What does artificial intelligence change in private equity value creation?
AI does not change what a fund is looking for — EBITDA growth — but it changes who can go and get it, and at what cost. It absorbs the analytical work that used to occupy whole teams: reading an entire data room, consolidating reporting across ten portfolio companies, preparing a monthly review. One senior operator, properly tooled, now covers what once required a team, which puts operational value creation within reach of small and mid-cap funds. And it moves the boundary of the deliverable: the recommendation is no longer enough — what counts is the tool that carries it out and stays in service after the provider leaves.
Why are funds moving now?
Because yesterday’s levers no longer deliver. In its Global Private Equity Report 2026, Bain sums up the coming decade in one formula: “12 is the new 5”. With high borrowing costs and entry multiples that refuse to soften, the return once achieved on 5 % annual EBITDA growth now demands 10 to 12 %. Financial leverage and multiple expansion no longer close the gap: it has to be found in the operations.
And adoption is already widespread on the transaction side. Deloitte’s GenAI in M&A Survey (2025, around a thousand senior executives) finds that 86 % of dealmakers use generative AI in their M&A processes, 65 % of them having started within the past year, and that 88 % of the private equity funds surveyed have invested more than one million dollars in it.
Where does AI actually act in the investment cycle?
Before the investment, on reading. The same Deloitte survey puts usage at 40 % in deal strategy, 35 % in target identification and 35 % in due diligence. In practice: working through a full data room in hours, surfacing non-standard contracts, cross-checking market data, exposing inconsistencies that sampling would have missed.
After closing, on steering and on execution. Consolidating heterogeneous reporting without waiting for the close, keeping a value creation plan current instead of a file reworked the night before the committee — and above all automating the processes that cost real money in portfolio companies: quotations, tender responses, customer service, industrial planning.
The useful distinction is not between “strategy” and “technology”. It is between what AI saves in analysis — time — and what it earns in operations — margin. The first shows up in weeks; the second is measured in the P&L.
Why do most AI investments produce nothing?
Because they start with the tool instead of the subject. A fund deploying a generic assistant across ten portfolio companies gets ten experiments and no line of result. What produces an effect is the opposite path: identify the two or three processes that genuinely weigh on margin, then build the tool that handles them.
The second reason is adoption. A tool nobody uses is worth nothing, and usage cannot be decreed: it is built with the teams who will do the work, during the engagement, not in a training session afterwards.
The third is ownership. In a small or mid-cap fund, AI often has nobody to carry it: not the investment team, busy with deals, and not the portfolio companies, waiting for a doctrine. Without an owner, the subject stays an agenda item.
Who does this work today?
Three families of players, serving different funds. The large firms — Alvarez & Marsal, Bain, EY-Parthenon — have built AI-enabled value creation practices sized for large portfolios. European platforms have formed around the same ground, such as OMMAX and Singulier, which merged in 2026 into a group of some four hundred people across Paris, London and Munich. Finally, small operating partner outfits work the small and mid-cap end, where the others do not reach.
On the fund side, the shift is measurable: according to France Invest’s Operating Partners Club and Alvarez & Marsal, 84 French management companies had operating partners in 2024, against 47 in 2019, for 185 people — a headcount multiplied by 2.4 in five years. But small and mid-cap funds rarely have more than one, and often none.
Off-the-shelf tools or bespoke ones?
Both, in that order. For everything standard — office work, document search, transcription — market products are better and cheaper than anything one would build. Bespoke work is only justified where the process is specific to the company and where it weighs on margin: precisely what vendors do not cover, since they sell the common denominator.
The real criterion is not the technology but what remains. A plan dies with the engagement that produced it; a tool in production keeps running. That is why we deliver tools, not only recommendations.