Don’t Believe Those Who Say that…

1. Key Themes

Theme 1: The "SaaSpocalypse" Was Panic-Driven, Not Data-Driven

The February 2025 selloff erased $300B in software market cap and spawned a death narrative — but actual spending behavior tells a different story.

"The cleanest problem with the SaaSpocalypse story is that the spending data does not behave like an obituary. Ramp's contract data shows seat-based contracts still representing roughly 65% to 75% of SaaS spend... Consumption-based spend, the model many people assume will rapidly replace seats, remains only 4% to 6% of the total. Those lines barely moved over twelve months."

Theme 2: Shallow SaaS Is Dying — Deep, Vertical SaaS Is Not

The article draws a sharp line between software that was protected by inertia and distribution vs. software embedded in mission-critical workflows.

"What is dying is not SaaS, but the idea that every thin software wrapper deserves durable SaaS valuations... The vertical SaaS market is still projected to grow from $133.5 billion in 2025 to $194 billion by 2029, which is difficult to square with a true category death."

Theme 3: Agentic AI Has a Wide Gap Between Demo-Ready and Enterprise-Ready

The replacement thesis assumes AI agents already work reliably in production. Security and reliability data say otherwise.

"A 2025 arXiv paper on LLM agent vulnerabilities found 94.4% of tested agents susceptible to prompt injection, 83.3% to retrieval-based backdoors, and 100% to inter-agent trust exploits... For regulated industries, that is not a mild adoption concern. It can block deployment entirely."

Theme 4: Per-Seat Pricing Is Under Pressure, But Hasn't Collapsed

The bear case has real merit in pricing erosion, but the timeline for disruption is being dramatically overstated.

"Per-seat pricing is under pressure for a simple reason. If AI agents can produce the output of several employees, software contracts tied to employee count start to look exposed... But a model under pressure is different from an industry in terminal decline."

Theme 5: The Transition to Outcome-Based Pricing Is Real But Slow

AI features are currently being absorbed into existing pricing structures rather than forcing wholesale contract rewrites.

"Adobe still shows roughly 99% seat-based billing in Ramp's spend panel even after adding Firefly AI credits, which suggests that AI features can be absorbed into existing pricing before they force a complete rewrite of the contract model. HubSpot's entire non-subscription revenue line sits around 2% of FY2025 total, which means Breeze AI's usage-based contribution is a rounding error."

2. Contrarian Perspectives

Contrarian 1: Consumption-Based Pricing Hasn't Actually Taken Over — And May Not Anytime Soon

The consensus assumes AI is rapidly forcing a shift from seats to usage-based billing. The contract data contradicts this.

"Consumption-based spend, the model many people assume will rapidly replace seats, remains only 4% to 6% of the total. Those lines barely moved over twelve months. That is the part the 'death of SaaS' argument has to explain and usually does not."

Supporting evidence: Adobe at 99% seat-based billing post-AI feature launch; HubSpot's usage-based revenue at ~2% of total despite a Spring 2025 AI product launch.

Contrarian 2: AI-Generated Code Is Less Safe Than Human-Written Code, Making the "Build vs. Buy" Shift Slower Than Assumed

Many bulls on the "build internally with AI" thesis overlook the quality and security deficiencies of AI-generated code.

"AI-generated code has been found to contain 1.7 times more major issues than human-written code, with 45% introducing known security vulnerabilities. A USENIX Security paper titled 'We Have a Package for You' found 205,474 unique hallucinated package names across 576,000 generated code samples, a pattern attackers have learned to exploit."

This undermines the narrative that enterprises can cheaply rebuild SaaS tools in-house using AI coding agents.

Contrarian 3: Agentic AI Creates New Financial Risk for CFOs That Seat Pricing Does Not

Usage-based AI pricing is often framed as modernization. But it introduces variable cost risk that enterprises are poorly equipped to manage.

"Agentic AI workflows can burn tokens unpredictably when tasks branch, loop, or require repeated calls across tools. That creates denial-of-wallet risk, where a broken or manipulated agent consumes far more compute than intended. A seat-based contract may look old-fashioned, but it gives CFOs a cost ceiling they understand."

3. Companies Identified

Ramp

Description: Corporate spend management platform
Why mentioned: Its contract-level spending data is used as primary evidence that seat-based SaaS has not collapsed
Quote: "Ramp's contract data shows seat-based contracts still representing roughly 65% to 75% of SaaS spend."

Adobe

Description: Creative and document software platform
Why mentioned: Used as a case study showing AI features being absorbed into legacy seat pricing without forcing contract model changes
Quote: "Adobe still shows roughly 99% seat-based billing in Ramp's spend panel even after adding Firefly AI credits."

HubSpot

Description: CRM and marketing platform
Why mentioned: Used to illustrate how AI products (Breeze AI) have had negligible financial impact on pricing model despite high-profile launch
Quote: "HubSpot's entire non-subscription revenue line sits around 2% of FY2025 total, which means Breeze AI's usage-based contribution is a rounding error."

Workday

Description: Enterprise HR and financial management software
Why mentioned: Its 8.5% workforce cut in early 2025 created negative optics for the sector, as it directly manages the headcount systems under AI pressure
Quote: "Workday's decision to cut 8.5% of its own workforce in early 2025 also landed awkwardly for the sector, because it came from a company whose products sit close to the very headcount systems now being questioned."

Zylo

Description: SaaS management platform tracking enterprise software portfolios
Why mentioned: Its 2026 SaaS Management Index provides data on enterprise vendor consolidation
Quote: "Zylo's 2026 SaaS Management Index found that the average enterprise portfolio had already shrunk from 374 to 342 applications, a sign that buyers are consolidating rather than expanding their vendor footprint."

Epic

Description: Electronic health records platform
Why mentioned: Cited as an example of vertical SaaS so deeply embedded it is effectively irreplaceable
Quote: "Healthcare platforms like Epic and Cerner sit so deep inside clinical operations, billing flows, records, compliance requirements, and institutional memory that replacing them is less a software decision than an institutional one."

Cerner

Description: Healthcare IT and EHR platform
Why mentioned: Co-cited with Epic as a model for deeply embedded vertical SaaS with high switching costs
Quote: Same as above.

IQVIA

Description: Life sciences data and technology company
Why mentioned: Cited as a vertical SaaS example where proprietary data networks and regulatory context create durable moats
Quote: "Life sciences infrastructure like IQVIA has a similar kind of depth, where data networks and regulatory context matter as much as software features."

Microsoft (Copilot)

Description: Enterprise software and AI platform
Why mentioned: The EchoLeak vulnerability in Microsoft Copilot is cited as evidence that agentic AI security risks are real, not theoretical
Quote: "The EchoLeak vulnerability in Microsoft Copilot made the risk feel less theoretical, because it showed how sensitive enterprise data could be exposed through the systems companies are now being asked to trust more deeply."

Clay

Description: Data enrichment and outbound sales platform
Why mentioned: Sponsor/partner for a live workshop on AI-powered prospecting for seed-stage founders; also cited as a tool founders winning on distribution are using
Quote: "HubSpot for Startups and Clay are running a live build-along workshop on June 18 at 11am ET for seed and pre-seed founders."

4. People Identified

Ruben Dominguez

Description: Author of The VC Corner newsletter
Why mentioned: Author of this piece; provides the analytical framework separating structural SaaS decline from cyclical/model-level correction
Quote: "The right response to the SaaSpocalypse is underwriting discipline, not category panic."

(No other named individuals are cited in the article.)

5. Operating Insights

Insight 1: Audit Whether Your Revenue Model Is a Headcount Bet in Disguise

Founders should pressure-test whether their expansion revenue is tied to customer headcount rather than product-delivered value — before investors or buyers do it for them.

"If your revenue expands mainly because customers hire more people, if your product has no proprietary data advantage, and if the workflow can be copied by an internal team using AI agents, the market is going to ask harder questions. Better to ask them yourself while you still have room to adjust."

Insight 2: Move Toward Outcome-Based Pricing Before Renewal Pressure Forces It

Rather than abandoning per-seat pricing entirely, founders should identify where their value is better measured in outcomes and begin introducing that framing into contract conversations proactively.

"If the value sits closer to completed tasks, reduced labor, faster throughput, or measurable business outcomes, then outcome-based pricing should become part of the contract conversation before renewal pressure forces the issue."

Insight 3: Investors Should Separate "Embedded Software" from "Feature With a Subscription"

The broad selloff punished strong vertical SaaS alongside weak horizontal tools — creating a differentiated buying opportunity for disciplined underwriters.

"What the selloff actually surfaced was a misclassification that had been building for years: software priced as infrastructure that was really just a feature with a subscription attached."

6. Overlooked Insights

Insight 1: AI Hallucinated Package Names Are an Active Attack Surface in Production

This security detail goes beyond general "AI is unreliable" commentary — it describes a specific, exploitable attack vector already in use.

"A USENIX Security paper titled 'We Have a Package for You' found 205,474 unique hallucinated package names across 576,000 generated code samples, a pattern attackers have learned to exploit because fake dependencies become real attack surfaces once developers copy them into production."

For founders building internal tools with AI coding agents, this is a concrete, underappreciated operational risk that slows the build-vs-buy shift.

Insight 2: The "Denial-of-Wallet" Risk of Agentic AI Has CFO-Level Implications

Beyond security, the unpredictable cost structure of agentic workflows creates a procurement and budgeting problem that hasn't received enough attention in the pricing model debate.

"Agentic AI workflows can burn tokens unpredictably when tasks branch, loop, or require repeated calls across tools. That creates denial-of-wallet risk, where a broken or manipulated agent consumes far more compute than intended."

This gives traditional seat-based SaaS vendors an underappreciated retention argument with finance-savvy enterprise buyers: cost predictability is itself a feature.