In January 2026, Monday.com CEO Eran Zinman quietly made an announcement that should have stopped every SaaS CFO in their tracks. He had replaced the company’s entire 100-person SDR team with AI agents. The results were not equivocal: response times dropped from 24 hours to 3 minutes and conversion rates went up. The agents were better. They were faster. They cost less. And Monday.com — a company whose valuation was built on selling seats — had just demonstrated that its own customers could run a core function with a fraction of the headcount that used to buy those seats.

This is the budget war that’s already started. Per-seat SaaS pricing — the model that powered a multi-trillion-dollar industry for two decades — is structurally incompatible with a world where AI agents replace the humans who use software. The February 2026 SaaSpocalypse wiped $285 billion from SaaS stocks in 48 hours, but the seat compression underneath it has been compounding since late 2025. The question now is not whether this changes your software budget. It’s who controls that change — you, your vendor, or a faster competitor.

For more context on the broader valuation collapse driving this, see our analysis of The Great SaaS Reset.

The Math That’s Breaking SaaS

The per-seat SaaS model worked on a single beautiful assumption: as your customers grow, they hire more people, those people use your software, and your revenue grows automatically. It was predictable, scalable, and loved by Wall Street for exactly those reasons. The model built companies worth hundreds of billions. Then AI broke the underlying logic in one sentence: if 10 AI agents can do the work of 100 sales reps, you don’t need 100 Salesforce seats. You need 10.

That is not a marginal change. That is a 90% reduction in seat revenue for the same business output. And it compounds across every human-facing function in the enterprise simultaneously. Paid.ai, which has conducted over 250 conversations with AI agent founders, found 75% of AI companies actively struggling with this pricing paradox. One founder summarised it precisely: “I am actually reducing headcount to serve the same purpose. So when I have fewer seats, the credits won’t make up for the difference, because the credits are attached to a seat. My expansion became a nightmare.”

“I fundamentally believe seats are going to die. Because seats are being destroyed when workflow consumption is going up.” — AI agent founder, via Paid.ai (2025)

The brutal irony is that the more successful an AI agent product becomes — the more tasks it automates, the more headcount it replaces — the worse the economics get for any vendor still running seat-based pricing. Variable LLM and compute costs go up as usage scales. Fixed seat revenue stays flat or contracts as headcount falls. The math doesn’t work. Paid.ai’s data shows 318 separate mentions of seat pricing challenges across their customer conversations, with one company’s board issuing a directive: move off seat pricing or face existential competitive pressure from tools that already have.

Which SaaS Categories Are Hit Hardest

Seat compression does not hit all SaaS categories equally. The exposure scales with how much of a product’s value is delivered through a human logging in, navigating a UI, and performing tasks. Products where an AI agent can replicate the entire workflow — query the data, take action, generate output — without a human in the loop are structurally most exposed. Products where the human is the point (creative judgment, strategic decisions, relationship management) are more durable, though not immune to compression as AI raises output per worker.

Highest Exposure

Sales engagement and CRM tools (Salesforce, HubSpot, Outreach) face the most direct threat. The SDR function — prospecting, sequencing, outreach, follow-up — is precisely the workflow AI agents automate end-to-end. Salesforce shares fell 26% since early 2026, explicitly linked to “customer seat compression” in analyst reports. Project management tools (Monday.com, Atlassian, Asana) are next: Atlassian reported its first seat-count decline in Q1 2026 and saw a 35% stock drop as investors connected fewer developers to declining seat revenue. Customer support platforms (Zendesk, Intercom) are already seeing agents handle 60–80% of tier-1 support without human involvement.

Medium Exposure

HR and workforce tools (Workday, Rippling) face compression as AI handles routine HR workflows — onboarding paperwork, benefits administration, compliance tracking. Workday fell 22% year-to-date on per-seat licensing concerns. Analytics and BI tools face a different version of the problem: AI can now generate the reports that once required a dedicated analyst team, reducing the headcount that needs seats. Marketing automation is bifurcated — platforms with deep data moats (audience history, proprietary signals) are more defensible; those that primarily automate content creation are not.

Lower Exposure (But Not Zero)

Creative tools (Adobe) face what analysts call “process compression” rather than direct replacement — AI raises output per creative, which can still pressure seat growth even if it doesn’t eliminate roles outright. Security and compliance tools benefit from governance requirements that keep humans in the loop and from the fact that AI agents themselves create new security requirements. Data infrastructure tools (Snowflake, Databricks) are positioned as AI infrastructure rather than AI’s target — though API data toll fights are emerging as a new battleground.

The New Pricing Models Replacing Seats

The industry has not converged on a single replacement for seat-based pricing. Four distinct models are competing for dominance, and the winner varies by product category, buyer sophistication, and vendor bargaining power. For operators and buyers, understanding these models is now a procurement skill — not just a finance exercise.

Usage-Based / Consumption Pricing

Pay per API call, token, workflow run, or action completed. The model aligns cost with actual use — buyers don’t pay for idle capacity, and vendors capture revenue when usage scales. The problem: variable monthly costs are hard to forecast and budget, and CFOs hate uncontrolled consumption bills. Buyers can also throttle usage to cut costs at renewal time. Interestingly, the 2026 Guide to SaaS Pricing notes a potential reversal: as AI drives down cost-to-serve dramatically, some vendors are finding they can offer competitive per-seat pricing again — undercutting usage-based competitors with the simplicity of a flat fee.

Outcome-Based / Value Pricing

Pay for the business result: closed deals, resolved tickets, processed invoices. This is theoretically the most aligned model — the vendor only wins when the customer wins. The implementation challenges are formidable: measuring outcomes requires shared data access and agreed definitions, attribution is contested, and billing disputes are common. Gartner projected over 30% of enterprise SaaS would incorporate outcome-based components by 2025, up from 15% in 2022. The trajectory is clear even if implementation is still maturing.

Agentic Enterprise License Agreements (AELAs)

A flat-fee all-you-can-eat model specifically designed for the agent era. Salesforce pioneered the AELA with Agentforce — a flat fee where customers can scale AI agent usage without per-interaction charges, framed as “shared risk.” Constellation Research predicts AELAs will become the norm as CIOs push back against unpredictable consumption models. The appeal for buyers: budget predictability. The risk: vendor lock-in at enterprise scale.

Credit-Based Hybrid Models

A pool of credits that can be applied across different agent actions at different rates. More flexible than pure seat pricing, more predictable than pure consumption. The trap, as one Paid.ai founder noted: when credits are attached to seats, you inherit all the structural problems of the model you were trying to escape. The best implementations decouple credits from users entirely — selling capacity, not licenses.

The Buyer Playbook: Using Seat Compression as Leverage

For enterprise IT and procurement teams, seat compression is not just a vendor problem — it is a negotiating weapon. Your SaaS vendors are acutely aware that their seat counts are declining. Their ARR forecasts are under pressure. Their investors are asking hard questions at every earnings call. That combination makes 2026 the best renewal negotiating environment in a decade, but only if you come prepared with the right data.

The first move is a usage audit before every renewal. BetterCloud’s research shows that traditional SaaS licenses are the primary funding source for new AI projects — the industry term is “Cost Reallocation Potential” (CRP). Your existing seat contracts represent the budget you can redeploy. Audit actual seat utilisation, identify idle licenses, and quantify which workflows are being partially or fully automated. That data is your opening position.

The second move is pushing for contract flexibility clauses in every renewal. Specifically: terms that allow you to reduce fixed seats mid-contract if automation demonstrably reduces human usage. This turns a rigid cost centre into a strategic lever. Vendors who refuse these clauses are signalling that they know their seat counts are at risk — and that’s information you can use.

The third move is demanding AI roadmap transparency from every strategic vendor. The question is not “do you have AI features?” — every vendor claims that in 2026. The question is: does your product serve as infrastructure for AI agents, or does it compete with what AI agents can now do directly? Vendors building agent-readable APIs and data layers are bets worth renewing. Vendors rebranding existing features as “AI-powered” without architectural changes are those to consolidate away from. For a full negotiation framework, see our guide to the SaaS Reset buyer playbook.

Vendor Responses: Who’s Adapting and Who’s at Risk

Not every incumbent SaaS vendor is losing this battle. The ones navigating it best share a common characteristic: they’ve positioned their product as infrastructure for AI agents rather than competing with what agents can do. The ones struggling are trying to patch seat-based models with AI feature labels rather than rebuilding the underlying pricing and architecture.

8 Moves for the Seat Compression Era

The seat compression era rewards those who act on structural clarity rather than waiting for consensus. The following moves are sequenced by who they apply to — but the underlying logic is consistent: understand the new economics, reprice your position accordingly, and move before your counterparty does.

The Three-Stage Disruption Map

Taskade’s analysis of the seat compression timeline identifies three distinct stages of disruption — and understanding where you are in this sequence changes which moves are urgent versus which can wait.

Stage 1 — Seat Compression (now): AI agents reduce the number of humans using legacy SaaS tools. Revenue contracts. Stock prices fall. This is the stage we’re in. Enterprise buyers are reducing headcount, seat counts are declining at renewal, and SaaS valuations are repricing downward. The operational response is audit-and-renegotiate: identify where seats are being displaced and redeploy that budget to AI tooling.

Stage 2 — Feature Extraction (2026–2027): Teams begin pulling individual capabilities out of monolithic platforms and replacing them with AI-native point solutions. Instead of Salesforce for CRM plus marketing plus analytics, enterprises build workflows across three specialised tools that each cost less and outperform the equivalent Salesforce module for their specific use case. This is already beginning in marketing automation and customer support. The operational response is stack rationalisation — mapping which capabilities in each monolith are genuinely sticky versus which are being unbundled.

Stage 3 — Workspace Consolidation (2027+): Surviving platforms become the operating system for teams — the central workspace where AI agents, automations, and human collaborators converge. Everything else becomes an integration feeding data to the central layer. This is the “services as software” thesis in action: the platform that sits in the execution path, captures decision traces, and owns the context graph wins. Everything below it becomes a queryable database. The Foundation Capital analysis articulates this clearly: AI startups are targeting the services market, not just the software market — selling outcomes, not seats.