Among financial institutions, there is virtually unanimous agreement about the promise that artificial intelligence (AI) tools hold for improving compliance. As much hype as there has been, however, it is still early days for this use of the technology. In a recent webinar, Oliver Wyman Partner David Choi and Principal David Carretero discussed the most effective strategies for implementing AI in compliance, and what the near future might hold for organizations using these tools.

Here we dive into are some of their key insights and learnings from the event, hosted by Guðmundur Kristjánsson, founder and CEO of the software firm Lucinity.

It’s time to shift from hype to effective AI application in business

Some of the initial buzz over AI in compliance has died down, as there is less of a call to adopt AI simply because it’s a shiny new object. But that doesn’t mean interest has waned. Businesses have recognized the need for practical applications, ramping up pilots for generative AI and accelerating their adoption of machine learning. This work is very important in identifying use cases that will yield the greatest benefits. Quiet as it may seem, the groundwork is now being laid for a big payoff in the near term.    

Humans need final decision-making authority in AI compliance

While compliance teams are bullish on AI use cases and committed to accelerating adoption of the technology, one principle has remained very clear: AI is not replacing humans for final decision making. AI is helping to automate many of the manual steps that compliance professionals must make, but assessing the risk and making determinations on how to manage it are still, crucially, in the hands of people.

AI integration must have top-down and bottom-up approaches

The first challenge associated with AI adoption revolves around the technology itself. Compliance teams need a tool that is flexible and can be customized to accommodate multiple, complex use cases. AI tools also must be seamlessly integrated into workflows — and, further, embedded into employees’ daily lives — safely and effectively. Doing so requires a top-down approach where executives are pushing the technology and offering proper technical expertise, training, and support.

At the same time, there must be a bottom-up component. Defining and implementing the use cases requires that employees closest to the process are intimately involved so that the solutions are comprehensive and practically improve the user’s experience. A platform accessible to many people also enables the company to build and maintain momentum by highlighting the stories of those who have engaged with it successfully.

AI is transforming compliance with promising use cases

One use case for which there has already been real traction is around regulatory change management. For large financial institutions, it is a cumbersome process keeping up with the myriad regulations their business lines and product offerings are subject to, as well as the various geographies and jurisdictions in which they operate. AI systems can help by, for example, identifying policies and procedures that align with the regulations and finding any gaps the firm might have. A related application that is gaining interest is using large language models to streamline the controls a company needs to maintain, thereby saving costs and reducing complexity.

Data collection is another area of focus. AI can aggregate and summarize data for a case, freeing up compliance professionals’ time to review it, make judgments, and perform other higher-value work. Finally, many compliance departments are looking to AI for quality assurance and quality control, such as executing secondary reviews with large population samples.

Prioritize effectiveness for better ROI in AI compliance

Compliance teams should measure their return on investment (ROI) from AI primarily on effectiveness and improving the quality of their risk management. Certainly, there will be efficiency gains that directly result from machines augmenting people for some processes, especially when it comes to automating many of the manual tasks that most compliance professionals perform today. But in practice, most of those gains will be a by-product of the better use cases, improved quality control, and other boosts in effectiveness. With the assistance of AI, there will be a gradual shift in how banks manage their risk, from a reactive to a proactive approach.

Navigating AI in compliance — from getting started to regulatory success

Implementing AI in compliance requires careful planning and testing to secure buy-in and acceptance from regulators and other stakeholders like model risk management and audit. Compliance teams should start with less-complex use cases such as using AI to QA Know Your Client (KYC) files or sanctions investigations, activities that are typically done manually on a sample basis and that could be expanded to larger samples with AI. By starting with simpler use cases, teams gain familiarity with AI technologies and learn how to design solutions that both meet business needs and regulatory acceptance, ultimately paving the way for successful implementation.