Key Takeaways

• Published seat rates run from GitHub Team at "$4 USD per user/month" to Salesforce Agentforce 1 Sales at "$550 USD/User/Month", so the model fixes no price level.

• Twilio Flex publishes three seat definitions for one product: "$150 per named user", "$1 per active user hour", "$35 per monthly active user".

• Slack bills only members who acted within 28 days, and credits the rest back prorated.

• Google Workspace's annual plan lets you "reduce licenses only when renewing", pinning the contract floor at its high-water mark.

• Intercom prices a human at "$85 per seat/mo" and its Fin agent at "$0.99 per outcome", so the AI works without a seat.

What actually counts as a billable seat?

A seat is whatever the contract says it is, and the definitions in live use bill identical behavior differently. A provisioned account charges whether the person logs in or not, a named license can't float, an active-user window drops anyone dormant, and an hourly basis charges no license at all. Some participants never count at all: Notion states that "Guests are free of charge."

Published seat prices and the basis each one uses, captured 26 August 2026:

Twilio Flex proves the label hides the basis: one product, three prices. Seat counting is an entitlement problem before a billing one, so whoever owns feature gating decides who lands on the invoice.

What happens to the bill when seat counts change mid-cycle?

Additions prorate up on a daily rate, and reductions often don't prorate down at all. In our experience that asymmetry, not the headline rate, drives most renewal-call surprises.

Worked on Notion's $20 per member / month Business rate over a 30-day cycle:

1. Daily seat rate: $20 ÷ 30 = $0.6667 per seat per day.

2. Cycle opens at 40 seats: 40 x $20 = $800.00.

3. Four seats added on day 19, 11 billable days left: 4 x 11 x $0.6667 = $29.33.

4. Three seats removed on day 24, 6 unused days: credit of 3 x 6 x $0.6667 = $12.00.

5. Cycle total: $800.00 + $29.33 - $12.00 = $817.33.

Slack publishes the same arithmetic as "($8.75 ÷ 30 days) x (20 days) = $5.83" and issues the step 4 credit. Google Workspace's Annual/Fixed-Term Plan does the opposite: you "can reduce licenses only when renewing your plan at the end of the contract". Its Flexible Plan drops the ratchet.

That ratchet moves numbers on both sides:

• Customers keep paying for departed employees until renewal, so real cost per active user drifts above list.

• Vendors book contraction at renewal instead of the month it happened, distorting MRR movements.

• Seat counts creep upward across a term, since adding is instant and removing needs a negotiation.

Why do AI products strain per-seat pricing?

An AI agent does the work a seat used to justify while consuming inference cost no seat covers. As agents resolve more tickets, the buyer needs fewer seats and the vendor's cost per resolution holds, so revenue and cost move apart.

Intercom prices both halves on one page: a human on Advanced at "$85 per seat/mo", Fin at "$0.99 per outcome", charged once per conversation. Automate 2,000 conversations a month and that bill is roughly $1,980, for work that used to justify headcount.

Salesforce restructured around the gap, offering "consumption-based pricing, with Flex Credits or Conversations, or per-user licensing". Its published rates decouple the work from the seat:

• Flex Credits list at "$500 USD/Per 100k Credits", half a cent per credit.

• "Agentforce actions are 20 Flex Credits, while Agentforce Voice actions are 30 Flex Credits", so $0.10 and $0.15 per action.

• The Agentforce User License is "$5 USD/User/Month (Requires Flex Credits)", a seat that grants access and pays for none of the work.

The pressures that follow once agents carry real workload:

• Seat count stops tracking value: the better the agent works, the fewer seats the buyer needs.

• Cost per account stops being fixed, because inference scales with agent activity while the seat fee doesn't.

• Agent identities consume compute without matching a person, and seat-based systems have nowhere to put them, which is the gap AI agent pricing fills.

Related terms

Seat pricing rarely gets decided on its own, and these terms come up alongside it.

• Hybrid Pricing Model covers seats, usage, and credits landing on one invoice.

• Consumption-Based Pricing is the pay-for-what-you-use alternative weighed against seats.

• AI Agent Pricing sets out models for when the worker isn't a person.

• Feature Gating controls what a given seat tier can reach.

• MRR Movements tracks expansion and contraction seat changes create.

• Subscription Lifecycle covers the events where seat counts change.

FAQ

Is per-seat pricing the same as per-user pricing?

Yes, in almost every commercial context the two names describe the same model. Notion prices "per member / month", Intercom "per seat/mo", GitHub "per user/month", all for identical billing. What matters isn't the label, it's whether the seat means a provisioned account or an active one.

Is per-seat pricing still a good fit for collaboration software?

It fits where value scales with the number of people involved. Linear, Notion, and GitHub still price per user because a second person in the workspace creates real value for the first, and their rates sit in a narrow $4 to $20 band. It gets shaky when one person's usage runs an order of magnitude above another's.

How does per-seat pricing affect expansion revenue?

It ties expansion to the customer's hiring rate, which the vendor can't influence. Accounts grow when the buyer hires and stall when they freeze. Usage-priced vendors expand when the product gets used more, a lever product work can actually move.