Generative AI: Connecting insight to (inter)action

Over the past decade, AI has enabled banks to derive insights from vast information troves and the value of analytics has shifted from historical insight to prediction, leading to better decisions.

Generative AI is the latest evolution in the AI paradigm shift from automation to intelligence to creation and is bridging the gap between insight and action.

When generative AI systems and large language models (LLMs) can both draw insight from large data sets, as well as understand, interpret, and generate human language, the time and effort required to act on insights is drastically shortened. And with LLMs we can unlock previously impracticable insights and manage the documentation to boot.

With this combination of forward insight and creative capabilities, the financial services sector is ripe for novel innovation, from customer services improvements through to credit risk understanding. A recent report on AI in the finance sector by The Alan Turing Institute found that: “AI has been used only for implementing ‘small benefits' such as cost reduction and process optimisation.” Now with generative AI, we can connect insight to action and even to interaction leading to highly personalised financial advice and much improved customer experience.

One example is compliance with Consumer Duty regulations, which requires lenders to ensure consumer understanding. This can be challenging for automated and self-service processes, but generative AI can help by creating materials for different levels of understanding and language comprehension, improving customer care and connecting analytics to conversational capabilities.

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 automate activities like reconciliations, journal entries, and financial consolidation. The diagram below from a recent CGI article shows the varying impact of AI tools across many financial services operational areas.

More generally, a recent Harvard study on management consultants showed that support in the form of ChatGPT raised all workers to the top tiers of performance. A Cornell University study on potential labour market impacts, found broad applicability of generative AI to work tasks. Starting off, about 80% of people could do 10% of their tasks better and or faster; but once LLMs are integrated in the tools we use daily, everyone will eventually see half their work tasks supported by AI.

Even though AI has much to contribute, it’s important to understand its shortcomings. Human oversight and collaboration are still essential for ensuring the quality and appropriateness of generated responses, but that work is offset by the efficiencies gained. As Meta’s Chief AI Scientist Yann LeCunn said: “Large language models are prone to hallucinations, and have no concept of how the world works, no capacity to plan, and no ability to complete tasks.” Financial institutions should also have a clear and well-defined policy for the use of LLMs outlining the risks and benefits of using them, as well as the steps that will be taken to mitigate the risks.

As generative AI technology continues to develop, we can expect to see it being used in even more innovative ways. We do however need to acknowledge the current limitations and consider how we could use a multi-model approach to create the most value. Using a mix of AI, financial institutions can gain access to insights to help them take action to better manage risk, optimise operations, and gain operating savings - all while improving both customer and employee experience.

I’m always happy to chat about all things AI, so please get in touch to discuss in more detail, or to find out how we can help you on your AI or LLM journey.