For the past two years, the easiest way for a portfolio company to demonstrate an AI strategy was to buy a few licenses, launch some pilots and put “AI” somewhere on the product roadmap. That bar is disappearing quickly.

The more relevant question for private equity is becoming much harder: Is AI actually changing the economics of the business?

That means more than automating meeting notes or helping engineers write code faster. It means asking whether AI can increase win rates, reduce customer acquisition costs, improve retention, accelerate product development, expand margins and, eventually, support a better exit multiple. In other words, the AI conversation is moving from adoption to enterprise value.

Recent research suggests the gap is already becoming visible. McKinsey analyzed 471 PE-backed companies across 31 industries and found that companies broadly embracing AI traded at median revenue multiples roughly 130% higher than businesses using it primarily opportunistically. More importantly, the valuation step-up became meaningful when AI moved beyond productivity and into the actual product and business model. Companies at McKinsey’s highest maturity level traded at a median 31x revenue, versus 20x for companies embedding AI into existing products and services, 14x for operating-model enhancement and 13x for opportunistic adoption.

That distinction gets at the emerging divide inside PE portfolios: between companies that use AI and companies that are becoming AI-first.

Arvind Kumar, Partner and Co-Head of Private Equity Technology at EQT, frames the objective more directly: “We want to make sure our portfolio companies are AI winners, not AI laggards or AI losers.”

The distinction is increasingly relevant because AI does not affect every company equally. Some businesses are genuinely vulnerable to AI-native competitors. Others have structural advantages that become more valuable when paired with AI. The investment question, therefore, is not simply whether a company is exposed to AI disruption. It is whether the company possesses assets that AI can amplify.

For EQT, that begins with what Kumar calls defensive attributes. Proprietary, first-party and domain-specific datasets matter because they are difficult for new entrants to recreate. Complex business-process knowledge matters for the same reason. So does being deeply embedded as a system of record, operating in a regulated market or benefiting from network effects. These characteristics existed before the generative AI boom; the difference is that they can now become inputs into an increasingly powerful offensive strategy.

As Kumar puts it: “It’s really the offense...that may need to be rethought. Not so much the defense. You want to use the offense to strengthen your moats.”

That creates an interesting underwriting framework. A company with weak customer lock-in, little proprietary data and a product easily replicated by a hyperscaler may face structural disruption regardless of how many AI initiatives management launches. But a company sitting on proprietary data, complex workflows and deep customer integration has something AI-native challengers still need to acquire: context.

For sponsors, the opportunity is to buy the latter and then accelerate the offense.

The problem is that most companies are still closer to the beginning of this journey.

BCG describes the progression as Deploy → Reshape → Invent. Deploy means distributing AI tools while leaving the underlying organization largely unchanged. Reshape means redesigning workflows, roles and the operating model so productivity gains actually flow into the P&L. Invent goes further, embedding AI into the customer value proposition and creating new products or revenue streams. BCG’s conclusion is fairly unforgiving: deploying AI can improve individual productivity, but rarely creates meaningful, measurable value on its own.

PwC finds a similar divide. Among PE-backed CEOs it surveyed, only 14% said AI had contributed to both higher revenue and lower costs, while more than half reported no upside at all. The companies producing stronger results tended to have something less exciting than a shiny AI demo: stronger technology operations and better foundations across data, governance, workforce and infrastructure.

This is where the definition of an AI laggard becomes useful. The laggard is not necessarily a company that has ignored AI. In fact, it may have hundreds of employees using AI every day. The more important warning signs are organizational: AI projects with no defined P&L outcome, pilots disconnected from core workflows, weak data infrastructure, management teams treating AI as an IT initiative and product roadmaps where “AI” means adding a chatbot to the existing interface.

Kumar specifically identifies incrementalism as a danger: “If you’re just saying, ‘Hey, we added a chatbot’...that to me is a pitfall.”

Instead, management should be thinking about how AI can create better core products and entirely new, “needle-moving” ones. That changes the discussion from How many people are using Copilot? to What can this company now do economically that it could not do before?

The most compelling examples are beginning to show up in operating results.

Kumar points to WorkWave, an EQT portfolio company providing field-service management software and payments. He describes the business as having been an AI laggard. EQT brought in a product-focused, AI-forward CEO, who in turn hired a new CTO and CPO. Rather than rewriting its entire legacy platform, the company used AI to innovate around the existing technology while developing new products.

The commercial impact came quickly. According to Kumar, WorkWave went from essentially zero to $4 million of new AI-oriented bookings in one year, with expectations to exceed $10 million the following year.

Suggested caption: WorkWave illustrates the sequence PE sponsors are looking for: AI-forward leadership → product transformation → measurable revenue. Source: Arvind Kumar / EQT interview.

The important part of the example is not the use of AI itself. It is that sequence: leadership → organizational change → product innovation → revenue. AI became a value-creation lever because it was connected to something investors can eventually underwrite.

Kumar also points to Billtrust, another EQT portfolio company, where AI is changing the underlying product rather than simply improving internal efficiency. In accounts-receivable automation, the company is using agentic AI to move toward a more autonomous workflow spanning invoicing through collections, with the objective of getting customers paid faster and reducing errors. The distinction matters: the product itself becomes more valuable because of AI.

This is the standard sponsors should increasingly apply. AI initiatives should ultimately show up in familiar KPIs: revenue growth, gross retention, net retention, win rates, CAC, gross margin, EBITDA and revenue per employee. The technology may be new. The scorecard is not.

Kumar reduces the ROI test to four questions:

“Are we going to grow faster? Are we going to retain customers at a higher rate? Are we going to get higher margins? Are we going to have a better exit multiple?”

“If it doesn’t fit into those four buckets,” he argues, “you probably shouldn’t do it.”

This also explains why becoming AI-first is as much a people problem as a technology problem.

Asked where he spends his time evaluating whether a portfolio company can execute, Kumar puts the CEO first, estimating that the role accounts for roughly 70% of the agenda. The signals he looks for are surprisingly practical: Is the CEO intellectually curious? Are they personally using AI tools? Do they understand what is happening in engineering and product? Are AI KPIs tied to growth, retention, margins and exit value?

The inverse is equally revealing. Kumar cites waiting too long to replace a CTO or CPO, excessive loyalty to executives who built the previous version of the company and incremental product changes as some of the biggest leadership pitfalls. EQT has gone as far as replacing CEOs who cannot keep pace, while using specialist teams to accelerate engineering and product transformations.

But AI-forward leadership does not necessarily mean replacing the entire workforce with AI natives. Kumar says his initial assumption was that many existing employees would need to be swapped out. In practice, he found transformation could happen much faster when companies combined an AI-forward CEO with broad access to tools, experimentation, hackathons, internal sharing of successful use cases and, eventually, business goals tied to AI usage.

“It actually has been in weeks instead of months.”

That may be one of the more important lessons for sponsors. The constraint is not always technical talent. Sometimes it is simply permission, incentives and leadership willing to force the organization up the learning curve.

If AI transformation works, the biggest impact may eventually appear in a metric PE has always cared about: how much output the company can generate from a fixed cost base.

That is why focusing solely on headcount reduction misses much of the AI opportunity. The more interesting scenario is a company that can sell into more markets, develop products faster, serve more customers and generate materially more revenue without proportional additions to payroll. In that world, AI becomes both a growth and margin lever.

McKinsey’s research points in this direction. At the highest level of its AI maturity framework, median revenue per employee reached $180,000, versus $118,000 at level three—a 52% increase. Crucially, these companies were not merely automating existing tasks; they were using AI to create new businesses and revenue streams.

That creates a different way to think about AI productivity. The first-order question is how many hours a tool saves. The second-order question is what management does with the capacity it creates. The real PE opportunity appears when that capacity is reinvested into growth while the cost base scales more slowly.

The implication for PE is that AI increasingly belongs in underwriting rather than in a post-close digital transformation workstream.

Diligence should ask two separate questions. First, how defensible is the asset against AI disruption? That means examining proprietary data, customer embeddedness, business-process complexity, network effects, pricing resilience and the likelihood that AI-native entrants or hyperscalers can replicate the value proposition.

Second, how much AI offense is available under PE ownership? Can engineering cycles compress? Can the product become more valuable? Can AI create entirely new SKUs or revenue streams? Can customer acquisition costs fall? Can salespeople cover more accounts? Can retention improve because the company understands usage and churn risk earlier?

That framework creates three broad categories: an AI loser, where structural disruption overwhelms the existing moat; an AI transformation opportunity, where the assets are defensible but execution is behind; and an AI winner, where defensibility and AI-forward execution reinforce each other.

The middle category may be particularly interesting for private equity. Public markets do not always have the patience—or governance control—to replace management, redesign incentive structures, rebuild product roadmaps and endure the messy middle of transformation. PE does.

Kumar sees that disconnect as an opportunity, particularly in software, where companies are increasingly being judged through the lens of AI disruption. His argument is not that SaaS is immune from AI. Quite the opposite: some software companies will be disrupted. But businesses with the right defensive characteristics, paired with aggressive AI-forward management, can use the same technology threatening the category to accelerate both growth and profitability.

“It’s our job as owners to find businesses with those great defensive attributes, and pair them with great executives that can unlock that offense.”

This ultimately brings the argument back to valuation.

The market is unlikely to pay a premium because a portfolio company has thousands of AI licenses or ran 50 successful pilots. It may pay one for a business growing faster, retaining more customers, producing higher margins, shipping better products and demonstrating that those economics are durable.

McKinsey’s data offers an early indication of that distinction. Moving from opportunistic adoption to operating-model enhancement barely changed median revenue multiples—13x to 14x. The larger step came when AI began changing the product itself: median multiples increased to 20x at level three and 31x when companies used AI for new business building.

Correlation is not causation, and AI maturity is almost certainly intertwined with management quality, growth and the attractiveness of the underlying business. But for PE investors, that caveat does not make the signal irrelevant. It makes the diligence question more important.

The exit story of an AI-first portfolio company should therefore be quantitative. How much incremental revenue came from AI-enabled products? What happened to revenue per employee? Did CAC decline? Did retention improve? How much faster can engineering ship? What portion of EBITDA improvement can be traced to redesigned workflows?

Those are harder questions than asking whether a company has an AI strategy. They are also much closer to the questions the next buyer will ask.

The first phase of enterprise AI was about access. The next is about redesign.

For PE firms, the winners will not necessarily be the portfolio companies spending the most on AI. They will be the ones with defensible assets, AI-forward leadership and a measurable path from technology to P&L. Proprietary data without execution can become a stranded asset. AI talent without a moat can become an expensive science project. The combination is where things get interesting.

The diligence question is therefore evolving from How exposed is this company to AI? toward something more useful:

What does this company become if AI works?

The best answer should not simply be “more efficient.” It should be a fundamentally better business.