal to a personal AI agent. The agent analyzes her financial data, queries dozens of lenders, and returns an optimized recommendation in seconds, ranking her bank of 23 years fourth on the list. This is not a distant future; the underlying technologies exist today. The quiet accumulation of such delegated decisions will steadily erode the market share of banks that are not the objectively best choice.

The second supporting point is grounded in Ronald Coase's transaction cost theory. Firms, including banks, exist because the costs of transacting in an open market are high. Banks have historically bundled services, relying on the friction of switching, the opacity of alternatives, and the cognitive load of comparison to retain customers. This inertia is often mistaken for loyalty. However, an AI agent acting on the customer's behalf can eliminate these transaction costs almost entirely. When the cost of transacting in the open market collapses, the rationale for the bundled relationship weakens, exposing banks that have relied on inertia rather than superior value.

The final piece of evidence is the emergence of the trust paradox in agent-mediated commerce. To be selected by a customer's agent, a bank must be a trusted data source and execution partner. This requires a demonstrable commitment to acting in the customer's best interest, even if it means recommending a competitor's product. This creates a paradox: the very behavior that builds the necessary trust for the agentic future - radical transparency and customer-centricity - is also the behavior that is most likely to compress the bank's own product margins in the short term. The UK's Financial Conduct Authority (FCA) and its Consumer Duty framework, which mandates that firms deliver good outcomes for customers, inadvertently provides a blueprint for the kind of trust architecture that will thrive in this new environment.

If the thesis of the Identic AI reckoning is correct, financial institutions must fundamentally shift their strategy from being a destination to being trusted infrastructure. The age of relying on customer inertia and opaque product structures is ending. Product leaders and designers must prepare for a future where their primary user is not a human, but an AI agent. This requires a strategic pivot towards creating an Intelligence Layer - a set of capabilities that agents can invoke, rather than an app that humans visit.

Practically, this means banks must first audit their dependency on transaction costs, honestly assessing how much of their customer retention is due to genuine preference versus switching friction. Second, they must build for agent legibility, ensuring their products, pricing, and data are machine-readable and accessible via APIs. Third, they must invest in the trust signal, going beyond regulatory compliance to prove they are acting in the customer's best interest. Finally, they must begin developing a Know Your Agent (KYA) framework to address the regulatory and liability questions that will inevitably arise when agents transact on behalf of customers. The banks that survive will be those that are chosen by the agents, and that choice will be based on demonstrable value and trust, not on brand marketing or historical relationships.