Wall Street, Manhattan, New York
Andrey Denisyuk | Moment | Getty Images
In the feverish race to adopt artificial intelligence, the financial world stands at a critical juncture. The allure of general-purpose AI, the kind championed by tech giants, is undeniable. But for finance, a realm of intricate regulations and specialized jargon, this approach is a dangerous mirage.
It's time for a reality check: finance needs its own AI, not a one-size-fits-all solution.
The idea that a generalized large language model (LLM) can seamlessly navigate the complexities of wealth management, asset management, or insurance is fundamentally flawed. These are domains with their own jargon, private data, specialized workflows and intermediaries, akin to healthcare or law.
A model trained on broad internet data will struggle with the precision required for financial calculations and regulatory compliance. Nor will it infer the multi-step process to navigate decision trees unless provided a framework.
Models fine tuned using private, public and user generated real world data and further enhanced by synthetic or simulated data using foundational large (and sometimes small) language models, for specific use cases using knowledge graphs and detailed workflow schemas to enable reasoning will soon determine the quality of your AI application in finance.
Extracting language from a document is one thing; reasoning and interacting with a specialist in a finance context, with its unique methodologies and schemas, is another. This leads to a natural inference: even the hyperscale horizontal players — the Microsofts and Amazons — and the application developers — the Salesforces and Palantirs of the world — need specialized collaborators in finance. Their generalist AI platforms, while powerful, lack the necessary domain expertise.
Specialized AI
The depth required in areas like wealth management and asset management is simply too granular. These leaders will inevitably need to collaborate with industry specialists who possess the intimate knowledge of workflows, regulations, and user experiences in finance.
The era of bulldozing LLMs through domains is over. The future lies in verticalization, where specialized AI is built in collaboration with experts who understand the intricacies of the financial world. This vertical of complex financial services is also large enough to justify these partnerships. At the same time, traditional financial service firms need to abandon the hubris of using these general platforms to build in-house. The initial impulse to build and own the technology due to domain expertise is understandable — sometimes because vendors are not mature or stable enough in an emerging industry. But this is a costly and often futile endeavor.
The AI landscape is evolving at breakneck speed. What's cutting-edge today is outdated tomorrow. This requires repeated reassessments, a culture of clean sheet thinking and an organizational design that rewards speed. Financial institutions risk getting trapped in a perpetual cycle of development and maintenance, diverting resources from their core business. If a use case is common to the industry, chances are that a fintech focused on that use case will build, scale, learn and maintain its way to a better product faster than an internal team can.
A relevant parallel is the early evolution of CRM systems: trying to build your own in-house solution in the early 2000s when specialized partners emerged is now clearly proven to have been shortsighted. In some cases, where the firm is large — e.g. a JPMorgan or a Morgan Stanley — and has the resources to deploy towards building internal teams tackling use cases unique to them, this may make sense. It may also make sense if the platform is being used to generate and enhance their core intellectual property. Assuming that they can move fast.
As a result, for the generalist technology players as well as for the incumbent financial service firms, the smart move is to embrace partnerships. Firms should focus on what makes them unique — their special sauce — and let emergent fintechs handle the complementary heavy lifting.
In conclusion, the financial world must recognize that its AI needs are distinct. It needs specialized solutions. It needs more strategic partnerships between tech giants and finance experts. It needs traditional firms to resist an isolationist go-it-alone approach. The stakes are high. Generalist technology firms and specialized financial incumbents: beware.
Dr. Vinay Nair is the founder and CEO of TIFIN, a fintech wealth platform using AI and investment intelligence to serve the wealth and asset management industries. Previously, Nair was the founder 55ip, which was acquired by JPMorgan Chase.
