The AI strategy conversation in private equity has evolved. Firms are no longer debating how boldly to commit. That battle is settled. What’s driving the most interesting discussions among Operating Partners and portfolio ops leadership now is something messier: the how.

How do you actually implement? How do you drive adoption? How do you make the change stick when consultants leave? How do you make sure you’re applying AI where it’s going to create real returns — not just where it’s easiest to deploy? And how do you accurately measure those returns? 

These are the questions we keep hearing from Operating Partners and portfolio ops leaders who are hands-on in their portcos. And notably, even the most experienced operators don’t have clear answers yet. Here is where frontline thinking actually stands.

Implementation is the focus

The firms making real progress on AI have internalized a skill that sounds simple but is actually quite rare: identifying the problem or value creation opportunity first and then determining whether AI is the right tool for the job. For some problems, AI is the best solution. For others, it’s not. The important skill is asking the question every time rather than assuming AI is the answer.

The approaches that are working tend to share a few characteristics.

We are seeing successful operators take a group of functional experts (marketing, operations, finance) trained to automate their specific part of the stack and team them with a cross-functional enabler who is accountable for integrating the whole thing into a coherent system. Rather than a top-down AI mandate, this model starts with people who understand their function deeply and empowers them to redesign it. The result is automation that reflects operational reality, not a theoretical process diagram. 

A second approach that’s gaining traction is to build in parallel rather than transforming from within. Rather than inserting AI at the margins of an existing process, some operators have found it more effective to stand up an AI-native version of a process or function alongside the original, proving the model in isolation, generating results that can’t be dismissed, and then replacing the legacy process once the case is airtight. It requires more upfront work, but the adoption story is considerably easier to tell.

What both approaches have in common is that they treat the existing organization as a source of knowledge rather than an obstacle to route around. The firms stumbling are the ones forcing AI onto their teams without explaining the “why” or the “how.” The firms succeeding are the ones building change with the people who know where the real problems are.

Ambition is not always the right size

There’s a board-level dynamic shaping AI decisions in ways that don’t get discussed openly enough. Boards demanding short-cycle proof of AI ROI creates an incentive for management teams to shrink their ambitions. Instead of rethinking a full commercial stack, someone deploys a call center tool, removes a handful of agents, and calls it done. The execution is complete, the proof is fast, and the larger opportunity, which would have taken 18 months and real cross-functional alignment, never gets started.

The over-focus on productivity leads to a similar problem. Productivity gains often seem like the low-hanging fruit of AI, but hours saved in a workflow don’t hit the P&L unless they’re redeployed into revenue-generating activity, converted into actual headcount reduction, or applied to something that moves EBITDA. Freeing up capacity only creates value when someone decides what to do with it.

The Operating Partners thinking most carefully about ROI are pushing toward different metrics. Revenue per employee is a popular option. In some cases, it is simply growth. The question becomes less “how much did AI save us” and more “how are we creating more value with the same people, or the same value with fewer?” That reframe changes which investments get made and how long leadership is willing to wait for the return.

AI is creating a people problem

The biggest conversation in AI right now for private equity is about people.

There’s a temptation, particularly among operators who’ve worked in a lot of distressed companies, to treat AI transformation the way you’d treat a restructuring: move fast, replace what needs replacing, and deal with the fallout later. This is a reasonable approach for a turnaround situation. In a well-functioning portco, this can create more problems than it solves.

In a business with capable people who understand why certain processes exist, institutional context is a real asset. Those people know why 100 employees are doing something that looks automatable from the outside. Sometimes there’s a regulatory reason or an internal systems hurdle, and you need them to tell you. Trying to automate around that knowledge is how firms end up with expensive downstream problems that looked like savings on a slide.

In one notable case, an operations leader granted employees both the autonomy to build their own AI workflows and a direct financial stake in the savings those workflows generated. The team became invested in outcomes rather than just compliant with mandates. The best performers moved into broader organizational roles, and headcount was eventually reduced. But the people on the team participated in and owned the process, and that changed the cultural math entirely. 

The talent question beneath all of this is thorny. Early-career professionals are often the most aggressive AI adopters — they default to incorporating AI in processes because they don’t have a history of using traditional workflows. That’s a genuine asset. And yet, deep industry knowledge, the ability to ask the right question rather than just run the right query, and pattern recognition built across decades of similar situations are not replicable by a model. The work isn’t choosing between experience and technical fluency. It’s building organizations where both coexist productively.

What this means for how you’re operating now

We work with a bunch of portfolio ops leaders who are deploying AI exceptionally well, and many of them share a few habits worth naming directly.

They treat AI as a portfolio of company-by-company bets rather than a firm-wide program. Each portco gets a use-case prioritization based on where EBITDA levers actually live. This isn’t just a playbook repeated uniformly across the portfolio, though solutions to common challenges can be addressed more quickly based on learnings and infrastructure from other portcos.

They front-load alignment. Getting the deal team, the operating team, and the portfolio company CEO on the same page about AI investment during underwriting is uncomfortable and easy to defer, but it’s also where a substantial amount of value creation gets set up or squandered.

They think carefully about innovation architecture. Whether that’s a dedicated team, a separate pilot environment, or a small group empowered to build without being constrained by compliance and risk processes that weren’t designed for this. The existing line usually can’t transform itself. You have to build alongside it.

They’re honest with their investors about ROI. The small, fast wins are real. The larger opportunity — rethinking how a company creates and captures value — takes longer and requires a different conversation. Having that conversation early is better than having it after the budget has been allocated to point solutions. 

None of this is simple, but all of it is worthwhile. The firms that win the AI race won’t be those with the biggest budgets but those with the clearest operational discipline.