Since OpenAI launched ChatGPT-4, the biggest names in technology as well as myriads of wannabe start-ups have scrambled to create marketable solutions powered by generative AI.
It didn’t take long for the analysts to work out that financial services is the sweet spot for gen AI. Banks, insurers and capital markets firms are a lot more complex than the average manufacturer or retailer, with a large proportion of processes that lend themselves to automation or augmentation (see the chart below). They are also subject to more regulation, with a compliance burden that demands a vast amount of data and manual effort. And then there’s the fact that financial services involves an awful lot of language tasks, which gen AI can handle without breaking a sweat.
Financial services firms have, of course, been investing in AI for many years. Banks in particular have aggressively recruited technology and data scientists from universities. This has not only allowed them to build their AI capabilities in areas like fraud management; it has also created appealing research and work environments for academics and other scarce specialists, facilitating further recruitment as well as partnerships with academia.
What then is the likely trajectory of gen AI in financial services? I believe we will see two main trends, working in different directions.
First, the technology giants will continue to invest massively in large, complex systems that address companies’ horizontal functions in a generic but scaled way. Finance, IT, sales and marketing, HR and more will all be profoundly affected. These systems will be expensive to build, train and use, so there are likely to be relatively few contenders for the prize of market dominance.
In many cases, though, it will be difficult to audit, manage and govern these models to the standards a financial services organization would require. In some ways, gen AI in financial services is a bit like the steam engine when it burst onto the scene. There was a lot of initial excitement, but to get value from it people needed faster ways of laying out railroads and manufacturing rolling stock, along with creating the other important infrastructure. The groundwork must be done first for financial services to fully harness generative AI.
The second trend, I believe, will be the proliferation of highly specialized start-ups targeting not only specific industries but also niches within each industry. These vertical systems will be smaller, simpler and cheaper, and enabled by advanced data segmentation and modeling. In financial services they will address many of the 73% of banking tasks and 70% of insurance tasks that are ripe for automation or augmentation.
This verticalization and specialization of the models creates opportunities for fintechs to take over particular niches and train their models on very specific bodies of data. This will enable outsized results. Some of the applications will supercharge existing tasks and processes; others will completely transform them or introduce services or capabilities we haven’t yet thought of.
The allure of this opportunity is reflected in the upsurge of VC funding which to some degree has reversed the slump in fintech funding caused by the rise in interest rates. Meanwhile, many existing fintechs are jumping in and leveraging gen AI in specialized ways, to the benefit of financial services providers. According to Cambridge Centre for Alternative Finance, roughly 90% of all fintech companies are already using some level of AI in their business models, all working simultaneously to carve out their respective niche of the market.
One example is Diveplane, which helps businesses solve data availability problems by generating synthetic data that is private, fully auditable and usable for any task. The auditing of gen AI models and the data they use is highly evolved and incredibly important for financial services players to ensure there is no decision-making bias.
In wealth management there’s Responsive AI, a next-best-action platform that uses gen AI for document analysis and custom advisor email generation. Other examples include SkyHive, a workforce reskilling solution that harnesses gen AI to organize workplace data, automate HR processes and empower a dynamic, skill-based labor economy, and Nuclia, which embeds AI search and generative answers into third-party products.
While FS companies need to manage the risks around explainability, privacy and security, the adoption rate of gen AI powered solutions could be dramatically faster than in any other industry as firms prioritize use cases, including software development and knowledge management chatbots to support front office staff. However, banks and insurers will need to be mindful of regulations around gen AI as they’re developed and ensure that they are following the right standards and guidelines across multiple geographies.
Adoption will be driven not only by the rapid advancement of technology in general, but more importantly by the inherent ability of AI to perpetually self-improve. Recent surveys have shown consistently that a large majority of business executives acknowledge this and report increased investment in the technology.
It’s too early to predict the exact impact of gen AI on financial services, but it seems certain that there will be significant opportunities to increase personalization, augment relationship management and customer service, and improve efficiency through automation of language-intensive tasks. Fintech companies may well be at the vanguard of this movement.
Read the full article here