Few corners of finance have absorbed the marketing claims around artificial intelligence as quickly as private equity. Pitch decks promise sourcing engines that surface targets before bankers do, diligence tools that read a data room overnight, and operating dashboards that find margin no human spotted. Some of that is real. A good deal of it is salesmanship dressed in technical vocabulary, and the gap between the two is where experienced investors earn their keep.
Jean-Pierre Conte has spent roughly three decades on the buy side of that question. The American businessman, who joined a San Francisco middle-market private equity firm in 1995 and rose to managing partner, built a record on sector specialization, patient ownership, and value creation through operating improvement rather than financial engineering. He launched his family office, Lupine Crest Capital, in March 2025. JP Conte tends to read a new technology the same way he reads a company: by separating the cash-generating core from the story around it.
A Buy-Side Industry That Adopted Fast
Adoption numbers leave little doubt that the technology has moved past the pilot stage. Deloitte’s inaugural 2025 GenAI in M&A survey, which polled 1,000 senior corporate and private equity leaders in the United States, found that 86% of responding organizations had folded generative AI into their dealmaking workflows, and 65% of those did so within the prior year. Spending tracks the enthusiasm: 88% of private equity respondents had committed $1 million or more to the technology for their deal teams, a higher share than corporate buyers reported.
What the same survey shows, though, is where that money actually lands. Most traction sits in the pre-signing phase of a deal. Adopters pointed to deal strategy and market assessment (40%), target identification and screening (35%), and due diligence (35%) as the areas seeing the most use. The work after a deal closes, where a sector specialist like Conte has historically created the most value, remains the harder problem and the later adopter.
Sourcing Means More Names, Not Better Judgment
Deal sourcing is the function where AI has the cleanest fit, because it is mostly a pattern-matching and screening exercise across messy, scattered data. Models can scan filings, web traffic, hiring patterns, and trade records to flag companies that match a firm’s criteria, then rank them faster than an associate working a spreadsheet. That widens the funnel, which matters in a middle market where many of the best businesses are privately held and never formally marketed.
Wider is not the same as better, and that distinction is where a disciplined investor pushes back. A longer list of names does not tell a buyer whether a founder is ready to sell, whether a management team can scale, or whether a fragmented market can be consolidated without destroying what made the first company work. Those judgments draw on relationships and sector knowledge that a screening model does not hold. For a firm like Jean-Pierre Conte’s, built on knowing a handful of industries deeply, AI sourcing works as a research assistant that surfaces candidates the partners still have to vet by hand. It compresses the top of the pipeline; it doesn’t replace the conviction needed to act.
Due Diligence Gains Speed With a Caveat
Diligence is the clearest near-term win, and the outside data backs it up. Bain & Company, drawing on a survey of private investors representing $3.2 trillion in assets, reported that leading firms using AI-assisted diligence can compress timelines by 30% to 40% while widening the range of material a team can review. A tool that reads thousands of contracts, customer records, and financial statements overnight lets a deal team spend its hours on interpretation rather than retrieval.
Speed carries a caveat that the practitioners themselves raise. Two-thirds of Deloitte’s respondents named data security as their leading concern, and 65% flagged data quality and availability. A model that summarizes a data room can also miss the buried liability or invent a clean answer where none exists, and a private business often lacks the tidy records a model assumes. JP Conte’s value-creation approach has always rested on understanding how a company earns its money, which is exactly the kind of grounded reading that an unverified machine summary can quietly undercut. The output speeds the work; it doesn’t check it.
Portfolio Operations Hold the Slower, Bigger Prize
The largest payoff, and the slowest, sits inside the companies a firm already owns. Bain found that a majority of portfolio companies were in some phase of testing generative AI, but only close to 20% had operationalized it and were seeing concrete results. The distance between experiment and result is the whole game, and it favors investors who treat AI as a tool serving a business priority rather than a goal in itself.
That framing fits a hands-on owner. Backing management teams and improving operations over a long hold is slow work that does not lend itself to a quick technology bolt-on. Conte’s portfolio sectors, including healthcare, financial services, software, and industrial technology, vary widely in how much clean data they generate and how heavily they are regulated, which means a use case that pays off in software may stall in healthcare. A patient owner can run those experiments across a multiyear hold and keep the ones that move revenue or margin, rather than chasing the version that demos well.
What the Discipline Looks Like
A reasonable reading of the current moment treats AI as useful infrastructure inside the investing process, not a new source of edge on its own. Deloitte’s data captures the tension neatly: 81% of private equity respondents expect a measurable return on their AI spending within one to three years. That’s a demand for proof, not a leap of faith, and it sits comfortably with an investor who built a career on operating results rather than narrative.
JP Conte’s track record suggests where he is likely to land. AI earns its place where it removes friction, in screening, document review, and routine analysis, and doesn’t substitute where judgment, relationships, and sector knowledge do the real work. Lupine Crest Capital targets companies generating $50 million to $500 million in revenue across the same sectors he has long backed, the part of the market where human read still carries real weight. The technology changes the speed and reach of the investing process. Whether it changes the returns will depend, as it always has, on what the owner does with the company after the wire clears.