AI firms like OpenAI, Anthropic embed deeper in bank operations

The movement of AI companies into banking operations is happening across multiple fronts, each revealing a consistent pattern: external AI providers are establishing footholds inside financial institutions through practical, immediate-value applications rather than wholesale technology overhauls. Anthropic's participation in the Financial Conduct Authority's Supercharged Sandbox exemplifies this approach. Through the programme, selected firms will use Claude to explore applications in areas such as payments, fraud prevention, compliance and responsible AI adoption. OpenAI is following a similar path through partnerships with financial institutions, including BBVA, where its technology is being applied across customer experience, risk analysis, operations, software development and employee productivity. Wells Fargo, meanwhile, is using Google Agentspace (now part of Gemini Enterprise) to support employee search, information synthesis and agent-assisted workflows.

What these examples reveal is that major AI providers are entering banks through internal productivity and operational workflows. Rather than reshaping entire technology estates, they are helping employees find information, analyse content and complete knowledge-intensive tasks more efficiently. This starting point is important because internal workflows offer a relatively controlled environment in which banks can test the technology's value, limitations and governance requirements. They also create a path for wider adoption if the tools prove reliable and useful. The current reality, according to the article, is not that AI companies already operate across every banking function, but that they are becoming part of the operational environment and their role may expand as banks gain confidence.

The strategic value to AI companies is substantial. Banks combine large volumes of high-value decision-making, regulated data and long-standing customer relationships. The company whose model becomes embedded in that environment gains reference credibility in the most demanding enterprise context, and long-term revenue that is difficult for competitors to displace. That value grows over time: once a model becomes part of how employees access knowledge, evaluate information and complete critical tasks, it starts to influence architecture decisions, procurement standards and investment planning across the institution. Success with one bank in a highly regulated environment strengthens the provider's position with every other bank, and across every regulated sector where control, resilience and accountability matter to the buyer.

Yet the article cautions that model capability alone is not the differentiator. There is no single "best" model for a bank. A model that performs well analysing complex documents may be unnecessarily expensive or slow for a high-volume customer-service process. A smaller, more predictable model may be better suited to a narrowly defined operational task, while a more capable model may be justified where the work requires deeper reasoning across multiple sources. The long-term differentiator will not be access to a model that competitors can also purchase, but the bank's ability to select the right AI model for each level of risk, integrate it with trusted data and change course without rebuilding the entire service. As the article notes, newer models do not automatically mean better for every banking use case.

The most significant challenge ahead is one of control. As companies such as Google, OpenAI and Anthropic become more deeply embedded in financial services, control will become one of the defining questions of AI adoption. The influence of large AI providers is not inherently problematic—they can bring advanced research, scalable infrastructure and experience gained across industries, helping banks modernise faster, improve productivity and experiment with new services without building every capability from the ground up. The tension emerges as successful experiments become embedded across the institution. The more functions that depend on the same provider, the more its technology, development choices and commercial direction become part of the bank's operating environment. What begins as an effective partnership can gradually reduce flexibility if changing course later becomes complex or expensive. The article suggests the relationship will be defined by balance rather than resistance, with the opportunity lying in combining the strengths of AI providers with the customer context, regulated responsibility and institutional judgement that banks must retain. Banks that benefit most will treat AI provider relationships the way they already treat other critical infrastructure decisions: with clear standards, deliberate redundancy and an operating model that assumes providers will change over time. Accountability for the outcome, the article concludes, sits with the bank.

Anthropic, OpenAI and Google are advancing a strategic shift in how they position themselves within financial services. Rather than attempting to become banks themselves, these companies are embedding their models into the operational fabric of existing institutions—starting with employee-facing productivity tools and gradually expanding into compliance, risk analysis and customer experience. This approach is deliberate: banks combine large volumes of high-value decision-making, regulated data and long-standing customer relationships. Once an AI model becomes embedded in that environment, it gains what the article describes as reference credibility in the most demanding enterprise context, along with long-term revenue that competitors find difficult to displace.

The real strategic value lies in how this foothold expands over time. When a model becomes part of how employees access knowledge and complete critical tasks, it begins to influence architecture decisions, procurement standards and investment planning across the institution. Success with one bank in a highly regulated environment strengthens the provider's position with every other bank and every other regulated sector where control and accountability are paramount. This creates a reinforcing cycle: early adoption by prestigious institutions like Wells Fargo and BBVA validates the technology and makes it easier for other financial institutions to justify similar investments.

The central tension the article identifies is one of control and dependency. Banks gain access to capabilities that would be difficult and costly to develop independently, potentially accelerating innovation in customer service, compliance, risk and operations. Yet the more functions that depend on the same external provider, the more that provider's technology, development choices and commercial direction become part of the bank's operating environment. The article suggests the most sophisticated banks will manage this relationship by treating AI providers like other critical infrastructure—with clear standards, deliberate redundancy and an operating model that assumes providers will change over time—ensuring accountability for outcomes remains with the bank itself.

No comments yet. Be the first to share your thoughts!