Bifurcation Under Stress: How AI Is Resetting Enterprise SaaS
A market perspective on why the reset is real but not indiscriminate — and what disciplined operators and investors should do before the next refinancing window.
Introduction
Enterprise software is being repriced in real time. Public SaaS multiples have collapsed from roughly 13.5x forward revenue in 2021 to about 5.3x in 2025 — decade-plus lows — even as the underlying businesses grew more profitable. It is tempting to read that as a “SaaSpocalypse.” It isn’t. What we are watching is bifurcation under stress: a stack of pressures — higher-for-longer rates, slowing seat growth, leveraged 2021-vintage balance sheets and, now, an agentic demand shift — landing at once, not a single AI shock. As in every prior technology wave, this is not the moment to turn away from the technology, nor the moment to break the glass and go all in. It is the moment for disciplined conviction.
AI inverts three pillars of legacy SaaS economics
Where the pressure turns structural is unit economics. AI inverts the three pillars legacy SaaS was built on — how it earns, how it prices and how it renews.
- Gross margin: from software to substance. Classic SaaS runs near 80% gross margin. AI-native delivery carries inference as a variable cost of goods — averaging around 23% of revenue — dragging margins toward 50–60%. Software-like scalability does not automatically survive the shift to agentic delivery.
- Pricing: from seats to outcomes. The unit of value is moving from per-seat access to usage- and outcome-based pricing; Gartner expects roughly 40% of SaaS spend to price this way by 2030. Intercom already charges $0.99 for each support ticket an agent resolves — revenue indexed directly to the labor it replaces. The catch: usage pricing compresses net revenue retention (“NRR”) before it expands annual recurring revenue.
- Renewal: from auto-renew to agentic RFP. Renewals are no longer a foregone conclusion. Customers now run competitive RFPs at renewal to hunt for AI-native alternatives rather than rolling over the incumbent — and they do so precisely as 2021 LBO debt comes due.
The trap is that re-architecture is expensive — and the balance sheet is already stretched
Becoming AI-native is not a feature release. It is a full re-architecture of product, pricing, go-to-market and operating model — the workflow redesign and change management that value actually comes from, not a line item you buy. That would be hard in any market. It is harder now because software is a crowded, leveraged asset class: software-related exposure is roughly 13% of the ~$1.5T U.S. leveraged loan index, and an estimated 25–33% inside private-credit business development companies, with a wall of maturities landing in 2028 and 2029. Many leveraged vendors simply cannot self-fund the transition out of current cash flow while still servicing debt and supporting legacy customers. Call it the self-funding gap — the distance between what the pivot costs and what the P&L can spare — and it is what turns a strategic challenge into a solvency one.
Value is moving up the stack — from selling software to selling work
The upside is the larger half of this story. As foundation models commoditize, value accrues to whoever owns the workflow, the data context and the outcome — not to generic interface wrappers on top of someone else’s model. The companies that win share three traits, what Sequoia frames as the “MAD” stack: Moats (proprietary data plus deep workflow integration), Affordances (selling outcomes, not seats) and Diffusion (product-led growth embedded in the workflow). The tape already shows it. Bessemer’s Cloud 100 has AI-native names trading around 24x ARR against 19x for peers — a ~26% premium for genuine AI positioning — and companies with deeply embedded AI are growing roughly twice as fast as those treating AI as a supporting feature.
Underneath the re-rating sits a bigger reframe: the shift from selling access to selling work. Price a product around output delivered rather than seats occupied, and the addressable market stops being “software” and becomes a share of the far larger labor and services economy. On a directional basis, that reframe could expand a ~$200B traditional SaaS market by an order of magnitude. Coding is simply the first category to prove it, as developers pay for completed code rather than a tool that helps them write it; legal, finance and customer service — all discrete, verifiable outputs — are the likely next dominoes.
How to move forward
The disciplined move in a structural reset is the one it always is: sort the assets honestly, then fund only the fights worth winning. Triage every product line — and every portfolio company — into four archetypes:
- Compound: regulated, vertical or systems-of-record names with proprietary data. Extend the platform with embedded AI.
- Transform: moated systems of record facing seat erosion. Convert the business model to outcome and agent pricing.
- Build: attractive categories where an AI-native offering can win. Build the data and workflow moat deliberately.
- Harvest: horizontal, seat-priced point solutions. Amend-and-extend, run for cash, and don’t over-invest in a losing pivot.
For C-suite operators, that resolves into four moves: score your own products for AI exposure before the market does it for you; reset the commercial model from seat to outcome pricing and build the telemetry to forecast usage-based revenue; rebuild the operating model around rapid test-and-learn, because AI-native operational muscle is the new moat; and make hard portfolio choices by product line, because capital discipline beats AI ambition.
For investors — particularly in private equity and private credit — the same logic runs through the portfolio: re-rate every portfolio company (“PortCo”) on AI exposure and sort it into defend, rebuild, harvest or exit before the next refi window; make AI survivability first-class diligence, haircutting NRR durability and gross margin for inference exposure on every new deal; pre-empt the maturity wall by addressing refinancings 24-plus months early; and hunt the dislocation, where mispriced credit sits on an intact AI roadmap and an overlevered sponsor.
The reset is real, but it is not indiscriminate. The question facing SaaS is no longer whether AI compresses the old model — it plainly does — but whether a company can re-architect toward selling work fast enough, and fund it honestly enough, to land on the winning side of the split.
Sources: FTI Research & Analysis, drawing on Software Equity Group, Sequoia Capital, Bessemer, Gartner, Barclays, and JP Morgan, among others.