Enhanced due diligence works best when AI strengthens human judgement

Enhanced due diligence (EDD) helps organisations make informed decisions about high-risk third parties, business partners and transactions. Artificial intelligence (AI) can improve efficiency, but EDD is most effective when AI enhances, rather than replaces, human judgement. For high-risk decisions, the strongest outcomes combine expert-led analysis with AI-enabled speed, consistency and scalability.

• EDD requires more than information gathering; it requires contextual risk assessment and defensible decision-making.
• An AI-only approach to EDD can miss critical information, context, and materiality considerations that experienced analysts identify.
• Manual-only EDD preserves expert judgement but can be slower, more resource-intensive, and less scalable.
• A hybrid, expert-led and AI-enhanced approach preserves human judgement while improving speed, consistency and scalability.
• Compliance teams should assess how providers balance AI capabilities with analyst accountability, source validity and source coverage.

The future divide in enhanced due diligence (EDD) will not be between AI and humans. It will be between organisations that use AI to scale expertise and those that use it to simulate expertise. A key distinction will be whether an organisation can explain how judgement was applied in a defensible way, taking account of local and use case-specific context.

EDD is required when a subject presents elevated risk and a standard review is no longer sufficient. The consequences of weak analysis are materially higher: gaps may lead to a business relationship with a party involved in bribery, corruption, money laundering, sanctions evasion, fraud or other misconduct. EDD therefore provides a deeper, evidence-led assessment to help organisations understand not only who the subject is, but also how they operate, where funds originate, who ultimately controls or benefits from them, and whether the relationship can be justified, mitigated or avoided. The objective is not simply to gather more information. It is to develop a comprehensive, contextualised understanding of risk so that a defensible decision can be made.

This is why much of the current debate about AI in EDD misses the point. The issue is not simply whether AI can produce a report faster, but whether that report provides trusted, defensible insight that can withstand scrutiny in higher-risk decisions. In practice, a defensible report is sufficiently transparent about its sources, research methodology and conclusions. It also explains where AI was used, where expert judgement was applied, and how materiality and risk significance were assessed.

For compliance teams, the strongest model is increasingly a hybrid, expert-led and AI-enhanced approach: AI improves speed, clarity and consistency, while experienced analysts remain accountable for assessing sources, interpreting findings and making the final judgement.

The limitations of AI-only EDD extend beyond commonly cited challenges such as false positives, misinterpreted risks and hallucinations. The deeper issue is that EDD for higher-risk subjects often depends on information that AI tools cannot readily access, interpret or validate.

Material risk information may not be discoverable on the open web. It may instead sit within fee-based court-record systems, subscription regulatory databases or official documents obtained directly from local authorities. Other valuable insight may come from discreet interviews with people familiar with the subject’s business reputation. Asking AI alone to synthesise an incomplete source set can produce an incomplete risk picture, undermining the purpose of EDD. The problem is compounded by the ability of AI-generated output to sound credible and well-reasoned even when important gaps remain in the underlying evidence and analysis.

AI-only EDD is limited not only by access to information, but also by its ability to apply contextual expertise and judgement. The challenge is not simply to identify a fact. It is to determine whether that fact is relevant, credible and material, and whether it forms part of a wider pattern.

A name match may be irrelevant or consequential depending on context. An adverse media reference may be significant or misleading depending on source quality, timing, and jurisdiction. Corporate relationships that appear benign at first glance may take on greater significance once political exposure, litigation history, ownership structures, or local nuances are understood. 

Consider an illustrative example. An entity may have legitimate incorporation documents but little online presence. An AI-only review could produce a clean summary based on publicly available records. A deeper review, however, might reveal links to a sanctioned individual, a lack of genuine operations, and changes in ownership or control shortly after sanctions were imposed. Each fact in isolation may not appear significant, but together they could indicate a heightened risk that a shell company is being used to obscure sanctioned activity. This is where expert judgement makes the difference.

AI-only EDD can therefore create false confidence. A report may appear complete and authoritative while resting on incomplete evidence or inaccurate interpretation. In higher-risk decisions, that is a material weakness.

None of this means that fully manual processes should be preserved. Manual-only EDD has a different weakness: it protects judgement but applies it inefficiently. It can also make it harder to maintain consistent structure, language and quality checks across high volumes of cases, jurisdictions and analyst teams. Experienced analysts add value by assessing source credibility, identifying material risk, interpreting facts in context and determining what further research is required. Yet too much of their time is spent on tasks that do not require that level of expertise, including information aggregation, repetitive drafting, language clean-up, grammar checks and consistency reviews.

The issue is not that manual work lacks value. It is that experienced analysts can spend too much time on activities that today’s AI tools can often handle more efficiently. The result is longer turnaround times, higher costs and reduced scalability.

These constraints matter as compliance teams face growing workloads and increasingly complex risk environments. LSEG Risk Intelligence research found that 90% of organisations had experienced an increase in enhanced due diligence requests over the previous three years. Despite this growth in demand, organisations continued to prioritise quality and oversight, with the survey underscoring the importance of responsible AI to compliance accuracy and risk mitigation. The findings highlight the central challenge for compliance teams: improving efficiency and scalability without weakening judgement or defensibility.

Human analysts should remain responsible for the parts of the process that determine the quality of the outcome. They assess source credibility, resolve ambiguity, identify when deeper follow-up is necessary and interpret findings in the appropriate context.

AI should support the handling of verified information once it has been gathered and assessed. In a responsible hybrid model, AI works from analyst-sourced and verified content rather than simply summarising the information that is easiest to find. It can accelerate drafting, improve readability, flag inconsistencies and support quality review.

This makes the hybrid approach stronger than either extreme. It does not automate expertise and judgement away; it protects them by reducing routine tasks that can distract analysts from deeper assessment.

It also preserves human accountability where it matters most while improving speed and consistency where technology can genuinely help. For compliance teams, the benefit extends beyond efficiency: due diligence reporting becomes more actionable and scalable, and better able to withstand scrutiny in higher-risk decisions.

EDD works best when efficiency and expertise are treated as complements rather than substitutes. AI can improve the speed, clarity and consistency of EDD reports. High-risk decisions, however, still require credible sources, contextual interpretation, investigative follow-up and accountable judgement — capabilities that remain beyond AI alone.

A hybrid, expert-led and AI-enhanced approach is better suited to this reality. It uses AI where the technology adds genuine value while keeping experienced analysts responsible for decisions that depend on trusted sources, defensible reasoning and accountability.

For compliance teams, the advantage is not simply faster or cheaper reporting. It is reporting that remains robust when decisions matter most. The strongest EDD outcomes will come not from automating expert judgement away, but from using AI to scale that judgement without diluting it.