Is per-seat pricing really dead?

Not entirely. Per-seat still works when each user genuinely represents a separate person getting distinct value. Slack, Figma, and Linear charge per seat because each seat corresponds to a real human doing differentiated work. But for tools where AI agents can act on behalf of users (CRMs, analytics, support, content creation), per-seat pricing is losing to usage-based and outcome-based models. The question to ask: would an AI agent need its own seat to use your product? If yes, your pricing model has a problem.

What is Workflow-as-a-Service (WaaS)?

WaaS is a pricing model where customers pay per task or workflow completed, rather than per user or per month. Think of it as turning your SaaS into a utility. You pay for what you use, measured in actions rather than access. Stripe is the canonical example: they don't charge per seat. They charge per transaction. The more transactions you process, the more you pay, and the more value you're getting.

How should early-stage startups think about PLG pricing in 2026?

Start with a generous free tier to build adoption, but design your pricing around a value metric from day one. Measure the action your product enables (documents created, analyses run, decisions made) and plan to charge per action once you have [product-market fit](/glossary/product-market-fit). Don't default to per-seat just because everyone else did. The infrastructure you build for metering and usage tracking will pay for itself when you're ready to monetize.

PLG Pricing in the AI Era

TL;DR: Product-led growth pricing is being rewritten. Per-seat models are dying as AI agents replace human users. The new models: per-task, per-outcome, and hybrid PLG.

Per-seat pricing made sense when humans were the only users of software. One person, one login, one seat. Simple.

That assumption is crumbling. In 2026, AI agents are the fastest-growing "user" segment of most SaaS products. One PM with Claude Code replaces a 5-person team. A single marketer with AI workflows produces the output of an entire content department. Klarna replaced hundreds of customer service reps with AI, and their per-seat SaaS spend cratered overnight.

The per-seat model penalizes efficiency. The more productive your customers become with AI, the fewer seats they need, and the less they pay you. That's a death spiral for SaaS revenue.

TechCrunch's "SaaSpocalypse" coverage in March 2026 captured what many founders already felt: the pricing model that built the SaaS industry is breaking. Here's what replaces it.

Quick Answer (TL;DR)

Per-seat SaaS pricing is breaking down because AI agents, not humans, are becoming the primary "users" of software. The replacements: per-task pricing (Workflow-as-a-Service), per-outcome pricing (Results-as-a-Service), and hybrid self-serve + sales models. PMs who don't rethink pricing now will watch revenue erode as customers consolidate seats.

Why Per-Seat Pricing Is Breaking

The math is simple. If your product charges $20/seat/month and a customer has 50 seats, you earn $1,000/month. When that customer deploys AI agents that let 10 people do the work of 50, they drop to 10 seats. Your revenue falls to $200/month. The customer is getting more value from your product than ever, but you're earning 80% less.

This is already happening:

AI agents now "use" software on behalf of humans. API calls, automated workflows, and agent-driven processes don't need seats. They need access.
One person + AI can replace entire teams. This isn't theoretical. Engineering teams using AI coding assistants are shipping 3-5x more code with fewer developers.
Per-seat penalizes your best customers. The most efficient, AI-forward companies pay you the least. That's backwards.
Companies are consolidating tools. Why pay for 5 specialized products when one AI-powered platform handles all five use cases?

The product-led growth playbook that drove the last decade of SaaS growth assumed human users as the growth vector. That assumption needs an update.

The Three Replacement Models

Per-Task Pricing (Workflow-as-a-Service)

Charge per action completed, not per person logged in. This is the utility model applied to software.

How it works: Define the discrete actions your product enables. Charge per action. Scale linearly with usage, regardless of how many humans or AI agents are doing the work.

Examples in the wild:

Stripe charges per transaction processed
Twilio charges per message sent
Vercel charges per build minute and function invocation
AWS Lambda charges per function execution

Best for: Products with measurable, discrete actions where the cost to serve scales with usage. If you can count the "thing" your product does, you can price per-task.

The PM move: Calculate your unit economics first. Use the LTV/CAC Calculator to model what per-task pricing looks like for your margins. Know your cost-to-serve per action before you set prices.

Per-Outcome Pricing (Results-as-a-Service)

Charge based on value delivered, not usage volume. This is the hardest model to implement but creates the strongest alignment between your revenue and your customer's success.

How it works: Identify the outcome your customer cares about. Tie your pricing to that outcome. You only get paid when you deliver results.

Examples in the wild:

Performance marketing platforms charge per qualified lead
AI coding tools experimenting with pricing per merged PR
Recruiting platforms charging per successful hire
SEO tools charging based on ranking improvements

Best for: Products where the outcome is clearly measurable and attributable. If you can prove your product caused the result, you can charge for it.

The catch: You need bulletproof measurement. If customers dispute whether your product delivered the outcome, you have a billing nightmare. Build your metering and attribution infrastructure before you launch outcome-based pricing.

Hybrid PLG + Sales

Self-serve for individuals and small teams. Sales-assisted for enterprise. This isn't new, but it's becoming the dominant model for a reason: it works at every scale.

How it works: Free or cheap entry point drives bottom-up adoption. Individual contributors discover and love your product. When their company wants to roll it out, your sales team handles the enterprise deal with custom pricing, SSO, compliance, and higher margins.

Why it's winning in 2026:

Most $100M+ PLG companies are now hybrid (Notion, Figma, Canva, Linear)
Pure self-serve leaves enterprise money on the table
Pure sales-led can't compete with bottoms-up adoption speed
AI makes the individual user more powerful, which actually accelerates bottom-up adoption

The key insight: product-led growth isn't opposed to sales. It's a distribution strategy that makes sales more efficient. Your sales team closes warm leads who already use and love the product.

What This Means for PMs Building SaaS

Five concrete steps to take this quarter:

Audit your pricing model. Model what happens when AI agents replace 50% of your per-seat revenue. If the answer is "we lose half our revenue while customers get more value," you have a pricing problem.

Identify your value metric. What action or outcome does your product enable? Documents created, analyses completed, decisions made, pipelines built? That's your pricing anchor.

Build metering infrastructure. You can't charge per-task if you don't measure tasks. Instrument your product to track the actions and outcomes that matter.

Plan your product roadmap around pricing. Pricing isn't a sales decision. It's a product decision. Your roadmap should include metering, usage dashboards, and flexible billing as first-class features.

Size your market correctly. Use a TAM calculator that accounts for AI agent "users." If your TAM model counts seats, it's probably wrong. Count workflows, tasks, or outcomes instead.

The Free-to-Paid Transition Is Changing Too

The classic PLG playbook said: offer a generous free tier, let users fall in love, then convert them to paid. That playbook has a new problem: AI bots.

Free tiers are getting hammered by automated agents, scrapers, and AI-driven workflows that consume resources without converting to paid. The response:

Time-boxed trials replacing unlimited free tiers. 14-day full access beats "free forever with limits." It creates urgency and lets users experience the full product.
Usage caps replacing feature-gating. Instead of hiding features behind a paywall, let users access everything but cap volume. This shows the full value before asking for money.
Investment-based conversion. The best PLG products make users invest time and data before they pay. Once your workflow lives inside the product, switching costs are real. This is the core of friction maxxing: strategic friction that creates value, not frustration.

The pattern to watch: products that combine time-limited full access with usage-based pricing after the trial. You get the PLG adoption curve without the free-tier abuse.

Real Examples: Who's Getting It Right

Vercel went all-in on usage-based pricing. Build minutes, bandwidth, function invocations, and edge requests all have metered pricing. The result: revenue scales with customer success, not headcount.

Anthropic charges per token for API access. Pure usage-based. No seats, no tiers for the API product. Customers pay exactly for what they consume.

Notion runs a textbook hybrid model. Free for personal use (PLG adoption), per-seat for teams (revenue at scale), and enterprise sales for large organizations (high-margin deals with custom terms).

Linear still charges per seat, but it works because they've built enough strategic friction that each seat represents genuine, distinct value. Not every product needs to abandon per-seat. But you need a reason to keep it.

IdeaPlan uses a free tools + annual pro tier model. Free interactive tools drive adoption and email capture. The pro tier charges for research access, not seats. Value-based, not headcount-based.

Explore more SaaS product ideas in the Ideas Lab to see how different pricing models apply to different product categories.

PLG Pricing in the AI Era: Per-Seat Is Dead, Here's What Replaces It