McKinsey: AI Cut Due Diligence Time by 46% – But Deal Teams Are Still Hiring More Expert Calls, Not Fewer

If AI is making due diligence dramatically faster, the expert network industry should be shrinking. Fewer hours spent digging through documents should mean fewer reasons to pay $500 to $2,000 an hour for a conversation with an industry specialist. That’s not what’s happening. McKinsey’s research on generative AI in M&A found that AI has cut due diligence time by roughly 46%. Over that same period, the expert network industry has kept growing at 12% to 16% a year, and by one industry estimate, the number of firms using expert networks has grown by around 150% since 2022.

Two numbers moving in opposite directions usually means one of them is wrong. In this case, both are right, and the reason why says a lot about what AI is actually replacing in due diligence, and what it isn’t.

What McKinsey Actually Found

The efficiency gains McKinsey documented are concentrated in specific, document-heavy tasks. CIM extraction, the process of pulling key details out of a confidential information memorandum, used to take a deal team 10 to 40 hours. With generative AI, it now takes under an hour. Quality-of-earnings review and broader due diligence work overall move about 46% faster with AI assistance. Drafting a first version of an investment committee memo, once a 15-hour task, now takes roughly two hours.

Every one of these tasks has something in common: they’re built on information that already exists in writing. Financial statements, contracts, transcripts, and filings are exactly the kind of structured, document-based material large language models are built to process quickly.

Why the Two Trends Aren’t a Contradiction

Expert network calls were never really about processing documents faster. They exist to answer questions no document can answer: why a customer actually churned, whether a regulatory approval is likely to hold up under review, or what a supply chain disruption really means for a company’s ability to deliver next quarter. That kind of insight lives in someone’s head, not in a data room, and no amount of AI-powered document processing generates it.

If anything, the math behind why expert calls are growing gets stronger, not weaker, as AI speeds up the rest of the process. According to industry data, private equity, hedge funds, venture capital, and asset management together make up around 44% of expert network clients and account for roughly 42% of total industry spending. Consulting firms, despite being a much smaller share of clients by count, drive close to half of all industry spend because of how frequently they run projects. As AI compresses the time spent on document review, deal teams have more hours freed up in the deal timeline, and they’re spending a meaningful share of that freed-up time on expert conversations rather than fewer of them.

The Real Shift: Sequencing, Not Substitution

What’s actually changing is where expert calls sit in the process, not whether they happen. A few years ago, a deal team might spend three weeks combing through a data room before they knew which questions were worth asking an expert. Now that AI can surface contradictions, flag unusual patterns, and summarize a data room in a fraction of the time, deal teams get to the expert-call stage faster, and they arrive with sharper, more specific questions.

Some firms have taken this further, using AI to review call transcripts after the fact, cross-referencing what one expert said against public filings or what another expert said on a separate call, to catch inconsistencies before they reach the investment committee. The AI is doing more work around the call. It still isn’t the one having the conversation.

What This Means for Deal Teams

The lesson in the data isn’t that AI failed to deliver on its promise. It delivered exactly what document-processing tools are good at: speed on structured, written information. The lesson is that due diligence was never only a document-processing problem. A meaningful part of it has always depended on judgment, tacit knowledge, and information that simply isn’t written down anywhere, and that part of the process is, if anything, getting more attention now that AI has cleared away the document backlog that used to eat most of a deal team’s time.

For funds building their 2026 diligence process, the practical implication is straightforward: budget AI tools for the document-heavy first pass, and budget real time and money for expert calls at the stage where judgment actually determines the deal. Cutting expert network spend because “AI does diligence now” misreads what the McKinsey numbers are actually measuring.