Unpredictable AI bills: why Salesforce buyers want seats again
Yuri VonchitzkiUpdated 4 min readFor two years, AI vendors pushed consumption-based pricing. They wanted us to pay per token, per task or per credit. In December 2025, Salesforce told investors that seat-based deals for Agentforce and Data Cloud had doubled year on year in its third fiscal quarter. Customers asked for flat-fee contracts to avoid paying by consumption.
Businesses do not want variable AI bills. They want a number they can put in a budget. Seat-based deals doubling at Salesforce shows that the market is turning toward flat-rate AI pricing. If you buy AI tools, lock in flat-rate or seat-based contracts now, before vendors see how much compute you will use.
What is consumption-based AI pricing?
Consumption-based pricing (also called pay-per-use) charges you for each unit of AI work: each token, each task or each credit. The bill changes every month with your usage. Seat-based pricing charges a fixed fee for each user. A flat-rate contract charges one fixed fee, whatever the usage.
| Pricing model | What you pay for | Monthly bill |
|---|---|---|
| Consumption-based (pay-per-use) | Tokens, tasks or credits | Changes with usage |
| Seat-based | Each user | Fixed per user |
| Flat-rate | Access, with no usage meter | Fixed |
Why are Salesforce customers choosing seat-based Agentforce pricing?
Businesses hate unpredictable bills, and I think that explains the shift. Agentforce is the Salesforce platform for AI agents. Salesforce first pushed consumption-based pricing for Agentforce: the September 2024 launch priced it at $2 per conversation. In December 2025, seat-based deals were growing again.
The problem is budgeting. A CFO cannot budget for a tool that costs $100 this month and $1,000 next month. With pay-per-use, that jump can happen only because an agent got stuck in a loop. Nobody decided to spend more money. The bill went up without a business reason.
The flat-rate era of AI
The AI market is entering a flat-rate era. Businesses will happily pay a premium for predictability. I expect unlimited flat-rate AI contracts to become the premium standard for enterprise software in 2026.
Think of it as an all-you-can-eat buffet. The buffet is open today, but it will not stay cheap forever. Vendors do not yet know how much compute their customers will use. When they find out, the price will go up.
How to buy AI tools with predictable costs
I advise business owners to choose predictable pricing now. When you buy an AI tool, use these steps:
- Ask for a flat-rate or seat-based option. Some vendors still lead with pay-per-use. Ask for a fixed price before you sign.
- Check the loop risk. On a pay-per-use plan, ask what happens to the bill when an agent gets stuck in a loop. If the answer is "you pay for it", your cost is not under control.
- Pay a premium for predictability. A slightly higher fixed price is often better than a lower price that can grow ten times in one month.
- Lock in the flat rate early. Sign flat-rate terms before vendors see your real compute usage.
Before you sign, model the cost against the hours the tool saves. The ROI calculator helps you check if the tool pays for itself. Also check what you get for the money: many agents on the market are thin wrappers, as I explain in how to test AI agents before you buy. Falling model costs change the math too; see why cheaper AI reasoning makes agent workflows pay off.
The bottom line
Predictable AI costs are a business decision, not a technical one. The growth of seat-based Agentforce deals shows what buyers want: a bill they can plan for. Lock in flat-rate pricing while vendors still offer it at a good price.
Build a calmer business,Yuri
Frequently asked questions
Why are Salesforce customers choosing seat-based Agentforce pricing?
In December 2025, Salesforce said that seat-based deals for Agentforce and Data Cloud had doubled year on year in its third fiscal quarter, and that customers asked for flat-fee contracts to avoid paying by consumption. The main reason is that businesses hate unpredictable bills. A CFO cannot budget for an AI tool that costs $100 one month and $1,000 the next month because an agent got stuck in a loop.
What is the difference between consumption-based and seat-based AI pricing?
Consumption-based AI pricing charges per token, per task or per credit, so the bill changes with usage. Seat-based pricing charges a fixed fee per user, so the bill stays the same each month. Seat-based and flat-rate contracts give a business a predictable cost that a CFO can put in a budget.
Should a small business choose flat-rate AI pricing over pay-per-use?
Yes, in most cases a business should prefer flat-rate AI pricing when a vendor offers it. Pay-per-use bills can jump when an AI agent gets stuck in a loop. Flat-rate contracts make the cost predictable, and businesses will pay a premium for that predictability. Lock in flat-rate terms before vendors see how much compute you use.
Will flat-rate AI contracts stay cheap?
Probably not. In December 2025, Yuri Vonchitzki of Ninjabot expected unlimited flat-rate AI contracts to become the premium standard for enterprise software in 2026. His view is that the flat-rate offer is like an all-you-can-eat buffet that will not stay cheap forever. Vendors do not know yet how much compute each customer will use. Locking in flat-rate terms early protects a business from that price increase.