By Gavin Sinclair, Group Head of Diligence and Investigations
Artificial intelligence has earned its place in the due diligence process. It can rapidly analyse large volumes of information, identify potential risks early, and improve efficiency in traditionally labour-intensive investigations.
However, analysts and investigators increasingly warn of a specific danger. Threat Digital has described what might be termed the AI due diligence mirage: the tendency for AI-generated reports to appear comprehensive and authoritative while remaining limited to the digital record available to the system.
AI-generated outputs are often authoritative in tone, well-structured, and linguistically fluent. They produce a coherent narrative that appears comprehensive. Yet a report that looks thorough is not the same as a thorough report.
The danger is that a decision-maker reads an AI-generated report and believes the investigation is complete, when in fact the report is limited to the digital record, which may be incomplete, curated, or deliberately misleading.
The Structural Blind Spot
AI has a structural blind spot that remains difficult to overcome through technical advancement alone. Information that has never been digitised, published, filed, or publicly discussed does not exist within the data AI can access.
Furthermore, AI cannot independently assess whether a source is credible, whether it carries a hidden agenda, or whether information has been deliberately placed in the public domain to shape a particular narrative.
People share sensitive information with people they trust.
An AI system cannot visit a location, build relationships over time, or identify the nuanced risks and ambiguities that experienced investigators often detect through direct engagement.
The 6-Level Source Framework
To understand where AI stops, and human intelligence begins, human source enquiries and discreet conversations to gather actionable intelligence can be viewed through the Ground Truth framework:
- Level 1: Academics, researchers, think-tanks, e.g. university professors and policy researchers.
- Level 2: Industry analysts and trade journalists, e.g. sector commentators with open agendas.
- Level 3: Former employees and ex-directors, e.g. former insiders willing to speak in general terms.
- Level 4: Current industry peers and competitors, e.g. known parties who may share off-record views.
- Level 5: Active insiders with indirect access, e.g. associates of the subject within the sector.
- Level 6: High-level confidential sources, e.g. trusted insiders with direct, privileged knowledge.
AI can access Levels 1 and 2, which largely represent the public record.
The most valuable intelligence often sits at Levels 4, 5, and 6, requiring human trust, relationships, and presence to access.
The Three Irreplaceable Functions of Human Intelligence
Human source enquiries serve three distinct functions that AI cannot replicate.
- Finding what is hidden: Human sources provide access to information that is unavailable through OSINT research or publicly available records, helping organisations identify risks before they become widely known.
- Corroboration: AI can identify a piece of information. Human sources can help determine whether it is accurate, significant, misunderstood, or deliberately planted to influence perceptions.
- Contextualisation: Human sources provide local, specialist, political, and cultural context that transforms raw information into meaningful intelligence.
As Stephanie Bergeman notes, the value of source enquiries extends beyond mere information gathering. They help investigators uncover previously unknown lines of enquiry, validate existing findings, and identify emerging risks before they become public.
A company may appear compliant based on public filings, media reporting, and regulatory records. However, conversations with current industry participants may reveal concerns around governance, payment practices, operational culture, or reputational issues that have never entered the public domain. These issues may not prove wrongdoing, but they can materially affect investment, partnership, or market entry decisions.
The Essential Final Layer
In a market where AI-generated due diligence reports are becoming increasingly common and increasingly similar, it is human capability that distinguishes serious investigative firms from data aggregators.
As Jenny Tsao observes, the future of due diligence lies in combining technology with human judgement.
Technology can perform much of the heavy lifting, processing large volumes of information at speed and scale. Human analysts contribute the context, judgement, and critical thinking that technology cannot replicate.
The experienced investigator remains an essential part of the due diligence process, asking questions that do not arise from the data alone, testing assumptions, and providing context that technology cannot verify independently.
Human source enquiries, discreet conversations with carefully selected sources to gather actionable intelligence, remain a core component of investigative due diligence. In a world increasingly shaped by artificial intelligence, the human dimension remains the difference between information and intelligence.
