The PE Professional Training AI on the Math That Doesn't Make It Into Models

By James Whitmore, CFA, MBA — 2025-12-20

James Whitmore spent sixteen years in private equity — infrastructure investing at one large-cap firm, then healthcare investments at another. Due diligence at that level is an adversarial process: you're trying to find everything a motivated seller would prefer you not find. He came to IXO because he kept encountering AI tools that had mastered the documents in the data room without understanding the purpose of the process.

Beyond the data room

"PE due diligence has a structural logic that isn't captured in any document," he says. "You're asking: what are the three things that could kill this deal? What are the five things the management team believes that aren't true? What does the customer concentration table not tell you about customer quality? AI trained on CIM summaries and financial model outputs doesn't know how to ask those questions."

Whitmore's IXO work focuses on investment analysis evaluation — testing AI PE tools against the standard of what a thoughtful senior associate would produce at a quality firm. His annotations address not just technical accuracy but analytical framing: did the AI identify the right questions, did it weight the right risks, did it think about the exit scenario correctly.

The gross-to-net litmus test

"The gross-to-net analysis in healthcare investing is one I use as a litmus test. It requires understanding channel mix, payer dynamics, pricing strategy, and rebate structure simultaneously. AI tools that produce clean waterfalls without understanding the commercial logic underlying each line are producing numbers that look right and aren't."

Early warning signs

He also evaluates AI tools in the portfolio monitoring context — testing whether AI performance tracking tools identify the early warning signs that experienced operators recognize before they show up in the financials. "The best PE professionals see a problem in the business six months before it hits the P&L. Working out how to encode that pattern recognition in training data is one of the most interesting challenges I've encountered at IXO."

Read The PE Professional Training AI on the Math That Doesn't Make It Into Models on the IXO blog