Definition
The different ways an AI vendor charges you: a flat fee per user, charges that scale with usage, or a fee tied to the results delivered.
At a glance
- Four core models: per-seat (flat fee per user), usage-based (pay per token, call, or action), credit-based (prepay a pool you draw down), and outcome-based (pay only on a result)[1].
- Per-seat is fading because AI agents do work with nobody logged in, so seat counts no longer track value or cost.
- Hybrid (fixed base plus variable charges) is now the dominant model: a predictable floor with room to scale[5].
How the models differ
Per-seat: $500/month per attorney. Usage-based: pay per customer review the AI analyzes. Credit-based: buy 10,000 credits, spend 50 per task. Outcome-based: a recruiter pays only when a surfaced candidate gets hired. Every AI query burns real compute, so vendors run 50-60% gross margins versus 80-90% for old SaaS[2] pushing bills toward consumption.
What it means for your budget
Match the model to your priority. Want predictable costs? Choose a hybrid with a fixed base[5]. Swinging usage? Usage-based can be cheaper but harder to forecast. Care most about results? Outcome pricing aligns the bill with value: Intercom’s Fin charges $0.99 per resolution and nothing if it hands off[4]. Salesforce kept rechanging Agentforce after per-conversation bills proved unpredictable[3]. Before signing, pin down the billable unit and ask for a cap.
Bottom line
There is no single right model, only the one fitting how you buy increasingly a hybrid base plus a charge that tracks the value delivered.