AI adoption in portfolio companies: what the surveys say

AI adoption in portfolio companies is measured, and the measurements agree. Funds expect value from AI within a short window, budgets are rising to pay for it, and the share of portfolio companies with AI running at scale is small. Three surveys published between January 2025 and May 2026 give the numbers, and this page sets them side by side so an operating partner can cite them and act on them.

Each figure below is quoted as the publisher stated it. Where a survey reports what respondents said about themselves, the page says so, because a fund reporting that its AI initiatives met their business case is a self-report.

What did the three surveys find?

The three surveys asked different people different questions, and the table keeps them separate.

Survey	Who was asked	What it found
Mubadala and MGX with Bain & Company, Alpha Intelligence: The Investment Fund of the Future, press release 22 January 2025	Private equity firms representing more than $3.2 trillion in assets under management	93 percent of investment funds anticipate AI-driven value within three years; 2 percent expect substantial returns this year; 18 percent of portfolio companies are already seeing concrete value from operational AI use cases; some funds have AI teams of 40 to 70 specialists
BCG, Private Equity's Future: Digital First and AI Powered, 7 January 2026	100 senior private equity investors	More than 90 percent plan to expand portfolio-level digital budgets over the next three years; 90 percent cite competing priorities as the top blocker to digital transformation and 76 percent unclear return on investment; 15 percent of portfolio companies claim the top IT maturity rating and nearly 75 percent report moderate maturity
FTI Consulting, 2026 Private Equity AI Radar, 19 May 2026	200 fund and operating leaders	36 percent using AI across use cases and 7 percent at enterprise scale; 95 percent of funds report AI initiatives meeting or exceeding their original business case criteria; revenue acceleration the top priority at 41 percent; talent the primary constraint on scaling adoption at 35 percent

Read across the rows, the story is consistent. In January 2025, nearly all funds expected value within three years and fewer than one in five portfolio companies had it. A year later, budgets were rising and the blockers were priorities and unclear returns. By May 2026, a third of respondents used AI across use cases and 7 percent had it at enterprise scale, with talent the constraint.

Reveneau builds AI capability for portfolio companies as scoped engineering work in which the code is written by AI and every change is checked against an evaluation suite derived from the specification, so the fund gets the capability with a stability measure attached rather than a pilot with a slide; the surveys above are about the demand for that work, and the rest of this page is about how a fund acts on them. The pillar guide to technology after the deal sets AI adoption among the capability decisions of the second month after closing.

Why is the gap between expectation and adoption so wide?

The gap is wide because the surveys are measuring two different things: what a fund expects, and what a portfolio company has built. The first is a plan and the second is engineering work, and the surveys themselves name the reasons the work lags.

BCG's respondents named competing priorities (90 percent) and unclear return on investment (76 percent). FTI's respondents named talent (35 percent). Mubadala and MGX with Bain reported that some funds had built AI teams of 40 to 70 specialists, which describes the fund-level response to a talent constraint and says nothing about the portfolio companies, which do not have 40 specialists each.

There is a third reason the surveys do not measure directly, and DORA does. Google Cloud's announcement of the 2024 DORA report, on 22 October 2024, said more than 75 percent of respondents relied on AI for at least one daily professional responsibility, and that as AI adoption increased it was accompanied by an estimated 1.5 percent decrease in delivery throughput and a 7.2 percent reduction in delivery stability, with 39 percent reporting little to no trust in AI-generated code. The 2025 report, announced on 23 September 2025 from nearly 5,000 technology professionals, found 90 percent using AI at work, a positive relationship between AI adoption and throughput, a continued negative relationship with stability, and 30 percent reporting little or no trust.

A portfolio company that adopts AI in its engineering without a way to check the output gets the stability cost the surveys describe. That is a reason adoption stalls at pilot, and it is an engineering problem with an engineering answer, which the eval-driven development guide describes and the AI-generated code to production guide argues in full.

What can a fund do with these numbers?

A fund can do three things with them: set the expectation at the portfolio company in the surveys' own terms, put AI adoption on the engineering plan with a measure, and report it in the same format as everything else.

1. Set the expectation with the numbers. A board that has read that 7 percent of FTI's respondents have AI at enterprise scale asks a different question of its portfolio company than a board that has read a vendor's deck. The question becomes "what is the first use case, what does it cost in engineer-weeks, and how will we know it is working", which is the question the technology workstream in a 100-day plan is built to answer.
2. Put it on the plan as engineering work. An AI capability has a specification, a build, an evaluation and a stability measure, like any other feature. It goes into the same backlog as the diligence findings and the value creation builds, with the same owner, date, cost and measure of done, and it competes for the same capacity. That is how it stops being a competing priority and becomes a scheduled one.
3. Report it with the four delivery measures. DORA's change fail rate and failed deployment recovery time are the measures that show whether AI adoption in engineering is costing stability, and they are already in the quarterly format described in reporting engineering progress to the fund. A portfolio company whose fail rate rises after adopting AI coding tools has the DORA pattern, and the board can see it.

Which AI adoption is the survey talking about?

The surveys mix two kinds of AI adoption, and a fund should separate them because they are different engineering work.

AI in the product is a capability customers use: a feature that answers questions, classifies documents, forecasts demand, or runs an agent on the customer's behalf. It is built like any product feature, with a specification and an evaluation suite, and its return shows in revenue. FTI's finding that revenue acceleration is the top priority at 41 percent is about this kind.

AI in the operation is a capability the company uses on itself: AI-assisted engineering, AI in support, AI in finance. Its return shows in cost and speed, and the DORA figures above are about the engineering version of it. Mubadala and MGX with Bain's 18 percent figure names "operational AI use cases", which is this kind.

The two need different measures. The product kind is measured by adoption and revenue. The operational kind is measured by the cost and speed of the function it changed, and for engineering that is the four delivery measures. A board that asks "how is AI adoption going" should get two answers.

What should the operating partner ask for?

The operating partner should ask for four things, each of which is a document rather than a status:

• A list of the AI use cases in flight at the portfolio company, each labelled product or operational, with an owner, a cost in engineer-weeks and a measure.
• For each product use case, the specification and the evaluation suite that checks the model's output against it, and the pass rate on the last run.
• For each operational use case in engineering, the four delivery measures before and after adoption, so the stability effect DORA describes is visible.
• The talent plan: who builds this, whether they exist at the company today, and if not, which of the options in interim, fractional, or as-a-service CTO supplies them.

FTI's self-reported 95 percent business-case figure is worth holding against those documents. A fund whose AI initiatives met their business case can show the specification, the evaluation and the measure; a fund that cannot show them has a self-report.

What the surveys do not say

Three things the surveys do not say, so that nobody cites this page for them.

They do not say what share of portfolio companies will have AI at scale next year; each is a snapshot of its respondents on its date. They do not say what AI adoption returns in a portfolio company, because the return figures reported (BCG's, FTI's) are what respondents said about their own initiatives. And they do not measure whether AI-written code in a portfolio company is safe to ship; DORA measures its effect on delivery stability, and the security question belongs to the AI-generated code to production guide, which holds the security figures.

Bain & Company's press release of 23 February 2026 says typical deals now need 10 to 12 percent annual EBITDA growth to return 2.5x. Whether AI in the product or AI in the operation produces any of that growth at a given portfolio company is a question the company's own measures answer, and none of the three surveys can answer it for them. If the fund would rather the capability was built by a team that takes the outcome, building with investors and their portfolio companies describes how we do that.