Can you trust AI to automate UBO compliance?

In association with:

By Neil Isherwood – Dun & Bradstreet SME & Strategy Lead, Third-Party Risk & Compliance 

Amidst the hype around AI for know your customer (KYC) and ultimate beneficial ownership (UBO) checks, one important question simmers: can you really trust AI to make your compliance decisions for you?

With hallucinations and a “need-to-answer” bias prevalent in most large language models (LLMs), many businesses are tiptoeing rather than charging towards trusting AI with their critical business functions. 

However, a fine balance is required. There is no doubt that AI is revolutionising compliance. Being cautious needs to be weighed up with the risk of being left behind. Striking the right balance and adopting AI safely comes down to keeping your AI decisioning tight – and your data even tighter.

While the natural language processing capabilities of generative AI are relatively new, the machine learning approach to decision automation has been around for years. This provides some valuable pointers. For starters, you need to tightly control your decision variables. Essentially, you need an algorithmic approach that is predictable, repeatable, and observable. That means assessing the same parameters, in the same way, each time – and only using clearly-defined data that you can trace back to its source.

Once you have your decisioning process tight, you need to ensure your data is too. Here, it’s important to make the distinction clear between training data for AI (i.e. historical datasets AI learns from and uses to generate answers by default) and real-time decisioning data for AI (i.e. a snapshot of live data that is accurate and up-to-date at the moment of retrieval). In the context of KYC onboarding, it is this real-time decisioning data that should form the basis of assessment – and beware of AI filling the gaps by intermingling the historical datasets it was trained on, which are no longer accurate or up-to-date.

In most respects, the data characteristics needed for AI automation are the same as those needed for ‘traditional’ compliance automation. Data needs to be accurate, up-to-date, and traceable. When it comes to beneficial ownership, that means connecting multiple UBO data sources to create a ‘living’ source of data. This is the essential starting point for incorporating AI into your compliance onboarding and monitoring processes.   

In our recent Ultimate Beneficial Ownership Guide, we did a deep dive into the data that is required for KYC automation and always-on monitoring. It’s a helpful resource for understanding what data characteristics are needed, the current challenges in sourcing accurate UBO data, and the newly emerging data and technology solutions that are simplifying UBO compliance. I encourage you to download the UBO guide and have a read. It’s a great first step to making sure that your UBO data is ready for AI.