Mostly with a seat that includes an expiring pool of AI credits, plus metered overage. Outcome pricing is real but narrow.
The map
AI Momentum Partners mapped 41 software brands in October 2026 onto five models. 21 sell a seat plus a metered AI allowance. 9 bill purely on usage. 5 bundle AI into the plan. 2 sell a per-seat add-on. 4 charge per outcome, and all four (Fin at $0.99, HubSpot at $0.50, Zendesk from $1.50, and Sierra per task) price a resolved support conversation.
The map is a snapshot of traffic, not five stable options. Over the last twelve months:
| Company | From | To |
|---|---|---|
| Oracle | AI agents at no extra cost (2025) | An included monthly AI Unit allowance, then a meter (Release 26C, per partners) |
| GitHub | $0.04 per premium request (2025) | Token-metered AI Credits (June 1, 2026) |
| HubSpot | $1 per conversation in credits (2025) | 50 credits per resolved conversation (April 14, 2026) |
| Datadog | $500 for 20 investigations | $500 for 500 AI Credits (June 2026, per PricingSaaS) |
| Atlassian | Rovo included in paid plans | Overage at $0.01 per credit from December 3, 2026 |
Bundles are adding meters, outcomes are being billed in credits, and nobody is leaving the middle.
Why credits
A vendor selling AI manages two prices moving in opposite directions: cost per token falls roughly 10x a year, while tokens per agentic task run 5 to 30 times a chat. A credit puts a rate card between the customer and both. The rate card can change in a release note without the contract changing. OpenAI cut ChatGPT voice from 5 credits a minute to 1.25 that way in September 2026.
The three numbers that are the real price
- Exchange rate. Dollars per credit. Salesforce ($0.005), Microsoft ($0.01), and Atlassian ($0.01) publish one; OpenAI, Workday, SAP, and UiPath don't.
- Burn rate. Credits per action, set and changed by the vendor.
- Expiry. Most pools reset monthly or after 12 months with no rollover.
What I learned from the vendor side
Expiring credits create revenue that looks healthy because the customer isn't using what they bought. In my own move of a $40M ARR business from per-seat to hybrid outcome pricing, 12 accounts were classified as on plan. They paid their committed minimums while using less than the minimum. Nobody escalated, and we lost two of them at the sunset. Salesforce's CFO said more than half of one quarter's Agentforce bookings came from customers "refilling the tanks." A refill is the new renewal, and a full tank at renewal is the hardest conversation in it.
My rule if you set the price: publish all three numbers, show the burn before an action runs, and track expired credits as a warning light, not as planned revenue.
The full analysis, with every column and the five things the chart gets wrong, is in 41 Brands Price AI Five Ways. The Credit Is the New Seat. It sits under SaaS to AI Business Models.