The PE Portfolio AI Maturity Assessment

1. Why You Need One Score Across the Portfolio

A fund with twenty portfolio companies has twenty different answers to the question "how far along is this company with AI," and usually no way to compare them. One CEO is enthusiastic and behind. Another is quiet and ahead. The operating partner is guessing.

A maturity assessment replaces the guess with one score on one scale. Not because the number is precious, but because comparison is impossible without it. You cannot decide which company gets the next dollar of AI investment, or which one is the case study and which is the rescue, until they are measured the same way.

This is the portfolio-wide cousin of the single-company readiness diagnostic used at a deal. The diagnostic asks "is this one company ready." The assessment asks "across everything we own, where do we push first."

2. The Cost of Not Knowing

Not measuring has a price, and it is now showing up in valuations. BCG's 2026 work on PE and digital maturity found that when digital maturity lags, 40 percent of investors have experienced a valuation haircut of 5 percent or more, and only 8 percent said it had no impact on valuation. Maturity you did not manage becomes a discount someone else prices at exit.

Read those last two numbers together. Most firms neither score maturity nor connect it to the exit story, which means the haircut in the second tile lands on firms that never saw it coming. The assessment is how you see it coming.

3. What AI Maturity Actually Measures

AI maturity is not how much AI software a company has bought. A company can hold every license and be immature, and a company with a few well-run use cases can be ahead. Maturity measures whether AI is actually changing how the company works and shows up in results.

That distinction matters because vendors sell licenses and licenses are easy to count. The assessment deliberately scores the harder things: is the data usable, are use cases in production rather than pilots, do people use them, is someone accountable, and is it governed. Those are what separate a company that talks about AI from one that runs on it.

Keep the framing concrete. Every dimension is scored on evidence (something in production, a measured result, a named owner) not on intent or enthusiasm. Enthusiasm is the most overrated input in this assessment.

4. The Five Dimensions

Score each company on five dimensions. They are deliberately few, because a scorecard nobody can fill in is worse than a rough one everybody can.

1. Data foundation. Is the data that AI needs captured, clean, and accessible? This is the dimension that gates all the others, and the one companies most overrate about themselves.

2. Use cases in production. Is anything actually live and creating value, or is it all pilots and demos? Production is the line between maturity and theater.

3. Adoption. Do the people who are meant to use it actually use it, every day, the new way? A live tool nobody uses scores zero here, correctly.

4. Leadership and strategy. Is there a named owner, a plan tied to the value creation thesis, and a budget? AI without an owner drifts.

5. Governance and security. Are there controls on data, vendors, and risk? Maturity without governance is fragility, and it is a real diligence finding at exit.

5. The Four Maturity Levels

Roll the five dimensions into one level per company. Four levels are enough to act on and few enough to agree on.

The honest starting picture for most portfolios is sobering and useful: a long tail at Levels 1 and 2, which lines up with the finding that only about 15 percent of portfolio companies claim very mature capabilities. That is not a problem with the portfolio. It is the opportunity, named.

6. How to Run the Assessment

Keep it light enough to actually repeat. A heavy assessment gets run once, becomes a slide, and is never updated. A light one becomes a habit.

Score each company on the five dimensions, on evidence, through a short structured conversation with the company plus a look at the systems. Use the same rubric and ideally the same assessor across the portfolio, because consistency is the entire point. A score is only comparable if it was measured the same way next door.

The output is a simple grid: companies down the side, the five dimensions across, a level for each, and an overall level per company. That grid, kept current, is the single most useful artifact an operating partner has for AI value creation. Our AI Readiness Diagnostic is a fast way to score a single company on these lines.

7. From Scores to Priorities

The score is the input. The decision is where to push first, and the grid makes it obvious in a way intuition does not.

Two things drive priority: the gap and the prize. The gap is how far the company is from where it could be (low maturity, high headroom). The prize is how much value the company represents (size, exit timing, sector exposure). A large, low-maturity company near its exit window is the obvious first call. A small company that is already Level 3 is a case study to copy, not a place to spend.

This is also where the company-by-company instinct fails. Treating each company as its own project ignores that the same playbook, the same vendor, the same data pattern usually transfers across several. Prioritize by where a reusable move helps the most companies, a theme of the deployment playbook.

8. Turning the Score Into a Plan

A level is a diagnosis, not a treatment. Each level implies a different next move, and naming it stops the assessment from being an interesting chart that changes nothing.

A Level 1 company needs the basics: a named owner, the data foundation, and one use case taken to production. A Level 2 company needs to get something live and adopted, not start more pilots. A Level 3 company needs to broaden adoption and tie the gains to EBITDA. A Level 4 company needs to keep its lead and turn it into the exit story. The move follows from the level.

Training is part of every move, and it is cheaper than firms assume. Andrew Ng's long-standing guidance is that executives need only a few hours of AI literacy, division leaders carrying out projects somewhat more, and the people building need real depth. Most of the portfolio needs the first kind, which is hours, not months.

9. Re-Scoring and the Exit Link

Score once and you have a snapshot. Score every quarter or two and you have a trajectory, which is what actually matters. A company moving from Level 1 to Level 3 over a hold is a value-creation story. A static score is a missed one.

The trajectory is also the raw material for the exit narrative, and almost nobody builds it. Recall that only about 11 percent of firms link their digital progress to the exit story. The maturity assessment, tracked over the hold, is precisely the evidence a buyer will pay for, because it shows the capability is real and improving, not a last-minute paint job. Building that into the sale is the subject of the exit value-creation guide.

Keep the re-score consistent with the entry score, so the trajectory is honest. A trajectory built on a moving rubric persuades no one, least of all a diligence team.

10. Where to Start

Score the portfolio once, quickly, on the five dimensions. Even a rough first pass tells you which companies are the priorities and which are the case studies, and it is almost always different from the intuition in the room.

Then commit to re-scoring on a schedule, with the same rubric, so you build a trajectory rather than a snapshot. The discipline of measuring the same way every time is worth more than any single clever score.

If you want a consistent rubric and an outside assessor across the portfolio, a Readiness Sprint scores your companies on one scale and turns the grid into a prioritized plan. Start a single company now with our AI Readiness Diagnostic.