8 Questions PE Operating Partners Should Ask Every Portfolio Company About AI

An 880-person company we worked with had rolled out Microsoft Copilot to every employee, along with the vendor's 8-hour video training course. When we asked how many people had finished it, the answer was 50. All 50 said it was boring.

That company would have told its board it was adopting AI. Every board deck now has an AI slide, and most of them look the same: tools bought, licenses rolled out, spend going up.

The problem is what's going on behind the slides. According to Deloitte, companies are putting 93% of their AI budgets into technology and only 7% into the people expected to work with it.

The result is AI on paper: licenses paid for, very little change in how work gets done, and value left on the table during the hold period.

It also leaves a weak story for exit. When a buyer's due diligence team starts digging into AI, they'll ask the same questions you should be asking now.

A single conversation, and the right set of questions, can tell you where a company really stands.

Here are 8 questions to ask every portfolio company about AI, written for operating partners or whoever owns value creation at your firm. Each one comes with the red-flag and strong answers to listen for.

If you're a portfolio company CEO, read this as a preview of what your board is about to ask.

Questions that test real AI adoption across the team

Start with the people. These three questions show whether the team works with AI day to day, or whether the tools are sitting unused.

1. How many people with AI access used it for real work this week?

Access and adoption are two different numbers, and most companies only track the first.

In Source Capital's cohort surveys, 56% of PE professionals with AI access hadn't actually worked with it in their day-to-day jobs. Wider data tells a similar story: Recon Analytics found that only about a third of U.S. workers with paid AI subscriptions and Copilot available at work have made it their main AI tool.

Every unused seat is spend with no return, and that adds up quickly.

• Red flag: "We rolled it out to everyone." That's an answer about purchasing.
• Strong answer: Weekly active use by team, tied to specific work. Something like: "Nine of our 11 finance people work with AI every week, mostly on month-end close."

2. What training did your team get, and who finished it?

Handing people a tool and a course is where most rollouts stop. Source Capital's survey data found 97% of PE professionals had no prior AI training, even though 85% were already using ChatGPT in some form. People experiment on their own, with no shared approach and no one checking the results.

Untrained teams get shallow results, which makes the license spend hard to defend.

• Red flag: A vendor video library, with no idea who completed it. Think back to the 8-hour Copilot course.
• Strong answer: Hands-on sessions built around the team's own work, where people build tools they keep using afterward. When Source Capital trained 59 executives from 28 portfolio companies, participants left with working systems already running in their businesses.

3. How much of your AI budget goes to people?

This is where the 93/7 split shows up inside a single company. Microsoft's 2026 Work Trend Index found that organizational factors, such as culture, manager support, and how people are developed and rewarded, have more than twice the impact on AI success as individual mindset and habits, such as how keen someone is to experiment.

Tool-only budgets buy potential, which only turns into value when people change how they work.

• Red flag: Every AI dollar is a license, a platform, or a vendor contract.
• Strong answer: A dedicated budget for training, change management, and internal champions who help colleagues day to day. That's the thinking behind The ADOPT Method™: training teaches skills, and ADOPT transforms behavior.

Questions about AI governance and ownership

Next, find out whether anyone is steering. These two questions surface data risk and ownership gaps before a buyer finds them.

4. Is there an AI policy and a company-approved login?

Without a policy or an enterprise account, people sign up for whatever tools they like on personal accounts. Customer data, financials, and employee information end up in places the company can't see or control. None of the firms in Source Capital's most recent PE cohort had a formal AI policy, and 68% had nothing in place at all.

This is exactly the kind of risk a buyer's due diligence team will flag at exit.

• Red flag: "People use whatever works for them." Or a policy exists on paper, but nobody has been shown how to follow it.
• Strong answer: A written policy, clear rules on which data can go where, and one enterprise login the whole team works in. For companies starting from zero, the AI Policy Generator and AI Governance Framework Builder are quick first steps.

5. Who owns AI outcomes, and who checks the output?

AI can draft the report, screen the deal, or answer the customer. A person still has to own the result. As we put it at AI Operator: AI is not responsible for its output. You are.

Ownership matters at two levels: a named leader accountable for AI results, and clear human review before AI output reaches a customer, a deal, or the board. Unowned AI stalls, and unchecked AI creates errors that surface at the worst possible moment.

• Red flag: "IT is handling it," or "the vendor takes care of that."
• Strong answer: A named executive owns AI results, and the CEO is visibly involved. In founder-led companies, that's often the founder. At Source Capital, the managing partners trained alongside their team, and CEOs and presidents made up a third of the portfolio company cohort.

Questions that measure AI ROI

Finally, follow the money. These three questions separate spend from results, which is what your partners and a future buyer will care about most.

6. Which processes has AI changed, and how long did they take before?

A company getting value from AI can name the work that changed and put numbers on it. Vague answers usually mean there's no baseline, and without a baseline there's no way to show a return.

Hours saved on repeat work flow into margin, and margin flows into exit value.

• Red flag: "It's helping across the board."
• Strong answer: Specific before-and-after numbers from specific people. In Source Capital's portfolio program, Greg Rosenstein, CFO of manufacturer Mechanical Concepts, built AI agents covering controller, FP&A, and reporting work, worth an estimated $30,000+ a year. Martha Parker, CFO of CBS Enterprises, worked with AI on percentage-of-completion accounting across hundreds of open construction jobs.

7. What's your return on token?

Some companies point to their AI spend as proof of progress: "Everyone's on Claude, look how many tokens we're going through." Tokens are the units AI tools bill by, so rising token spend shows activity. The question to ask back is simple: what did you get for it?

Chasing usage for its own sake even has a name now, token maxxing, and it tends to drive up cost faster than results. Spend without a measured return is just cost, and cost shrinks margin.

• Red flag: A spend figure, followed by silence when you ask which processes it's going into.
• Strong answer: Spend mapped to specific workflows, each with a baseline and a measured result. If the answer is shaky, this guide to measuring AI ROI lays out how to set baselines and review results at week six.

If question 7 is hard to answer across your portfolio, the 2-minute AI Audit is a quick place to start. It shows where AI is delivering in a company and where the gaps are, with instant results.

8. What has your team stopped doing because of AI?

Real adoption shows up as work that disappears. It's the hardest answer on this list to fake, and a team that works with AI every day can answer it in seconds.

Work that disappears frees people for higher-value work, and that's where the gains start to compound.

• Red flag: A long pause, or a list of tools.
• Strong answer: A specific task or habit that's gone. One of the best answers we've heard came from inside a PE firm, where the team used to spend Thursday and Friday preparing for the Monday deal committee. Now everything is ready when they walk in. In their words: "We don't prepare for the Monday call anymore.”

How to use these questions across your portfolio

The questions work best asked at the right moments, compared across companies, and framed as a conversation.

Ask at three points in the hold period

In the 100-day plan, the answers give you a baseline. At quarterly board meetings, they show whether anything is moving.

Ask again 12 to 18 months before exit. That gives each company time to fix red flags before a buyer's due diligence team finds them.

Score the answers and compare across companies

Rate each answer red, amber, or green. One company's scorecard is useful, but scorecards from every company in the portfolio show who's ahead, who's stuck, and which gaps keep coming up.

The companies that score well become proof points for the rest. That's how Source Capital's program spread: results inside the firm led to a program for 28 portfolio companies, which then drew in 24 peer firms. To decide where to act first, see how to find the highest-value AI opportunities across your portfolio.

Keep it a conversation

Many portfolio company CEOs founded their business and still own part of it, so they benefit from a higher exit price too. Framed that way, these questions open a shared conversation about value.

Framed as an audit, they put people on the defensive, and you get polished answers instead of honest ones.

Turn the questions on your own firm

The same 8 questions work inside a PE firm, as the Monday deal committee example in question 8 shows.

A firm that can answer these questions well asks them of portfolio companies with more credibility. If you're weighing where to focus first, this breakdown of firm-level or portfolio-level AI can help.

Turning red flags into real Human + AI adoption

Most portfolio companies will score a few reds, and that's normal. Nearly every company is early, and the gaps these questions expose, like training, ownership, and measurement, are all fixable within a hold period. Source Capital built AI capability across its own team and 28 portfolio companies in seven months.

The fix starts with people. When humans and AI work together, with clear ownership and a way to measure results, the answers to all 8 questions can change quickly.

Want help turning your portfolio's red flags into results? Book a strategy call and we'll walk through where to start.

Frequently asked questions about AI in portfolio companies

What should operating partners ask portfolio companies about AI?

Focus on three areas: whether people actually work with AI, whether anyone owns and governs it, and whether it produces a measurable return. Questions about weekly active use, training, AI policy, ownership, and return on token cover all three in a single conversation.

How do you measure AI adoption in a portfolio company?

Track weekly active use by team, tied to specific workflows, alongside before-and-after numbers for the processes AI has changed. License counts and token spend measure activity, so they're a starting point at best.

What is return on token?

Return on token is the business value a company gets for its AI token spend. A strong answer maps spend to specific workflows, each with a baseline and a measured result, such as hours saved or faster cycle times.

How can PE firms tell real AI adoption from AI on paper?

Ask what the team has stopped doing because of AI. Companies with real adoption answer quickly with specific tasks, named owners, and numbers. Companies with AI on paper tend to answer with tool names and spend figures.