The AI value-creation thesis has a plumbing problem

PE's 2026 surveys agree AI creates value — for the funds whose portfolio companies can execute. The constraint isn't the model. It's the platform layer underneath it.

Open any investment committee memo written this year and you'll find the same paragraph: AI as a value-creation lever, AI as an exit differentiator, AI as the thing that justifies the multiple. The conviction is real and increasingly backed by capital. What's getting less attention is the sentence that should come right before it — the one about whether the portfolio company's core systems can actually support any of it.

The headline numbers are genuinely encouraging. FTI Consulting's 2026 Private Equity Value Creation Index, drawn from a survey of more than 550 senior PE leaders, found AI accelerating time-to-value across the standard levers, and its companion AI Radar reported that the overwhelming majority of deliberate portfolio AI initiatives are meeting or beating their business cases.1

But read past the headline and a more interesting pattern emerges. High-performing funds aren't adopting AI at meaningfully higher rates than everyone else — they're getting dramatically better results from comparable adoption. FTI found roughly four times as many high performers exceeding their AI business case as their peers.1 Industry summaries of portfolio research point the same direction from the other side: only a minority of PE-backed companies have moved generative AI into production with concrete, measurable results — the rest remain in pilots and testing.2

So the differentiator isn't enthusiasm, budget, or tooling. Everyone has those. The differentiator is the ability to execute — and execution keeps running into the same wall.

Ask why AI initiatives stall inside established mid-market companies and the answer is rarely the model. It's the data the model needs and the systems that hold it. Analyses of PE deployment patterns are blunt about it: legacy system fragmentation constrains the gains, because portfolio reporting and AI tooling alike depend on structured, consistent data — and companies running on aging ERPs, disconnected field service systems, and spreadsheet workarounds can't produce it without manual cleanup.3

This tracks exactly with what happens inside businesses on sunset platforms. When an ERP stopped receiving meaningful feature investment years ago, the organization compensates the only way it can: side spreadsheets, manual reconciliation, a thicket of point tools stitched around the core. Every one of those workarounds is a place where data fragments. By the time an AI initiative arrives, the "single source of truth" is neither single nor true — and the pilot that demoed beautifully on clean sample data fails in production: expensively, and in full view of the board that funded it.

The uncomfortable chain of logic for any hold-period plan: the AI thesis assumes usable data; usable data assumes a coherent platform; and a meaningful share of the mid-market is running platforms with published end-of-life dates and no funded replacement plan.

Here's where the timing gets unforgiving. A platform transition in a mid-market company — selection, contracting, implementation, stabilization — realistically consumes 12 to 24 months. A typical hold period runs five to seven years. If the AI value-creation story is supposed to be demonstrable at exit, with audited results rather than aspirations, then the platform work has to start early in the hold, not after the AI pilots disappoint.

We've written before that the real deadline for any end-of-life platform isn't the vendor's published date — it's the start-engagement-by date, worked backward through the implementation runway. The AI thesis adds a second clock: every quarter spent on a fragmenting legacy platform is a quarter the AI story can't compound. For a portfolio company acquired this year with an exit narrative built on AI-enabled margins, the platform decision isn't an IT matter to sequence later. It's the prerequisite for the equity story.

Three implications worth carrying into the next portfolio review.

First, platform end-of-life exposure belongs in the value-creation plan, not just the risk register. A portfolio company on Dynamics GP, SAP ECC, or an on-premise field service system isn't merely carrying technical risk — it's carrying a structural ceiling on the AI initiatives the deal model assumes.

Second, sequence honestly. The temptation is to run AI pilots in parallel with (or instead of) platform modernization, because pilots are cheap and demos are persuasive. There is an honest case for that ordering — pilots surface which use cases matter before you commit platform dollars — but it holds only if everyone understands the pilots are scouting, not the value creation itself. The 2026 data suggests the funds separating from the pack are the ones that embedded AI into fundamentals rather than bolting it on.1

Third, start the assessment early. Diligence increasingly asks AI-readiness questions; the same questions deserve answers across the existing portfolio. An honest inventory — what platforms, what support horizons, what data actually lives where — reframes the entire AI conversation from aspiration to plan. The first two answers come fast; mapping where the data actually lives takes real effort in a fragmented environment — but it's a fraction of the cost of discovering the answer mid-initiative. Our free EOL Radar covers the support-horizon piece in minutes, and it's built to be forwarded to a portfolio company CFO as-is.

AI didn't change the case for platform modernization. It repriced it.

What used to be a defensive spend — avoid the security gap, avoid the compliance finding — is now the enabling investment for the offense the whole industry is promising its LPs. The funds that internalize that ordering will have an AI story that survives diligence at exit. The rest will have pilots.

If a company in your portfolio is running on a platform with a published sunset date, the two conversations — the EOL clock and the AI thesis — are the same conversation. It's worth having early, with someone who has led these transitions from the operator's chair. Start a conversation, or forward the EOL Radar to the portfolio company that comes to mind first.