Most boards monitoring AI in their portfolio companies report adoption, capex, and maturity. Those numbers confirm activity. They do not measure value. Three other numbers determine whether AI investment becomes a multiple at exit. They measure whether AI has produced recurring economic value.
EBITDA absorption rate, not AI adoption rate.
The adoption rate is the most reported AI metric in mid-cap portfolios. The percentage of functions using AI tools. The percentage of employees with access to AI copilots. The percentage of workflows where AI has been deployed. These numbers go up reliably, they look like progress.
Adoption is not absorption. The EBITDA absorption rate measures the share of AI-generated productivity that has been captured into recurring EBITDA. Not the gross productivity gain at the individual level. The net gain that has reached the bottom line and stayed there: a portfolio company at 80% adoption can run at 5% absorption.
To calculate it, take the measured productivity gain by function for the last twelve months. Subtract the gain that was reinvested in scope expansion or absorbed by operational slack. The remainder is the gain that has reached EBITDA. Divide by the total measured gain. That ratio is the absorption rate.
The 80% gets reported every quarter. The 5% is invisible.
Run-rate of the asset created, not capex deployed.
The capex metric is the second most reported AI number. The total amount spent on AI initiatives. The number of platforms built. The number of partnerships signed. The boards monitoring AI capex track these numbers carefully because they are easy to count and easy to compare across portfolio companies.
Capex is not value. A portfolio company can deploy 12 million in AI capex and generate 300 000 in annualized run-rate value.
The run-rate of the asset created measures the recurring economic value the capex has produced. Not what was spent. What is now generating recurring economic value, quarter after quarter.
For each AI initiative, identify the recurring economic effect it produces. EBITDA contribution, cost avoidance, revenue uplift, working capital release. Multiply by twelve months. Sum across initiatives. That number is the run-rate of the asset created.
The capex is on every board deck. The run-rate is rarely shown with the same discipline.
Repricing gap, not maturity score.
The maturity score is the third most reported AI metric. The position of the company on a maturity curve. The sophistication of the roadmap. The completeness of the governance framework. The maturity score goes up predictably as initiatives accumulate. That predictability is what keeps it on every board agenda.
Maturity measures readiness. It does not measure impact. On any maturity scale, a level 4 portfolio can have generated less economic value than a level 2 portfolio.
The repricing gap measures the distance between the value implied by the board’s AI narrative and the value supported by actual EBITDA absorption and run-rate creation. It is the unrealized gap already sitting in the portfolio, hidden by the maturity metric.
The calculation requires two valuations. One based on capex deployed and maturity reached. One based on EBITDA absorption rate and run-rate of the asset created. The difference is the repricing gap.
The maturity score grows year over year. The repricing gap widens just as fast.
The 3 numbers behind the multiple.
Each of the 3 metrics is harder to compute than the metric it replaces. That is the reason they are not reported, and the reason they decide the multiple.
Activity metrics like adoption or capex confirm that work was done. The 3 metrics above measure whether that work has produced economic value.
Boards that report only activity indicators arrive at exit with a portfolio that looks valuable in the data room and prices badly in the offer. Boards that report value indicators arrive at exit with a portfolio whose multiple matches the story.
The board pack is not neutral. It trains the company to optimize what it reports.
In the AI transition, the multiple is shaped by what the board chose to measure 4 years earlier.
