AI and risk: re-imagining EDD

AI continues to sweep across numerous industries, offering ever-evolving tools to enhance speed and efficiency. The potential uses and applications of this technology are vast, but can generative AI streamline and optimise the enhanced due diligence (EDD) process without inadvertently increasing risk?

• EDD is a vital additional risk management step where heightened risk is identified and reviewed.
• AI can optimise the EDD research process, but it is essential to adopt a low-risk approach when using this technology in the due diligence space.
• A low-risk approach must retain a core element of human oversight.

Every customer or third-party relationship has the potential to introduce risk to your organisation, and this means that due diligence to identify possible financial crime and sanctions risk – and understand exactly who you’re doing business with – is an imperative step in every business relationship.

Where heightened risk is suspected or detected, initial screening may not be enough, and EDD becomes necessary. EDD reports deliver vital information about your most critical or risky relationships and can be crucial when deciding whether to engage with a new customer or third party.

Specific instances that should trigger EDD include:

• Where the customer or third party is high risk
• Where there is high value attached to the relationship
• Where initial findings suggest that red flags may be present

Robust EDD offers additional insights that typical initial screening does not. These reports include risk-relevant information gathered from a variety of credible sources and deliver detailed background information and contextualised insights on subjects. 

EDD, when executed well, helps you make informed decisions about engaging with new customers or third parties while meeting compliance and regulatory obligations. 

However, the EDD research process is time-consuming and resource intensive. Rushed reports raise the risk of quality errors that could result in considerable damage to your company. It is therefore important to consider how innovation can be harnessed to conduct high-quality EDD with speed and efficiency. 

AI offers a viable solution, but is there a risk attached to using AI in EDD?

All new technologies come with a degree of risk – and AI is no exception. There are, understandably, several concerns around the safe application of AI to business processes in general, and to EDD in particular.  

Adopting a low-risk approach can offer EDD researchers the tools they need to work smarter, work faster and deliver more robust EDD reports quickly, but what does such an approach look like?

Essentially, a low-risk approach means taking steps to avoid common pitfalls – such as AI hallucinations and bias, which is the biggest risk when using AI. 

Hallucinations are the tendency of some generative AI models to deliver incorrect information where output errors appear legitimate.

AI bias refers to results that skew the original training data or AI algorithm, leading to distorted outputs and potentially harmful outcomes. Bias can creep into algorithms in multiple ways, for example, via training data, which can include skewed human decisions or reflect historical or social inequities. Bias can also be introduced in the AI model via flawed data sampling. 

So how can you streamline and optimise the EDD process using AI, without introducing more risk?  

The safest approach is “grounding”, which means setting restrictions for the input of data. In the EDD process, the input provided to the AI writing tool can be grounded to research elements completed by an analyst using their existing research methodology. The AI can then simply use the collected data – that has been quality assured by human analysts – to create a summary. 

The key ingredient to minimise risk is human oversight. Where the application of generative AI capabilities simply creates a tool to augment the writing process, but analysts retain full control and accountability, risk is low. 

This approach optimises the more basic elements of report construction and frees analysts to concentrate their attention on areas of greater value-add, such as analysing the collected data and distilling the insights that drive better decisions.  

The benefits are many: leveraging a combination of generative AI and existing research capabilities within the report writing process boosts efficiency across an array of day-to-day tasks, such as creating executive summaries more efficiently. This approach also helps to improve writing quality and consistency across reports.

At LSEG we have adopted a low-risk approach to integrating AI into our existing EDD report processes. This conservative use of AI augments and optimises our in-house EDD capabilities, which are supported by analysts speaking 60+ languages across 200+ jurisdictions and producing tailored reports to meet business-critical needs. 

By leveraging this powerful combination of leading technology and trusted human intelligence, we deliver 30, 000 EDD reports a year, enabling businesses to understand and mitigate dynamic risk without slowing the pace of business.