Sunil Venkataram
Back to garden
•
🌰 seedling
Seat-Based to Token-Based SaaS Pricing Transition
type/notesaaspricingaibusiness-model

Seat-Based to Token-Based SaaS Pricing Transition

The forcing function

The shift is driven by customer-side budget dynamics, not by a vendor-side pricing fad. When a customer deploys AI to make knowledge workers 30% more productive, the most visible line item they can cut is seats on tools those knowledge workers used. Seats go away or stagnate. Meanwhile, new budget — for agents running in the background, for automation infrastructure, for outcome-based vendor contracts — grows rapidly.

A vendor with a 100% seat-based business model is simultaneously:

  • Losing existing ACV as customers reduce seat counts
  • Invisible to the growing budget categories their customers are spending on
  • Unable to reprice their existing product into consumption terms without cannibalizing their own GAAP metrics

The math is unkind to the middle. Seat growth for vendors whose products are “things knowledge workers actively use” will be negative or flat. The growth is in a different part of the customer’s budget, and only vendors who can price in that part get the check.

What “token-based” actually means

The term is loose. A more precise way to think about the transition: pricing should be based on what a vendor’s product does on behalf of the customer, not on how many humans touch it. That can take several shapes:

  • Per-token / per-call consumption pricing — simplest version, easy to meter, works when the product is substantially a wrapper around foundation model calls
  • Per-workflow completion pricing — vendor charges when a workflow (claim processed, document reviewed, incident resolved, appointment scheduled) is successfully completed
  • Outcome pricing — vendor charges a share of the value created (dollars of admin cost avoided, hours saved, revenue generated)
  • Hybrid with a seat floor — small seat fee per licensed user plus consumption charges on top, to capture both the shrinking seat budget and the growing consumption budget

The right choice depends on the vendor’s workflow. A coding assistant can meter tokens directly. A healthcare workflow vendor might meter per completed referral. A legal AI might meter per contract reviewed. The common thread is that the unit of billing tracks the unit of value delivered, not the unit of human access.

The agent-can-pay test

A useful filter from David George in the a16z two-paths essay:

If an agent can’t consume and pay for your product autonomously, you probably aren’t there yet.

The test cuts through a lot of hand-waving. A lot of “AI-native pricing” announcements in 2025-2026 are seat pricing with an AI feature bundled into the seat cost — which is still seat pricing. The real test: could an autonomous agent, acting on behalf of a human customer, discover this product, evaluate its output, decide to use it, and pay for it without a human intervening to approve the purchase? If no, the pricing model is still human-mediated and the vendor is still in the old world.

Agent-to-agent commerce is going to be a real channel for at least a subset of B2B software, and the vendors whose billing systems can accept agent payments will own early market share in that channel.

The internal transition pain

Moving from seat to consumption is not a clean pivot. It creates real short-term damage:

  • ACV compression in the first 12 months. Consumption pricing almost always starts below the equivalent seat price, and existing seat customers will migrate at the lower entry. This tanks the reported revenue growth rate for several quarters.
  • Sales comp redesign. AEs compensated on seat ACV will resist the change, and replacing sales comp plans is the single hardest internal org change a software company can attempt.
  • Forecasting gets harder. Consumption revenue is lumpier and less predictable than seat revenue. CFOs and boards hate this.
  • Churn semantics break. “Retention” in a seat world is whether a customer renewed their licenses. “Retention” in a consumption world is whether the customer’s usage is growing, flat, or shrinking — and “shrinking but not churned” is a category that doesn’t exist in seat-world dashboards.

Most incumbents who talk about the transition aren’t actually willing to accept 12-18 months of ugly optics to execute it. That’s why the transition will mostly be led by new entrants who never had to unlearn seat pricing.

Where seat pricing survives

Not every software category is going to fully abandon seats. Seat pricing will persist in:

  • Products where the human is the actual consumer of value, not a proxy for it (e.g., creative tools, learning platforms, communication tools where the social graph matters)
  • Highly regulated categories where audit and compliance are tied to named-user provisioning
  • Categories where the seat count is actually the usage signal (a video conferencing product billed per seat is approximately billing per meeting participant, which is a reasonable unit of value)
  • SMB segments where consumption pricing is too complex for the buyer to understand or forecast

The transition is not 100% — it’s a reallocation of where the growth sits. Vendors who stay in seat-based categories will be fine if those categories are the ones where the unit of value really is human access. Vendors whose seat pricing is just a billing convention inherited from the SaaS era will get squeezed.

Diagnostic questions

For any software vendor or investment thesis:

  • What fraction of customer value is delivered autonomously vs. by humans actively using the product? High autonomous share = strong case for consumption pricing.
  • If AI doubles customer knowledge-worker productivity, does that grow or shrink the ACV? If it shrinks ACV, the pricing model is short the transition.
  • Can an API caller consume the product without a named human user in the contract? If no, the product is locked to the human-seat world.
  • What percentage of new customer budget is flowing to the vendor’s billing unit of measurement? Seat budgets are flat or shrinking in most enterprise categories; consumption and outcome budgets are growing.
Backlinks (3)
Connected Notes