AI for Private Equity: Train the Firm or Transform the Portfolio?

When a private equity firm adopts AI, it faces a choice it rarely makes on purpose: deploy it on the firm, or deploy it inside the portfolio. Most default to the firm, because that's the safe, visible, close-to-home option. It's also the smaller of the two. This is a guide to where AI for private equity actually pays off, and how to sequence the two so you build fluency without mistaking it for value.

This is for partners and operating teams who are past "should we use AI" and onto "where do we point it first." Get the order right and you compound an advantage. Get it wrong and you'll have a slightly faster deal desk and a portfolio that looks exactly as it did at entry.

Why firms train themselves first

Almost every firm starts the same way: with itself. The partners want to understand the tools, so they get trained on diligence, modelling and memo-writing. It's a reasonable instinct. It's also where the impact is smallest.

"They come to us to train themselves first... but them being a bit faster in processing board decks is not going to move the needle. What's going to move the needle is the company on the board deck actually using AI."

— Alex Kantjas, 9x co-founder

That's the whole argument in two sentences. Firm-level AI makes the people who watch the business a little faster. Portfolio-level AI changes the business itself. One trims hours off internal work. The other shows up in the number you sell on.

None of this means skip the firm. Training the investment team is the right first step, just not the finish line. It builds real fluency, it's low-risk because a person still signs off on every model and memo, and it earns the credibility to push AI into portfolio companies with a straight face. The mistake isn't starting with the firm. It's stopping there.

What "transform the portfolio" actually means

Point AI at a portfolio company and the goal isn't a faster finance team. It's an operating company where the people who own processes build and automate their own tools, without waiting on a backlog or a systems integrator.

The most vivid version we've seen is a partner at a large firm who stopped delegating this and started building himself.

"He's building a mini app that's going to replace all the board decks... a single app for every portfolio company where you have financials, core initiatives, absolutely everything generated with Claude. And that's him and his co-head doing this, not the juniors."

— Alex Kantjas, 9x co-founder

Read that again, because the detail that matters is who's doing it. Not an outsourced dev shop, not the analysts, but the deal leads themselves, building working software with Claude. When the people closest to the value can build the tools they need, the reporting layer that used to take a quarter and a consultant becomes a weekend. Multiply that across a portfolio's operations, from finance to sales to reporting, and you're changing what the businesses are worth. That's the pattern McKinsey found across hundreds of PE-backed companies: the deeper the AI integration, the higher the revenue multiple.

You can run this across a whole portfolio, not just one company. RM Equity Partners did exactly that: a fund that normally avoids any central, cross-portfolio initiative made an exception for AI, putting 15 portfolio companies through one programme, training over 600 operators, and closing with a three-day hackathon where teams shipped more than 25 automations straight into production.

"This is the biggest technological change in 25 to 50 years. We needed to enable AI transformation not just within one company, but across our entire portfolio."

— Eugen B. Russ, Managing Partner, RM Equity Partners

The aim wasn't to turn every operator into a developer. What most of them gained was judgement: a real feel for what AI can and can't do, so they know what to ask for and when to reach for it. That capability outlasts any single automation.

The trap: automating before you optimise

There's a failure mode worth naming before you roll anything out. Firms get excited, pick a process, and automate it as-is, mess and all. Automating a broken process just makes the mess run faster.

The discipline that separates value from theatre is to fix the process first: question whether the step is needed at all, cut what isn't, simplify what's left, and only then automate. It's unglamorous, and it's the difference between a portfolio company that actually runs leaner and one with a pile of brittle automations nobody trusts. This is also why judgement matters more than tooling: knowing what to automate, and what to leave alone, is the skill.

How to sequence it

The order that works looks like this.

Train the firm first, on real deal work, to build fluency and credibility. Keep a senior person accountable for every output while the team finds its feet.

Then pick one willing portfolio company and point the same capability at its operations. Choose a management team that wants it, not one you have to drag, so the first case is a win you can point to.

Optimise before you automate inside that company, process by process, so what you build is worth keeping. Save the wins as repeatable agentic workflows that other portfolio companies can adopt rather than reinvent.

Then expand from proof. Once one operating company is visibly running leaner, the others stop needing to be convinced. If the work of driving this is landing on one person at the firm or the company, that's the Chief Automation Officer role emerging, and it's worth resourcing properly.

For the wider picture of where AI shows up across a firm and its holdings, our guide to AI in private equity covers the deal desk and the portfolio in more depth.

The bottom line

AI for private equity pays off at two levels, and they're not the same size. Training the firm builds fluency and is the right place to start, but it trims internal hours rather than moving the multiple. Transforming a portfolio company, where operators build and automate their own tools, is where value that reaches exit gets made. Start with the firm to earn credibility, move fast to a willing portfolio company, optimise before you automate, and expand from the first clear win.

If you want your firm and its portfolio companies building this properly, our hands-on Claude training for teams takes investment professionals and operators from their first task to workflows that run the work.

Common questions

Should a PE firm deploy AI on itself or its portfolio first? Itself first, to build fluency and credibility at low risk, but not only itself. The value that reaches exit is created inside portfolio companies, so the firm should move quickly from training itself to transforming a willing operating company.

What does AI value creation in a portfolio company look like? Operators and process owners building and automating their own tools, from finance reconciliation to reporting, without waiting on engineering or external consultants. The business runs leaner and ships faster by the time it's sold.

Why not automate everything straight away? Because automating a broken process just makes the mess faster. Optimise each process first, question and cut steps, then automate what remains. Judgement about what to automate matters more than the tools.

Who should own AI inside a portfolio company? Ideally a dedicated person or small team focused on automating processes for others, rather than a side project. As it scales, that's the Chief Automation Officer role, and it's worth resourcing rather than leaving to whoever has spare time.