Every private equity deal model should carry a line for AI: the revenue AI could reprice or take away during the hold, the costs it could remove, the evidence behind each figure, and the manager who will report it after close. A portfolio company preparing for a sale can build the same line on itself before a buyer does.
In EY's Private Equity Pulse, published in July 2026, the global value of tech-focused private equity deals in the first half of the year fell 50 percent from a year earlier, while non-tech deals rose 9 percent, and EY says sponsors now set a higher bar where "AI disruption is reshaping assumptions around growth, pricing power and long-term defensibility." Bain's midyear private equity report, published in June, ties February's nearly 30 percent drop in public market software valuations, since partly recovered, to uncertainty about AI's impacts.
In a Bain and StepStone survey of 103 private equity investment and investor relations professionals, mostly in North America and Europe, run from December 2025 to January 2026 and published in March, GPs most often named due diligence and deal sourcing when asked where generative AI returns the most within their firms, yet 39 percent of GPs do not expect AI to have any material financial impact on portfolio companies in 2026. That diligence work can fill the line, putting what AI finds in the data room about the target's exposure into the model for the investment committee to price.
Revenue at risk lists each revenue stream whose price or volume AI could change during the hold, with its share of the total, and costs removable lists what AI could take out by function, with the year each saving would land. Evidence is what stands behind each figure: a clause in the customer contracts, a test on the target's own data during diligence or a published benchmark. A saving with no evidence behind it goes into the model at zero, while a revenue stream nobody has tested for AI exposure goes into the downside case.
EY's US AI Pulse survey, which polled 534 US senior leaders at senior vice president level and above in April and May 2026 and was published in July, asked them to look ahead five years: 82 percent of those whose organization invests in AI anticipate that "the traditional enterprise per-seat Software-as-a-Service (SaaS) pricing model will become less relevant in their industry." For a software target, that puts seat-based revenue on the revenue entry.
For a medical billing company, the line might read: revenue at risk, the fee clients pay per claim or as a share of collections, which they may push down as AI cuts the work behind each claim; costs removable, the hours coders and denial staff spend on each claim; evidence, a de-identified sample of the target's own claims, run through an AI tool during diligence with the seller's agreement and checked by its coders; owner, the chief operating officer.
After close, the line moves into the value creation plan and the owner reports each entry at the monthly review. Repricing usually arrives at contract renewals and savings land over the hold, so day 90 shows the board which AI assumptions are on track in the company's own numbers and which are still untested.
Bain's midyear report notes that rising interest rates and shifting market dynamics had changed deal math well before the upheavals of early 2026: a deal that might have gotten by with 5 percent EBITDA growth ten years ago now needs 12 percent to reach a 2.5 times return over a five-year hold.
