The market correction is real, but PE investors who treat this as a blanket retreat from software will miss the cycle’s biggest opportunities.
Seemingly overnight, the media narrative has shifted from “software eats the world” to “AI eats software.”
When agentic tools capable of automating legal, financial, and data-services tasks launched in early February, relevant tech stocks sold off within hours. As of February 20, the North American Tech Software Index had declined about 30% from its peak in mid-September, according to data from S&P Capital IQ.* Private markets are feeling the pressure too, particularly the 2021 and 2022 PE vintages that deployed capital near peak valuations. Those companies are now seeing markdowns, tighter exit windows, and growing LP scrutiny about the actual value of their software portfolios
In our view, the market is right that a repricing is underway, but it’s missing the mark about what should be repriced. Software is still an important asset class. AI just highlights the difference between durable platforms and replaceable tools. Platforms based on essential workflows, unique data, and deep industry expertise will, we believe, see their position strengthen. Those built on surface-level features and seat-based growth models face sharper pressure as AI lowers barriers and accelerates competition.
That divergence creates opportunity for investors. Broad valuation compression and market uncertainty make it harder to distinguish between resilient, defensible, businesses where AI represents a growth accelerator and the exposed, existentially threatened, legacy models. An investor's edge now lies in disciplined diligence across the full spectrum of commercial, operational, financial, and technical considerations.
The headlines offer a conflicting narrative to the sector’s actual performance in recent years. US technology sector EBITDA margins for public companies have rebounded materially since 2022 and remain structurally elevated across most sub-sectors. Free cash flow generation across the best SaaS businesses is at record levels.
The next decade of successful software deals will be focused on growth acceleration, pricing power, and defensibility beyond the code. When determining valuations, investors should be ruthlessly focused on how the AI roadmap is enhancing the moat of the business and aligning with customer behavior.
Why software valuation models are changing
Traditionally, SaaS has sold access to a tool: a CRM platform, an analytics dashboard, a project management interface. Now, agentic software is about delivering outcomes: a closed lead, a resolved support ticket, a drafted contract reviewed against a proprietary playbook. This shift from “tool provider” to “outcome provider” doesn’t shrink the addressable market. For many firms, the change expands the market dramatically, converting labor costs into software investments and opening new opportunities that were previously too costly to address with employees alone.
The historical line between “software” and “services” TAM is now blurring. Software providers with clear agentic strategies can expand their TAM into what was historically labor-based services. Additionally, many services firms and system integrators are also using AI to codify and “productize” repeatable delivery. While that shift increases the TAM in certain subsectors, it may also shift competitive dynamics.
The largest tech operators are shifting from process to outcomes as AI experiments are scaled into enterprise-wide transformation. The organizations seeing real impact are the ones building centralized orchestration layers, setting clear metrics, and redesigning workflows from the ground up, not just bolting a copilot onto existing processes. Leading companies are defined not by which AI tools they use but by their willingness to fundamentally rethink how work is accomplished.
How PE should evaluate the ‘defensibility’ of a software target’s moat
Not every software subsector or company will be equally impacted by AI. Technology serving industries characterized by high regulatory scrutiny and operational complexity (e.g., public sector, financial services, healthcare, energy and aerospace) are more insulated due to compliance safeguards that aren’t easily replicated by AI. Many vertical and niche SaaS providers in these industries possess curated domain datasets, embedded compliance logic, and partner ecosystems built over many years. This can create an ecosystem moat that safeguards the “last mile” of workflow execution.
Similarly, platforms in highly complex sub-sectors like cybersecurity can see outsized acceleration from AI given their access to proprietary threat intelligence, continuous data streams, and deeply embedded positions within enterprise infrastructure that enable automated detection and response at scale. While the entire sector is facing new AI-related questions, we believe that for the top-performers AI can be a force multiplier rather than a threat.
That said, opportunity isn’t limited to a specific list of subsectors or solutions but, rather, to companies with common defensibility characteristics, such as:
Domain depth and ecosystem entrenchment
“Vibe-coding” sounds easy until you get to the last-mile and encounter industry-specific workflows, edge cases, compliance requirements, and the messy reality of change management. Defensibility is often forged by the product choices that look boring in a demo, but matter enormously in production: specialized data models, configurable rules engines, and compliance guardrails built over years of customer feedback. Code is now easy to write. What is hard to replicate is the years of accumulated domain expertise, regulatory understanding, and customer relationships that make enterprise software sticky.
Proprietary context, not just proprietary data
Having access to proprietary data matters, but so do data rights, provenance, and the ability to use the data securely. Trust and auditability are a defensible moat in regulated environments. The strongest businesses generate proprietary context that makes AI better over time in ways competitors can’t easily replicate, things like curated knowledge graphs, validated playbooks, and customer-specific configurations that cannot be scraped or synthetically reproduced.
Mission-criticality and ‘workflow gravity’
If a product owns the system of record and is tied to a financial or regulatory outcome, agents tend to layer on top rather than replace. These are the businesses where AI creates a compounding advantage with more automation, more data, more value captured per workflow cycle. In contrast, standalone BI tools, collaboration suites, and horizontal workflow products that compete primarily on UX are squarely in the crosshairs. Since AI copilots can replicate these interfaces natively, they can enhance competition and exert severe pricing pressure.
The more embedded software is within an organization, the more switching becomes less about the technology and more the organizational implications. Products with true “workflow gravity” tend to be systems of record tied to financial or regulatory outcomes. When AI agents are used, they often drive more automation through the platform rather than routing around it. In this way, efficient products accelerate usage and outcomes instead of cutting back.
The inverse of this can be seen in the structural disadvantage facing narrow AI startups that build on top of standard foundation models and target a single task. These companies can often burn through cash paying for model access while lacking the advantages that broader platforms have spent decades building.
Build vs. buy vs. partner considerations are being rewritten in real-time
AI lowers the friction to build, and that changes the math for everyone involved. Years of institutional logic deemed “building” a less efficient path for many operators. Smaller, more focused solutions emerged in many sub-verticals. They grew quickly with a goal of being acquired by a larger competitor. If those larger competitors can now build similar products in-house, the acquisition thesis weakens and makes it harder for PE-backed point solutions to achieve liquidity through M&A.
This is one reason the markets have been so jumpy. “Workflow software” that looked secure because it was hard to build may suddenly look less secure in the context of competitors “building” solutions with small teams prototyping quickly. That said, a prototype is not a product, and a product is not a platform a compliance team will sign off on. The companies that understand this distinction, and can articulate it to buyers, will hold pricing power even as the cost of writing code approaches zero.
Partnering is more important than ever. Companies that establish themselves as essential nodes in an ecosystem through APIs, deep integrations, and emerging interoperability standards like Model Context Protocol (MCP) that let AI agents orchestrate across tools, can create stickiness that transcends the build-versus-buy question entirely.
How the diligence playbook is changing
The implications for how investors conduct diligence, underwrite, and create value are continuing to evolve. Three shifts in particular deserve attention.
Retention metrics need reframing
Net revenue retention (NRR) has long been the marquee metric for SaaS quality, but in an AI-disrupted sector NRR can mask seat contraction beneath expansion revenue from AI add-ons. Gross revenue retention (GRR) can be more helpful in assessing near-term defensibility and underlying durability of the contract base, stripped of upsell. Investors need to disaggregate retention metrics by cohort, by product module, and by AI-impacted compared to non-impacted revenue streams.
Pricing models are in transition
Seat-based pricing, the foundational unit economics of SaaS 1.0, is under structural pressure. If an AI agent can do the work of three analysts, the customer does not need three seats. The winning companies are moving toward outcome-based or value-based pricing that ties revenue to the impact delivered rather than the number of humans touching the software. This is a healthier long-term alignment, but it also means revenue can be harder to forecast and requires new diligence frameworks to underwrite.
As software shifts from aiding human productivity to autonomously completing work, value-based pricing captures the genuine impact of AI and allows revenue to scale naturally as customers mature. The old SaaS playbook of pricing per seat needs to give way to models that reflect what the software accomplishes.
Value creation means growth acceleration, not just margin expansion
The traditional PE software playbook of acquire, cut costs, expand margins, and improve go-to-market is necessary but no longer sufficient. The companies that will command premium exits are those using AI to accelerate their own product roadmaps—shipping features faster, deploying AI capabilities that enhance customer stickiness, and building data moats that compound over time. For PE firms sitting on aging portfolios and compressed exit windows, the ability to articulate a credible AI value-creation story is no longer optional. It’s a prerequisite for liquidity.
Growth acceleration is becoming a critical success factor. The focus of management teams should be on increasing the slope of their trajectory through AI-native product development. Margin expansion matters, but product velocity is becoming the differentiator that separates premium exits from average ones.
