Skip to content

Annual data report · first edition

The Portco AI Benchmark 2026

The first benchmark of AI adoption, ownership and P&L impact inside lower-middle-market PE portfolio companies. Built from the AI Pilot-to-P&L Scorecard and a 12-question survey of operating partners and portco leadership. Results publish when the sample passes 50 companies. Respondents get the report first, plus their own company's position in it.

12 questions About 4 minutes Aggregated and anonymised Free

Why this benchmark exists

The fund-level surveys stop at the fund

The published surveys cover large-cap funds and fund-level views: what the sponsor intends, what the operating partner has seen across the portfolio. Nobody publishes the portco-level operating picture for funds with fewer than fifteen companies. Who owns adoption. Whether usage is measured. Whether anything has reached the P&L. That is the picture this benchmark builds.

36%

of PE-backed portfolio companies use AI in day-to-day operations. FTI Consulting, 2026 Private Equity AI Radar

7%

call AI fully integrated across the portfolio. FTI Consulting, 2026 Private Equity AI Radar

9%

of operating partners have seen a demonstrable AI premium in a completed transaction. Accordion / Wakefield, PE AI Adoption Benchmark, May 2026

What we measure

Six dimensions, one portfolio company at a time

Each is a yes-or-no with a number behind it. None of them is a technology question. All six map to the failure modes the Scorecard scores.

01

Adoption owner named

Yes or no, and how senior. Adoption is the whole job, and a committee is not an owner.

02

Weekly usage measured

Yes or no. Without a weekly number the renewal conversation is about spend and nothing else.

03

System of record vs sandbox

Where AI actually runs. A pilot on exported data wins the meeting and dies before the P&L.

04

Licence spend per employee against usage

What the portco pays per head per year, set against whether anyone uses it. Spend without usage is the License Trap.

05

Board reporting cadence and metric

How often AI reaches the board pack, and whether the metric is a P&L number or an activity count.

06

Playbook reuse across portcos

Whether a system working in one portfolio company has been rolled into a second. Reuse is where the fund-level return is.

How to read the results

What will be published, and in what form

No figures are shown until the sample passes 50 portfolio companies. What follows is the shape of the report, so a respondent knows what their answers turn into. The Scorecard's aggregate bands will be published alongside.

Question Reported by Published as
01Is there a single named adoption owner? Survey question 6 and Scorecard mode 06 Share of portcos with a named adoption owner, split by seniority
02Is weekly usage measured? Survey question 7 and Scorecard mode 01 Share measuring weekly usage; median weekly usage rate where reported
03Where does AI run today? Survey question 8 and Scorecard mode 03 Share in the system of record, in a sandbox, on licences only, or nowhere
04What is licence spend per employee, and is it used? Survey question 9 against question 7 Median spend band; spend band cross-tabulated with usage measured
05How does AI reach the board? Survey questions 5 and 10 Share reporting monthly with a P&L metric; share with a number in the value-creation plan
06Has anything reached the P&L, and been reused? Survey question 11; operating partners responding for two or more portcos bps of EBITDA attributed, median; share of funds with one system running in two or more portcos

Every cell is published as a band or a median, never as a company. Cells with fewer than five responses are withheld.

The survey

Twelve questions. One portfolio company.

Answer for one portco, not the fund. If you cover several, submit once per company. Answers are saved on this device until you submit.

Progress

0 / 12

12 questions left

Question 1 of 12: your role
01

Your role

Question 2 of 12: fund size
02

Fund size

Question 3 of 12: number of portfolio companies
03

Number of portfolio companies

Question 4 of 12: this portfolio company's revenue
04

This portco's revenue

Question 5 of 12: is AI named in the value-creation plan
05

Is AI named in the value-creation plan?

Question 6 of 12: is there a single named adoption owner
06

Is there a single named adoption owner?

One person whose review or weekly number depends on the workforce using the system.

Question 7 of 12: is weekly usage measured
07

Is weekly usage measured?

Question 8 of 12: where AI runs today
08

Where does AI run today?

Question 9 of 12: licence spend per employee per year
09

AI licence spend per employee per year

Question 10 of 12: board reporting
10

How is AI reported to the board?

Question 11 of 12: has any AI result been attributed in the P&L
11

Has any AI result been attributed in the P&L?

Question 12 of 12: work email and company
12

Where to send the report

A work email is required so each company is counted once. Company is optional and is never published.

Responses are aggregated and anonymised. You receive the report and your company's position first. Nothing is shared with a third party.

Methodology

How the sample is built and protected

Sample

Portfolio companies of private equity funds, with an emphasis on the lower middle market: $20M to $500M in revenue and funds with fewer than fifteen companies. One response counts for one portfolio company. An operating partner may respond for several companies, each counted once. Responses from vendors, advisers and anyone without a work email are excluded.

Anonymisation

Answers are aggregated into bands before anything is published. No company, fund or respondent is named. Any cell with fewer than five responses is withheld. The work email is used to count each company once and to send the report, and for nothing else.

Scorecard data

The AI Pilot-to-P&L Scorecard stores a result only when the respondent asks for the emailed breakdown. Those six mode scores and the band are matched to a survey response by work-email domain, then the identifiers are dropped. The Scorecard contributes the failure-mode picture; the survey contributes ownership, spend, reporting and attribution.

Publication

The first edition is published when the sample passes 50 portfolio companies. After that, annually each September, with the survey open all year. Respondents receive the report before publication and are never named.

Questions before you respond

Who can respond to the Portco AI Benchmark?
Operating partners, portfolio company CEOs, COOs, CFOs and CHROs, and board members of PE-backed companies. Each response covers one portfolio company. An operating partner who covers several companies can respond once for each. Responses from vendors, advisers and anyone without a work email are excluded from the sample.
Is the survey anonymous?
The published benchmark is. A work email is required so that each company is counted once and so that the report and your company's position can be sent to you. Results are aggregated into bands before anything is published, no company, fund or respondent is named, and any cell with fewer than five responses is withheld. Nothing is shared with a third party.
When is the benchmark published?
The first edition is published when the sample passes 50 portfolio companies. After that it is published annually each September, with the survey open all year. No figures are shown on this page until the first threshold is met, because a benchmark with a small sample is a guess with a chart on it.
What do respondents receive?
The full report before it is published, plus a one-page note showing where your portfolio company sits against the sample on each of the six dimensions. If you have also run the AI Pilot-to-P&L Scorecard, the note includes your band against the aggregate bands. There is no cost and no sequence you cannot leave in one click.

Put one portfolio company in the sample.

Twelve questions, about four minutes. Or score the company first and bring the band with you.

Take the survey Book a call