The buyer will price your AI maturity.

I sat inside AltaReturn through its $500M acquisition by Vistra. From the seller's side of the table, I watched what diligence actually looks like when the cheque is being written.

The buyer didn't want a vision. They wanted evidence. They wanted to know, function by function, where the work was being done, by whom, at what cost, and what would still be true 18 months after close.

Three years on, that experience is the most useful lens I have for reading the current AI conversation in private equity. Most of what gets written for GPs treats AI as a value creation lever. Useful. EBITDA-positive. A source of margin. All of it true. Almost none of it differentiating any more, because everyone is saying it.

The harder question, the one that will define returns over the next two exit cycles, is not whether AI creates value inside a portfolio company.

It is whether the next buyer will pay you for it.

That is a different question entirely. And the answer, on current evidence, is that most GPs are walking into it underprepared.

The diligence shift has already happened

The buyers I speak to, both strategic acquirers and secondaries, have stopped asking whether a target uses AI. They assume it does. The questions have moved on, and they are sharper than the seller-side narrative is ready for.

They want to know which functions have been redesigned, not augmented. They want governance artefacts. They want evidence that automation is operating in production with audit trails, not sitting in a pilot deck. They want to understand which roles have been removed, which have been redeployed, and what the run-rate cost base actually looks like once the human-AI operating model has settled.

A polished AI narrative is no longer a premium. It is the entry ticket. The premium now sits with the assets that can prove, in their numbers, that the operating model has already been rebuilt and is producing the savings the seller is claiming.

This is not a 2027 problem. It is happening in live processes right now. Vendor diligence questionnaires are asking for L3 process maps showing automation coverage by task. QofE adjustments are being contested on the basis of whether claimed AI-driven savings are sustainable. Where a sponsor cannot defend the operating reality behind the AI story, the buyer either drops out or reprices. I expect that pattern to harden materially over the next 12 months as more buy-side teams build standing AI diligence capability.

The gap between AI narrative and AI operating reality is becoming a discount in the bid.

The size of the gap, in the numbers that matter

Take a private equity SaaS asset, £30M EBITDA, exiting at 10x. That is £300M of enterprise value.

Now imagine the buyer concludes, six weeks into diligence, that 25 percent of the claimed AI-driven cost savings are not yet annualised, governance is informal, and the operating model still depends on substantially the same headcount the asset had at acquisition. They don't walk. They reprice.

A turn of multiple, in that example, is £30M.

Two turns is plausible where the gap is structural rather than presentational. That is £60M. On a fund-level basis, two assets with that profile in the same vintage compound into the difference between hitting and missing carry.

This is the calculation that sits underneath the polite "the buyer had concerns about the AI story" line in the post-mortem. It is rarely framed that crisply at the time, because it gets absorbed into broader negotiation. But the number is real, and the next 18 to 24 months of exits will be the period in which the market starts pricing it explicitly rather than implicitly.

If you are sitting on hold-period assets you intend to exit in 2027 or 2028, the window to do anything substantial about it is now. Not next year.

Why this gap exists, and why it will not close on its own

The reason the gap exists is structural, not technical. The technology is not the bottleneck. There are competent execution firms that can deploy agentic workflows in 12 weeks. Methodologies exist. Tooling has stopped being the constraint.

The constraint is sponsorship.

PortCo CEOs are running quarterly numbers. They have a board, a value creation plan, and a finite attention span. AI sits as one of perhaps fifteen workstreams competing for their time. Without an external force pulling AI maturity into the priority stack and holding it there across the hold period, it slips. It always slips.

Operating partners know this. The good ones are stretched across eight or ten assets and cannot embed inside any single one for long enough to drive the operating model rebuild. The diagnostic gets done. A roadmap gets written. Two quick wins get delivered. Then the CEO turns to the next thing, and the programme decays into a managed service relationship that delivers steady, unremarkable savings.

That is not what the buyer will pay a multiple for. The buyer will pay for evidence that the operating model is structurally different. Different cost base. Different headcount shape. Different unit economics. Different governance. That requires sustained, opinionated pressure from someone whose only job is to make sure the AI thesis lands in the numbers before exit.

That person does not currently exist in most portfolios. The execution firms cannot play that role. They are accountable to delivery, not to the buyer's diligence question. The CEO cannot play that role. They are running the business. The operating partner cannot play that role at the depth required, because they are spread across the portfolio.

This is the sponsorship layer, and it is empty in most funds I look at.

What changes if you treat AI as exit infrastructure

If the buyer is going to price AI maturity, then AI readiness stops being an operational improvement programme and becomes an exit positioning programme. That reframing changes who owns it, when it starts, and what counts as success.

Three things follow.

The work has to start in year one of hold, not year three. By the time a sale process opens, the operating model rebuild needs to be 18 to 24 months into operation with a settled run rate. Anything short of that reads as a story rather than a fact, and the seller spends the diligence period defending instead of pricing.

The metric is not cost saved. It is cost base shape. A buyer looking at two assets with identical EBITDA will pay more for the one whose SG&A is 40 percent lighter on FTE and 60 percent automated, because that asset is more defensible, more scalable, and easier to integrate. Headcount removed is a more durable signal than savings booked. The market premium for genuinely rebuilt operating models will widen as more assets cluster around polished narratives without operating substance.

The GP needs someone in the room whose mandate is the buyer's diligence question, not the CEO's quarterly one. Someone who reads the asset the way a buyer's team will read it in three years, and works backwards from that into the operating decisions that need to happen now.

The uncomfortable conclusion

The GPs who will outperform on the current vintage will not be those with the best AI strategy. The strategy layer is crowded and largely indistinguishable across funds.

They will be the ones who treated AI readiness as exit infrastructure from the day they signed the SPA, and who put the right person in the right portfolio companies to hold that line over four years against the daily pressure of running the business.

That is a sponsorship discipline, not a technology discipline. It does not require more vendors, more pilots, or larger consulting budgets. It requires a single, senior, accountable person per asset whose mandate is exit-shaped.

If you are a GP with assets entering exit prep over the next 24 months, three questions are worth sitting with before the next portfolio review.

How would your operating partner describe the AI thesis of your most exit-ready asset, and how would the next buyer's diligence team describe it? If those two descriptions are not nearly identical, you have a pricing problem you have not yet seen.

What is the run rate, in months, of the AI-driven savings in your model? Anything under 18 reads as narrative.

Who, specifically, in the asset is accountable for the answer the buyer will get when they ask these questions in 2027? If the name is not obvious, the gap is already open.

The execution layer is solved. The strategy layer is crowded. The gap that will quietly determine which funds clear their target IRR over the next two exit cycles sits in the sponsorship layer between the GP and the PortCo CEO.

That is the conversation worth having now, before the buyers price the gap for you.