The Great
SaaS Reckoning
How higher rates and agentic AI hit the software industry at the same time — and what Tier-1 LPs should do about it. Three foundational assumptions underpinned every software buyout from 2015 to 2025. All three are gone simultaneously. This is not a valuation reset. It is a structural one.
Two Regime Shifts,
One Industry
Between 2015 and 2025, PE sponsors completed more than 1,900 software acquisitions at 15–20x revenue. Three assumptions underpinned every deal: near-zero borrowing costs, perpetual double-digit ARR growth, and seat count as a durable proxy for software value. All three are gone at the same time.
Higher benchmark rates turned interest coverage ratios from comfortable to precarious. Where 5x levered buyouts required 10% revenue growth to service debt, companies delivering 4–6% are underwater on coverage. Meanwhile, agentic AI has broken the seat-based pricing assumption that justified those multiples in the first place.
The combination is a structural reset — the Great Bifurcation — between thick systems with data gravity and thin wrappers about to be displaced. This is not a cyclical correction. The playbook that worked from 2015 to 2021 does not work anymore.
Three Case Studies in
Peak-Multiple Collapse
MEDALLIA ($5.1B gone): Thoma Bravo took it private in 2021 at $6.4B, financing ~$3B with debt when leveraged-finance markets demanded almost no risk premium. By late 2025, BDCs were quietly marking it down: FS KKR Capital at ~79 cents, Apollo Debt Solutions at ~74. By April 2026, Blackstone, KKR, Apollo, and Antares took the keys. Thoma Bravo's equity went to zero.
PLURALSIGHT (~$4B losses): Vista Equity Partners bought it for $3.5B in 2021. Pandemic-era growth proved the high watermark. As rates rose and corporate training budgets contracted, topline went negative. Vista tried the classic J.Crew IP drop-down in early 2024. It didn't work. Blue Owl-led consortium took control August 2024, converting ~$1.2B debt to equity. Vista absorbed ~$4B in losses.
QUALTRICS (debt market says no): March 2026 — JPMorgan plus 10+ banks halted a $5.3B debt financing. Investors passed, explicitly citing AI disruption anxiety. Qualtrics' existing $1.5B term loan fell from near par to ~86 cents in days. The signal: credit markets now underwrite AI displacement risk alongside leverage ratios.
Multiple Compression as
AI Defensibility Filter
Public SaaS multiples have compressed sharply from the 2020–2021 peak, when growth-oriented SaaS indices traded at median EV/Revenue above 18x. By Q1 2026, the picture is bifurcated — and the bifurcation itself is the signal.
The new variable in multiple determination is an implicit AI defensibility filter. Two companies with identical growth rates and NRR can trade at very different multiples depending on whether they look like thick Systems of Action with proprietary data moats, or thin wrappers at risk of agentic displacement.
What the market actually pays for: Rule of 40 ≥40% commands a 60–70% premium versus sub-40 peers. NRR above 110% (120%+ ideal) moves multiple turns of ARR. Gross margin above 75% is material at scale. And Revenue per Employee (RPE) has become the new AI-era signal — a steep upward slope separates winners from displaced.
The Structural Shock to
Seat-Based SaaS
The shift from passive LLMs to agentic AI systems is moving faster than most underwriting models assume. These agents don't just answer questions — they perceive, plan, act via tools, and adapt based on outcomes. They ingest data from CRMs, ERPs, and ticketing systems. They chain reasoning steps, call APIs, execute workflows, and coordinate with other agents.
The Klarna number: Klarna's AI customer service deployment handled approximately 2.3M customer conversations in its first operational month — the equivalent work of 700 full-time agents. Resolution times fell from ~11 minutes to under 2 minutes. Repeat inquiries dropped ~25%. Projected contribution to annual profit: ~$40M. The revenue Klarna might otherwise have paid to a human-seat-priced CX software vendor did not rise with that productivity. It should have fallen.
From SaaS to SaS (Service-as-Software): The coherent response is a pricing pivot — from per-user-per-month tax on human headcount to per-outcome fee: per resolved ticket, per closed deal, per reconciled ledger entry. SaS pricing aligns vendor and customer incentives. The more work the agents do, the more revenue the vendor earns.
We discount heavily any business where revenue is both predominantly seat-based and where credible agentic use cases exist but pricing remains unchanged. We reward businesses with early SaS pilots showing measurable economics: a defined percentage of contracts on outcome-based terms, demonstrable gross-margin uplift per automated workflow, and RPE inflecting upward as agents scale.
Four Tiers for a
Fractured Universe
The SaaS universe is splitting into two camps. Thick systems — ERPs, core banking platforms, industrial SaaS — have deep proprietary data, regulated workflows, and integration complexity that makes them expensive to replace. Thin systems — horizontal point tools, UI-centric apps, simple workflow wrappers — are exactly what agents can replace, automate around, or commoditize.
Any GP who claims their entire software book sits in Tier 3 and 4 is either not looking hard enough or not being honest with LPs. Most portfolios have more Tier 1–2 exposure than the quarterly letters suggest. The toggle above switches between NAV exposure guidance and exit multiple guidance for each tier.
Tier 4 carries the deepest moats and the highest new-capital priority. Core banking, heavy industrial ERP, and vertical SaaS with proprietary operational data are the assets where AI budgets flow toward the vendor, not away from it. Don't discount them because the UI looks like it was built in 2008.
Five Questions Every LP
Should Be Asking Now
This framework isn't just for Icarus Asia's portfolio — it's for every GP in our allocation. The five questions below are ones we now bring to every AGM. Some GPs find them uncomfortable. That is usually a signal.
The monitoring framework covers three dimensions quarterly: Financial (ARR, growth, Rule of 40, NRR, gross margin, CAC payback, RPE trend); AI and Product (% workflows agent-augmented vs fully orchestrated; seat-based vs outcome/usage-based revenue mix; % value from proprietary data); and Risk and Credit (net leverage, interest coverage, covenant headroom, AI disruption risk assessment by tier).
Click each question node in the panel to expand the full analytical ask. The questions sequence from portfolio-level diagnostics through asset-level product architecture to balance sheet stress — in that order intentionally. Start with the map. End with the debt.
The Medallia and Pluralsight restructurings were predictable, in retrospect, by anyone willing to look at leverage, product thinness, and AI exposure simultaneously. Most allocators weren't looking at all three at once. We now do this as a baseline quarterly exercise across every GP relationship. LPs who don't ask these questions in 2026 will be reading restructuring announcements in 2027.