What AI Does Well Today
The clearest wins for AI in IT due diligence are in document processing at scale. A human analyst reviewing a Virtual Data Room of 400 documents in a three-week diligence window will, at best, read 60–80 of them closely. The rest get skimmed or sampled. AI changes this constraint fundamentally — the entire VDR is processed, every document is analyzed, and findings are surfaced with references to the specific page and passage that generated the flag.
This matters for a simple reason: the most important document in a diligence process is often not the one you expected to be important. Accounting irregularities, security incidents, and contractual liabilities that determine deal outcomes frequently appear in documents that weren't prioritized in the review plan. Coverage breadth is not a nice-to-have — it's a prerequisite for reliable findings.
Pattern recognition across documents
Beyond coverage, AI is effective at recognizing patterns that span multiple documents — the kind of cross-document analysis that's genuinely hard to perform manually at speed. Revenue recognition patterns that appear inconsistent across financial statements and individual contracts. Security policy claims in management documentation that contradict what appears in technical specifications. Customer concentration that's visible in aggregate financial data but more clearly visible in contract-level review. These patterns are discoverable with structured analysis; they're rarely discovered through sampling.
Information Request List generation and gap tracking
The Information Request List — the structured list of documentation and data the diligence team needs from the target — is one of the highest-leverage documents in any deal process. AI-generated IRLs, built from the actual documents received and gaps identified in the analysis, are more complete, more targeted, and more directly tied to findings than IRLs built from templates. The gap-tracking function — automatically identifying which IRL items remain open at any point in the process — reduces the administrative burden on deal teams substantially.
What Still Needs Human Judgment
Being clear about the limits of AI in diligence is not a hedge — it's a practical requirement for using the output correctly. The places where human judgment remains essential are specific and consistent.
Management interviews and qualitative assessment
Technical competence, organizational cohesion, and the quality of a management team's thinking about risk are not assessable from documents. The management interview — when conducted well — surfaces the gap between what a company knows and what it acts on. AI can prepare the interview agenda with unprecedented specificity (drawing on findings from the document review), but it cannot conduct or interpret the interview itself.
Novel risk assessment
AI systems trained on historical diligence data are, by definition, better at recognizing risks that have appeared before than risks that are genuinely novel. The first time a risk category appears — AI regulatory exposure in 2023, training data IP liability in 2024 — it has to be recognized by a practitioner with domain expertise, not extracted from a model's pattern library. The speed of change in technology means novel risk categories emerge regularly.
Deal thesis judgment
Whether a finding is deal-breaking depends on the investment thesis, not just the severity of the finding. A cybersecurity gap that would be disqualifying in a financial services acquisition might be acceptable in a manufacturing deal where customer PII exposure is limited. AI can flag the finding and score its technical severity; the Investment Committee implications require human synthesis of the finding against the thesis.
What's Coming
The near-term evolution of AI in diligence is happening in three areas that will materially change how deals are assessed within the next two to three years.
Continuous diligence and post-close monitoring
Current diligence is a point-in-time exercise: it produces a snapshot of the target at the moment of the process. Post-close, the acquirer often loses visibility into the technology risk posture of a portfolio company until something goes wrong. The next generation of AI diligence tools will connect to portfolio company systems — with appropriate access agreements — and produce ongoing risk monitoring that flags material changes between annual reviews.
Regulatory AI-in-finance frameworks
Regulators in multiple jurisdictions are developing frameworks for the use of AI in financial due diligence. The EU AI Act's provisions on high-risk AI systems include applications in credit assessment and investment decision support. As these frameworks mature, the use of AI in diligence will move from a competitive advantage to a compliance requirement — with documentation standards for AI-assisted findings that parallel existing audit standards.
Benchmark databases and cross-deal intelligence
Deal teams currently assess technology risk in relative isolation — they know what they found in this deal, but not how it compares to similar companies at similar stages. As AI diligence platforms accumulate anonymized data across deals, comparative benchmarking becomes possible: "this company's patch management posture is in the bottom quartile for mid-market SaaS businesses" is a more actionable finding than "patch management gaps identified." This capability is being built now and will be the defining competitive feature of the next generation of diligence platforms.
Our View
The framing we resist is "AI replaces diligence analysts." That's not what's happening, and it's not a useful frame for deal teams making decisions about how to work. The accurate framing is: AI removes the coverage constraint that has been the binding limitation on diligence quality since the practice began.
When a human team can only read 20% of the VDR, they make implicit judgments about which 20% matters. Those judgments are informed by experience and pattern recognition — but they're also subject to anchoring, to time pressure, and to the specific documents the other side chooses to organize prominently. AI-assisted analysis that covers 100% of the document set removes that constraint and surfaces the documents the team would have missed.
The human diligence lead's job doesn't disappear. It changes: from document reviewer to analyst, from analyst to synthesizer, from synthesizer to advisor. The Investment Committee deserves that level of analysis. AI makes it achievable within the timeline of a competitive deal process.