Generative banking: How financial institutions are embracing the new AIOctober 25, 2023 | By Andrey Slivka
For years, artificial intelligence has been a valuable part of banking technology, aiding in fraud detection and data analysis. As machine and deep learning have evolved, so have their roles. They’re now also assisting in identity validation and risk assessment and helping providing personalized customer experiences.
Enter generative AI — a niche within deep learning. Having hit the mainstream in late 2022, this technology has the potential to reshape banking through its ability to generate content and engage in humanlike interactions.
Developing consumer-facing solutions like AI-driven financial advisors or conversational banking would take time, partly due to issues associated with integrating gen AI into existing systems and partly because of the need to make sure that using it doesn’t lead to data privacy and veracity challenges. Today, banks are eyeing generative AI for game-changing internal applications, including knowledge management, credit decisioning and cybersecurity.
Three key impact areas
Large banks tackling knowledge management initiatives — sharing relevant information within an organization — often must grapple with siloed data, a problem worsened by legacy or outdated tech. Generative AI can help by processing vast amounts of data and promptly delivering information to those who need it. The ability to dynamically synthesize data means faster access to regulations for legal teams, product documentation for engineers, and branding guidelines for marketers — all of which boost efficiency.
As for credit decisioning, traditional lending often suffers from slow processes and complexity — think of acquiring a home mortgage. Working with open banking methods and in concert with other AI models, generative AI can streamline this. It can help serve those without standard credit histories or with thin credit profiles by considering alternative data, synthesizing information for lenders, supporting decision-making and proposing lending strategies. Moreover, AI can guide applicants seamlessly through procedures, enhancing user experience.
In cybersecurity, gen AI trained on vast datasets, including malware and synthetic data, can predict cyber threats, simulate security scenarios and pinpoint anomalies — providing a richer, real-time defense strategy. Security teams can use the technology to create models predictive of cyberattacks and propose methods of countering them.
Difficulties to finesse
Generative AI isn’t without its challenges.
First, there’s the data exposure issue. Gen AI models ingest as training data the information to which they’re exposed, such as transaction or chat information, and then repurpose it. This is a concern in banking, where data security is paramount and compliance with privacy regulations is required.
Information accuracy is another concern. Generative AI can sometimes produce inaccurate or “hallucinated” information — a challenge for knowledge management.
Third, AI might perpetuate bias. If a data pool reflects that a certain demographic has historically received fewer loans, the AI application could take that fact as prescriptive and discriminate against that group. Understanding and mitigating this is crucial.
Finally, if gen AI can be a powerful cybersecurity tool, it’s also true that criminals can exploit it, using it to produce “deep fakes” or churn out iterations of deceptive email copy in phishing expeditions. Security specialists will have to come to terms with this tech’s two-sided nature and stay a step ahead of bad actors.
Implementing gen AI safely
Financial institutions could build their own models to mitigate risks of various sorts, ensuring control over them — for example, by restricting their use to select clients or making them purely internal solutions so that the data security stakes are lower.
Banks might also adapt commercially available models for their specific needs, ensuring tighter controls. Opting for commercially available models can also be cost-effective for many institutions. (Well-known financial institutions have recently used or are using these strategies.)
The takeaway? “Restricted” is the watchword for many financial entities. The focus is on secure, regulated and in-house applications.
Starting small and ramping up
Don’t expect a complete gen AI overhaul of enterprise tech overnight. Generative AI will likely integrate gradually into banking, starting small with internal use cases and, as challenges are addressed, expanding toward more ambitious and public-facing deployments. But given its potential, it’s poised to deliver a significant transformation in bank operations over the next several years.
Look for a comprehensive exploration of generative AI’s role in banking in the next issue of Mastercard Signals.