If software investing has delivered anything to private equity investors so far in 2026, it is the disquieting certainty that the world has changed—probably forever.

Revenue growth that was running around 20% annually is now trending at half that. Net revenue retention (NRR) has dropped about 8 points since 2021 (see Figure 1). Dealmaking is at a crawl. Aging portfolios are suddenly a thing in tech investing. And looming over everything else is the specter of artificial intelligence, which, at best, threatens the once-unassailable SaaS value proposition and, at worst, raises fears that some software use cases are veering toward obsolescence.

Figure 1

Perhaps most ominous is the building evidence that software is losing the structural advantages that for a decade produced the buyout industry’s shiniest returns. The software assets that are transacting in the private markets continue to attract high prices, largely because sponsors have focused their efforts on selling their A-plus companies. Yet deals completed after the Covid-19 pandemic, including more seasoned investments from 2020 to 2022, are, to date, generating returns below pre-Covid averages (see Figure 2). While many of these investments are not yet fully realized, and performance could improve over time, that will be a challenge given the high prices sponsors paid for assets and the sector’s slowing growth.

Figure 2

What comes next?

While predictions aren’t worth a lot amid such uncertainty, it’s safe to say that we’re living through the Wild West atmosphere that attends any period of technological disruption. What we know from previous transformations (the dot-com revolution and the shift to cloud computing come to mind) is that winning business models will eventually emerge from the chaos, and the winners will all have something in common: a deep understanding of customers’ needs and a clear vision for how the emerging technologies can solve those real-world problems with real-world economics.

At the moment, bold proclamations about AI’s future impact trade off almost daily with reports that few companies are seeing measurable returns from their AI investments. Bid-ask spreads for some software categories might narrow somewhat as AI use cases and business models begin to prove out. But valuations in most categories are essentially up for grabs right now. The biggest issue is that the AI disruption is moving faster than the dot-com or cloud transitions. Most companies are still in discovery mode, searching for ways to meaningfully deploy AI to satisfy customer needs (or what they might need soon).

Yet, against this backdrop, patterns are emerging that suggest a number of practical actions private equity investors can take right now to keep pressing forward. We already know enough about AI’s disruptive power, in fact, to start making no-regrets shifts in every phase of the PE value proposition. Proactive general partners (GPs) are taking steps to:

• pragmatically reorient due diligence to more reliably underwrite AI risks and opportunities;
• refocus value-creation playbooks to zero in on how AI technology can transform offerings through a deeper understanding of customer workflows; and
• demonstrate bankable progress by marshaling data and metrics that can provide the next owner with compelling AI proof points, not just rosy narratives or random prototypes.

The new due diligence imperative

In the decade or so since the subscription model took over software, boosting value has been all about maximizing predictable, recurring revenue. Growth came steadily as vendors added seats, charged more for new features, and cross-sold new products. With no physical product to deliver, each new dollar of revenue fell in large part to the bottom line. And once a SaaS company was established with a customer, the relationship deepened into a competitive moat as software became embedded, users developed habits around it, and data accumulated within the platform.

AI is changing all that. These new technologies don’t just threaten to leapfrog some software solutions, allowing AI-native companies to swim across those moats (think customer service or software development). AI products also have different economics. The massive calculations powering AI add substantial cost to every query, which requires revenue models based not on seats but on actual usage and results. That shrinks margins and, today at least, makes growth projections uncertain.

The result is that traditional SaaS signals are increasingly misleading in due diligence. Revenue multiples, annual recurring revenue (ARR), and NRR were reliable proxies for value when software moats were more durable and margins were structurally high. But the correlation between growth rate and valuation is flattening, and traditional underwriting isn’t fully capturing the elements that will help predict which assets are equipped to thrive in a landscape reshaped by AI.

Diligence in this new environment has to start with an assessment of two things: How much can AI impact the user workflows the software supports, and how much risk is there that AI could displace the software altogether within those workflows? Stress testing requires specificity, since product moats, workflow moats, and data moats each come under different degrees of pressure from AI.

Strong diligence will also capture upside: How AI-ready is the company itself? Is it rapidly deploying AI internally to improve efficiency? Is it gaining traction with AI on the product side, either by adding features or by launching new products that customers value? The key here is evidence: The market is already bifurcating between companies that can demonstrate measurable AI traction and those that are still spinning a narrative without numbers behind it. In diligence, the key question is no longer “Do they have an AI strategy?” but “Can they show me the proof points?”

AI transformation during ownership

Using AI internally to improve efficiency and transform workflows is rapidly becoming table stakes for any company in your portfolio, software or otherwise. The opportunity to truly inflect performance still involves some experimentation and faith in the technology, but standing still isn’t an option.

The more complex question is how to reshape the product roadmap to generate AI-based revenue—and how to scale meaningful innovation at speed. That can be a tall order for an incumbent software company used to optimizing for a traditional seat-based platform, especially given all the daily requirements of managing the core business.

The companies seeing the most success tend to have a light-bulb moment when they recognize that simply helping humans do tasks faster with incremental product enhancements is increasingly missing the point in an AI world. If an agentic system, for instance, can actually manage a workflow end to end, the goal should be enabling a step change in measurable outcomes, not just user efficiency.

That means turning the traditional product development approach on its head. Instead of focusing solely on how customers use discrete products within workflows, the most innovative companies are looking to connect the dots across the entire process. They are going deep on the customer’s broad objective and how their solution can be rebuilt to achieve that outcome—sometimes freeing up workers for more productive pursuits, sometimes replacing them altogether.

For Zendesk, the $2-billion-revenue customer service platform, the tipping point arrived a couple of years ago when it became obvious that most inquiries being handled by customer service reps could be resolved using AI. Adding AI features to the existing platform might speed up human interaction to a limited degree. But once Zendesk asked itself how much of the customer service work could be done autonomously and reliably, the future was clear: It needed to completely rebuild an AI-based platform to sort queries, surface enterprise-wide knowledge, and work across the company to reach resolutions, all at speeds and accuracy never before possible. 

Assembling the AI capabilities fast enough posed a major challenge for a company that still needed to maintain growth and service a large installed customer base. The answer: a string of targeted acquisitions that would import expertise in everything from AI-powered automation and quality management to enterprise search and analytics.

Dialing up the clock speed also required rapidly refocusing the workforce so that more than half the company was working on AI, up from less than 10% just 15 months earlier. That involved a nervy program to prune initiatives, collapse overlapping work, kill bets that were no longer strategic, and otherwise make the painful trade-offs needed to free up capacity. The results have been impressive: In roughly 18 months, Zendesk reported $200 million in ARR from AI, with 20,000 AI customers and nearly 800 million AI interactions in 2025. 

The lesson here is that AI is not a game of incrementalism. Capturing the opportunity starts with zero-basing product assumptions and reimagining what’s possible. The answer won’t always be a complete rebuild. But you need to decide how you can take customer outcomes to a new level by matching AI technology to a deeper understanding of the relevant workflows. And you’ll need to assess what that will require in terms of adding talent and making organizational changes to support rapid execution.

Proving it

It’s no surprise that GPs and their portfolio company management teams are expending massive bandwidth to develop new AI features and products. But they’re probably not spending enough energy to develop the means to track progress and measure impact with concrete data. That helps explain the wide gap in value expectations we’re often seeing between buyers and sellers.

Because the traditional SaaS KPI stack (ARR, NRR, and gross margin) was built for a different value proposition (a world of near-zero marginal cost, seat-driven expansion, and predictable retention), it fails to capture AI impact reliably. ARR, for instance, is a great measure of the predictable revenue derived from subscription seats. But AI is priced on actual usage and outcomes, which can be bursty and unreliable.

Some companies try to isolate AI’s top-line impact by breaking out a broad category called “AI-associated revenue.” But that doesn’t tell the story any more than the label “cloud-associated revenue” did a dozen years ago. Understanding the true performance of these products requires separately tracking at least three distinct revenue buckets: traditional AI and machine learning (predictive models, risk scores), AI add-ons (copilots, assistant features), and agentic products (workflow automation, autonomous execution). Each has a distinct margin profile, growth trajectory, and competitive dynamics.

The cost side, too, is very different. AI products have significant variable costs per use that don’t exist for traditional SaaS solutions. A full understanding of what you’re spending requires tracking hosting and infrastructure costs, third-party model costs, and the fully loaded cost of each employee devoted in full or in part to AI-related R&D. 

The bottom line is that if you don’t change your metrics, you can’t evaluate AI impact. The revenues and costs directly attributable to AI first need to be separated from the core business and then broken down into their constituent parts. Without that, it’s impossible to answer with precision the three essential questions on everyone’s mind: Is AI driving incremental revenue? Is AI changing cost structures? And are AI-related products scaling efficiently?

Armed with firm answers, portfolio companies can not only design the best product roadmap and allocate resources accordingly, but GPs can start building the kind of evidence-based exit story buyers are demanding. As more and more due diligence processes recalibrate to focus on AI-specific metrics, the only way to get full credit for hard-won traction against plan is to back it up with clear, compelling proof points.

Twenty twenty-six will likely be remembered as the year AI truly redefined the software industry—the kind of no-turning-back moment that challenges all previous assumptions. Nothing about that is easy. But the chaos won’t last forever. The leaders coming out of this transformation are already hard at work rethinking how to underwrite risk, inflect portfolio company performance, and measure results in a world upended by AI.

One thing is clear: There’s no time to waste.