The most important skill in banking is no longer building financial products, but assembling them. As artificial intelligence remakes financial services, a great division is occurring. The industry is separating into two camps: nimble, AI-native firms building hyper-specialised tools, and incumbent giants whose new competitive edge lies in integrating them. This is not disruption. It is a new industrial arrangement.
Of the 20 newcomers on the 2026 Forbes Fintech 50 list, the most significant are not consumer-facing apps but sharp-edged business-to-business operations. They do not try to do everything. They use AI to solve one problem exceedingly well. Consider Antithesis, a firm that has built an autonomous platform to test software for bugs—a critical function for trading desks where a single error can cost millions. It already counts a major quantitative hedge fund among its clients. Rillet, another new entrant, has raised over $100m in less than a year for its AI-powered platform that automates the tedious work of closing a company’s books. In insurance, Reserv uses AI to overhaul claims processing. It now serves over 100 clients.
These firms have no ambition to hold customer deposits or offer mortgages. Their focus is on perfecting a single, complex workflow and selling it to the institutions that do. They are the precision machinists of the new financial industry, betting they can solve one problem faster and better than a sprawling bank ever could. The venture capital flowing into them confirms the wisdom of that bet. The market for such B2B fintech is projected to grow at a compound annual rate of 32%, reaching $285 billion, fuelled almost entirely by incumbent demand.
Instead, it has partnered with Anthropic, a leading AI firm, to co-develop agents that can automate core functions like trade accounting.
The world’s financial titans are not building competing tools. They are becoming expert assemblers. Rather than invent, they buy and integrate. Goldman Sachs has not tried to build its own large language model from scratch. Instead, it has partnered with Anthropic, a leading AI firm, to co-develop agents that can automate core functions like trade accounting. JPMorgan Chase, while developing some proprietary AI, now has over 450 distinct AI use cases in production, many of which rely on external technology. This pattern is not confined to America. In Europe, Deutsche Bank is pursuing similar partnerships to modernise its operations, signalling a global shift in strategy.
This pivot is a quiet admission that the pace of technological change is too fast for any single institution to master. But weaving these new services into decades-old infrastructure is a formidable task. Industry veterans report that integrating a single external tool can take 18 to 36 months and cost tens of millions of dollars. The work involves a kind of digital plumbing, connecting modern code to rigid, legacy systems and navigating vast data silos. The new core competency for a bank is not just risk management, but technology procurement and systems integration.
When incumbents draw from the same pool of elite AI vendors, the basis of competition shifts. If every bank can license the same claims-processing software from Reserv, the advantage is not the technology itself. It is how well the bank integrates that tool and redesigns its human workflows around it. The challenge becomes managing a hybrid workforce of people and algorithms. The moat is no longer built of capital. It is built of organisational agility.
To be sure, a handful of the largest institutions may defy this trend. A behemoth like JPMorgan Chase, with its vast resources, could pursue a vertically-integrated model. By building its own foundational models and proprietary tools, it could argue for a more secure and customised system, free from reliance on outside vendors. Owning the entire technology stack could become the ultimate prize, creating a fortress of intellectual property that assemblers cannot match. For the vast majority of the industry, however, this is a fantasy. The cost of trying to out-innovate hundreds of focused startups is prohibitive. For them, assembly is the only viable strategy.
This new arrangement also creates a new species of systemic risk. When dozens of banks rely on a single provider like Rillet for a core accounting function, that provider becomes a single point of failure. A bug, a security breach or a simple outage at one of these specialised firms could cascade through the financial system. Regulators are taking notice. A 2023 report from the Financial Stability Board highlighted the risks of depending on a small number of third-party providers for critical services. The Bank of England has gone further, proposing a framework to directly oversee these "critical third parties". The question for regulators is no longer just whether a bank is too big to fail, but whether its software supplier is.
The financial system is being re-plumbed. Monolithic banks are becoming modular platforms, their functions performed by specialist AI tools they have stitched together. This changes the nature of a financial institution. It becomes less a fortress of proprietary technology and more a hub whose value is determined by the quality of its connections. The most valuable employees at a bank in a decade may not be its traders, but the engineers who manage the APIs. For a century, banking has been about managing financial risk. The next will be about managing technological dependency.
