Financial Services' Generative AI Balancing Act

With regular announcements about new gains Generative AI models have made, the big question for financial institutions is how to apply it safely and responsibly.

For an industry that thrives on balancing risk and reward, completely ignoring this much-hyped tech will not be an option for much longer. According to the Gartner Financial Services Business Priority Tracker Survey from February 2023, disruptors were already investing in the new technology – giving them at least a half-year head start at this point. Then, the biggest brands were closely monitoring the situation, and the largest group was waiting for certain improvements.

The wait-and-see approach seemed sound with so much in flux from a regulatory perspective

Much of the regulatory dust has not yet settled, but it’s become clear that companies of all sizes and in all sectors need to learn how to harness the gains the new tech can provide, and internal use cases are the best place to begin.

Let’s examine this through the lens of existing laws; in the UK the FCA’s Consumer Duty regulations require explainability in decisioning and proof of compliance, neither of which is possible with LLMs such as ChatGPT fully in charge of customer communications. However, the key thing to remember is that generative AI is fundamentally an assistive tool and shouldn’t be employed for full automation. Employees working with AI support have been shown to do better quality work faster, which nullifies the adage: “Faster, better, cheaper – pick two”, proving that all three are probable.

Generative AI has been shown to improve outcomes on a range of high-value tasks

For example producing drafts for internal and external financial reporting, which will save considerable time during month and quarter end. It could also be used to do the heavy lifting in activities like reconciliations, journal entries, and financial consolidation.

Content creation and adaptation for marketing is an area where generative technologies can be used in a hands-on way, making it safe while saving enough time and costs to be irresistible. Marketing can be customised to niche groups or completely individualised without an undue amount of effort.

This individualised approach could also be employed during on-boarding to help meet regulatory requirements like the Consumer Duty outcome of consumer understanding. Use in response creation – both to tailor the wording to the correct reading level or primary language of the individual, and to determine whether they understand the product and the implications of signing up are both valuable endeavours that aren’t feasible at scale without the help of technology.

Use here should be as a timesaver and any responses it generates should be checked

The not-quite-ready-for-prime-time tech might not be suitable for unmediated customer-facing interactions but behind the scenes it can shine, even as a tool for data extraction and analysis and fraud detection. To use it successfully and safely, both employees and companies need to know how to navigate its pitfalls. It’s important that AI governance frameworks are updated for generative AI and that companies provide tools for safe use of AI and clearly delineate expectations around use and outcomes.

Despite its drawbacks, the potential upsides shouldn’t be ignored, and steps for safer implementations reduce risk enough to make it worthwhile.  The key to getting the most out of a technology touted as capable of anything lies in discovering innovative applications, rather than merely integrating it into our existing workflows. How can we transform our work processes to take advantage of its potential fully? By proposing new ideas, devising strategies, experimenting with various formats, and exploring different problems you’ve not yet been able to solve, to see if it has a part to play. Whether you're initiating a project or refining an existing one, it's essential to understand that the outcome is not the final destination; it's just a point where you can choose to take over and direct the next steps.

The best way for banks to start down the generative AI road is to:

1. Explore behind-the-scenes rather than starting with customer-facing use cases.
2. Employ it alongside people to enhance their expertise or enable faster execution.
3. Test thoroughly.

Using this approach they can make future investments that are sound, logical, and offer value while maintaining security and data integrity.

If you want to find out more about how Generative AI can work for your organisation, please don’t hesitate to get in touch for a chat.