Active-User Pricing: Examples & Companies

What is it

Active-User Pricing is a billing unit where customers are charged per monthly or daily active user rather than per provisioned seat. It is the seat model with the waste squeezed out: nobody pays for the licenses sitting unused after a big-bang rollout, because the meter counts engagement, not entitlement.

Glean is the corpus’s clearest enterprise specimen. Its confirmed Enterprise Flex model licenses per-user seats “to every employee,” but the intensity of how each employee uses premium AI is absorbed by a pooled layer of pay-per-use FlexCredits — a Thinking Mode query on premium models runs roughly 35 to 120 credits, slide generation 45 to 142. The seat aligns the contract with headcount; the credit pool absorbs the variance in who is actually active and how hard. That two-layer design is the recurring shape of the unit in this corpus.

The contact-center pair shows the unit’s enterprise ancestry. Observe.AI licenses per agent — third-party sources cite roughly $69/agent/month for a single module via AWS Marketplace, with a ~100-seat minimum and annual commitment — while Uniphore blends per-agent licensing with a per-interaction consumption rider (indicative third-party estimates put the agent component near $35/agent). In both cases the “active user” is a deployed call-center agent whose headcount the buyer already tracks daily, so the billing unit maps onto an operational metric the customer already trusts.

How it works

Active-user pricing swaps the denominator. Instead of billing rate x provisioned seats, the vendor bills rate x users active in the period — and then bounds the resulting variance with a second lever, because a naked fluctuating denominator is unplannable for both sides. The corpus shows five levers that recur:

Worked example — what inactivity is worth. Take a 1,000-employee company rolling out an enterprise assistant where 320 people actually use it in a given month. At a representative enterprise rate of ~$45/user/month (the third-party figure trackers report for Glean), provisioned-seat billing charges for all 1,000 users — $45,000. Active-user billing charges for the 320 who showed up — $14,400. The ~$30,600 gap is pure adoption risk, transferred from buyer to vendor — and the reason the floor in the table exists is to keep that transfer from running to zero. The usage-metric guide treats this as the canonical case of choosing the metric the buyer already audits: every company already tracks its own active headcount.

Companies using this

16 in-corpus companies bill on active or deployed users, clustering into five uses of the unit: enterprise assistants (Glean, Sana AI), contact-center/voice AI (Observe.AI, Uniphore), revenue intelligence (Clari, Gong), compliance-bound healthcare and legal (Abridge, Ambience Healthcare, Suki AI, Thomson Reuters CoCounsel), and the consumer edge where “active user” collapses to a single subscriber (Pi, Character.AI, the discontinued Phind). The table below sorts by pricing model, billing units, and free-tier availability.

Patterns observed

The unit never travels alone. Every enterprise vendor here pairs the active-user meter with a second mechanism that bounds variance — a pooled credit layer (Glean’s FlexCredits), a per-interaction consumption rider (Uniphore), or an explicit seat floor (Observe.AI, Sana AI). Pure presence-based billing — pay only when someone shows up, with no floor, pool, or cap — exists nowhere in this corpus, because neither side can plan around a naked fluctuating denominator.

The unit thrives where value is inseparable from individual participation. Revenue-intelligence platforms like Clari and Gong bill per user because the product’s value is anchored to each individual rep’s pipeline — a rep who never logs in generates no AI analysis and no bill. Gong even qualifies prospects into team-size brackets (1–50, 51–1,000, 1,001–9,999, 10,000+) before quoting, layering a platform fee on top of per-user licenses; Clari differentiates by charging no separate platform fee. In both cases the buyer’s own CRM already confirms which reps are active, which makes the per-user fee easy to defend in procurement.

Vertical compliance makes the per-user unit non-negotiable. Healthcare AI vendors — Abridge, Ambience Healthcare, and Suki AI — all license per clinician (PEPM-style) because the HIPAA compliance boundary is naturally per-provider: the audit trail, the data-access controls, and the consent framework all run to the individual clinician, so the billing unit mirrors the compliance unit. Abridge scales its per-clinician contracts by encounter volume and attached modules (Clinician, Nursing, Revenue Cycle, CDS); Ambience adds a one-time Epic-integration implementation fee. Legal AI (Thomson Reuters CoCounsel) follows the same logic against a licensed-attorney headcount its buyers already audit.

At the edges, the unit degrades into a subscription or a quota axis. On the consumer boundary, Pi never monetized at all — rate limits replace dollars. And where the seller is itself a billing platform, “active user” becomes one entitlement axis among events and transactions rather than the sole meter — a segmentation dimension inside a richer pricing surface (see the Schematic variant below).

Counterexamples & variants

Character.AI marks the consumer boundary where the unit stops being a meter. With no organization in the loop, “active user” collapses into the subscriber themselves — a $9.99/month (or $94.99/year) c.ai+ plan where activity gates experience (faster responses, no ads) rather than the bill. No headcount denominator, no floor, no pooled layer. Against a base that once peaked near 28 million monthly active users, “active user” is a retention metric here, not a line item.

Phind is the cautionary variant. It sold a flat ~$20/month Pro tier and ~$40/user Business tier with a daily cap, capturing nothing extra from heavy users — and shut down in January 2026. A flat fee inside a user tier never distinguishes engaged from dormant users at the price level, so cost per active user climbed as the denominator shrank with no lever to compensate: the failure of flat freemium under cost pressure, not of user-based billing itself.

Granola is the “priced per seat, not per active user” foil. Its Business ($14/user/month) and Enterprise ($35/user/month) plans are flat per-seat with no usage metering and no annual discount; the only usage limit is the 25-meeting lifetime cap on the free tier. It qualifies for this page on a billing dimension, but its paid model deliberately chooses provisioned seats for predictability.

Schematic is the structural variant from the platform side. Its product is runtime entitlements, so “active user” (its “monetized subscriptions” concept — Free at 10 subs, Growth at $200/month for 100 subs) is one axis alongside events and transactions, not the primary denominator — closer to a quota axis than a pure per-user bill.

What this means for buyers vs vendors

For buyers

Get the activity definition in writing — daily vs monthly active, and what action counts — because it moves the denominator more than the rate does. Then check what sits underneath: a floor can make the “pay for actual usage” story mostly notional at small scale, so price your own expected active headcount against it. Bring your own engagement data rather than trusting the vendor’s meter alone, and negotiate the activity definition, the floor, and overage handling in the initial contract — a well-drafted active-user clause then lets you scale users mid-cycle without reopening the base rate. In healthcare and legal, expect the per-clinician or per-attorney rate to sit above general SaaS pricing; the premium buys a billing unit your governance team can audit against a credentialing roster. And check for stacked platform fees, which can swing the total independent of the per-user rate.

For vendors

Active-user pricing is an adoption-risk transfer — take it on only when engagement is genuinely sticky and auditable, then bound it with the mechanisms this corpus validates: a seat floor under your downside, a pooled credit layer where intensity varies more than headcount (Glean’s FlexCredits are the template), and a consumption rider (Uniphore’s per-interaction component) where presence under-captures value. Set the floor in the low hundreds when the product is workflow-native and the buyer has committed to a deployment program — too low leaves revenue exposed to shortfalls unrelated to product quality; too high erases the fairness argument and loses to seat-based competitors. Define activity events precisely in your metering pipeline from day one: an active-user invoice the customer’s own analytics can’t reproduce is a dispute, not a bill. In compliance-bound verticals the per-user unit is accepted in procurement without explanation, and the compliance architecture behind it (access controls, audit logging, credentialing integrations) is real cost — price to reflect it rather than discounting to a horizontal benchmark. Above all, make the price tier or a pooled layer track value density; a flat fee inside a user tier inherits adoption risk with no way to price it.

Explore this theme in the knowledge graph →