Investors News

McKinsey bets AI diligence can see 36 months out

McKinsey partners split on AI diligence: Bundy bets on speed, Pralong bets on 36-month precision from proprietary data.

StartupHub.ai Staff

By StartupHub.ai Staff, reviewed by Daniel Singer, Editor

4 min read

David Pralong said the more granular the proprietary data, the more precise the view of the future will be, with what he called a good degree of predictability for about 36 months. Pralong, a senior partner based in the Carolinas and global leader of the transformation practice at McKinsey & Company, made the claim on Inside the Strategy Room during a discussion of its article on using generative AI for outside-in diligence.

McKinsey bets AI diligence can see 36 months out
McKinsey bets AI diligence can see 36 months out

Speed versus precision is the fight.

Will Bundy, a partner in Washington and co-lead of private capital value creation work, argued the biggest win right now is speed and ideation. He said the tools scan the web, internal databases and the confidential information memorandum, then assemble a full list of value creation ideas in minutes, removing what he put at the first four to five days of analyst work while still requiring analysts to test each idea for a topline and bottomline plan. Pralong and Chase, a New York partner in the transformation and strategy and corporate finance practices, pushed a different edge built on proprietary history, including millions of weekly initiative data points collected over 15 years and catalogued expert interviews, which Chase said can turn a vague investment committee claim like save 10 percent on indirect procurement into five to eight categories with savings ranges, costs and time frames.

Sean Brown, the host, pressed the panel on limits. Bundy conceded the models still struggle with accuracy, bias and complex reasoning, citing fabricated legal citations, election deepfakes and hallucinated financial summaries as public misfires that demand oversight, red or blue team agents and fact checking. He said the shift since the 2023 breakout is evidence based citation, multi step logic and deep research with sources and web pulls, which lets a team upload a memo, check what management claims and see where the data came from and why the model doubts it.

That market context is wider than McKinsey. A 2026 study by SS and C Intralinks and Reuters reported by Needl.ai found 84 percent of 400 senior deal professionals used AI in early stage or outside-in diligence and 78 percent in confirmatory diligence, which suggests McKinsey is describing an edge in method rather than exclusive access to the tool. The legal parallel is cautionary, with Spellbook warning that a software vendor does not stand in the position of outside counsel so supervision exposure stays with the lawyer and the firm.

Chase framed the blocker as cultural, not model quality. He cited McKinsey survey work in which he said 88 percent of organizations had launched AI or digital transformation efforts but less than 25 percent reported meaningful progress and less than 10 percent operate at scale, with 75 percent lacking a comprehensive strategy and 70 percent of transformations failing, leading to his rule that every 1 dollar on technology often needs 5 dollars for adoption and workforce readiness. In a separate cut of 40 global general partners across 200 investment teams, he said 50 percent reported some gains like faster research and investment committee memos but only 6 percent called results transformational, while more than 95 percent expect transformational impact within five years.

Laura Borton, a New York consultant in the transformation practice introduced as working from diligence through implementation, offered no separate thesis in the transcript provided. No speaker named the base models used, named a client, or priced the build, and the savings example stayed anonymized by industry and capital cycle context. Pralong advised starting small with one agent in one workflow because, as he put it, seeing is believing when a value creation proposal arrives in two days instead of two weeks, while Chase said lean investor teams will not win by cutting one analyst but by paying a sharper price or passing on the right deal.

If every bidder can produce a month by month P and L in two days from similar web data, what keeps that 36 month forecast proprietary once the model, the memo and the challenge agent are all for sale?

Sources

Get the free daily brief every morning

Every morning the day's AI news, new AI startups and new funding rounds. Free and takes a few minutes to read, unsubscribe at any time.

Join 64,000 other subscribers.

TopicsMckinseyDavid PralongWill BundyInside the Strategy RoomWaveSS&C IntralinksReutersSpellbook

StartupHub.ai StaffDaniel SingerWritten by StartupHub.ai Staff, Editorial teamReviewed by Daniel Singer, EditorAbout

StartupHub.ai Staff, Editorial team

The staff writers of StartupHub.ai, ranging from investment analysts to avid AI tool users, early adopters and critical enthusiasts. Backgrounds span engineering, business and the arts. We hold every piece to rigorous standards of research and review.

Daniel Singer, Editor

Daniel Singer is an AI expert who has been writing about, covering and building AI technology since 2018.

More from Daniel Singer

Test your Agent Readiness now

Paste your website. How well ChatGPT, Claude and the other agents can read and use a site, in about a minute.

Watch

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.