CRMs now charge in four ways: per seat, by usage credits, by outcome, or per workspace. Per-seat pricing is under pressure because AI agents do more of the work people used to log in for, so the big vendors are adding credits and outcome-based charges on top of seats rather than replacing them.
Here is how each model works, a real example of each as of 18 November 2026, and what each does to a growing team's bill.
The four models
| Model | You pay for | Real example | Grows when |
|---|---|---|---|
| Per seat | Each person who logs in | HubSpot's Core Seats; Salesforce's per-user Agentforce licences | You hire |
| Usage credits | Work done, drawn from a pool | HubSpot Credits; Salesforce Flex Credits | You or your agents do more |
| Outcomes | A result, such as a resolved conversation | HubSpot's Customer Agent | The AI succeeds more often |
| Per workspace | The account, with limits on capacity | Empiraa Signal, unlimited users | You need more capacity |
Per seat
The classic model. HubSpot's Core Seat is the one that lets someone edit records, and view-only seats are included at no extra cost, per HubSpot's seats FAQ. Salesforce also sells per-user Agentforce licences with unmetered use for employees, according to Salesforce.
Good for: predictable bills when headcount is stable. Bad for: growing teams, because every hire raises the bill, and it quietly discourages people from using the CRM at all. More in how per-seat pricing changes how teams work.
Usage credits
You buy a pool and AI work draws it down. HubSpot introduced Credits on 2 June 2025, and in October 2026 added default spend caps, caps per feature and a pay-as-you-go option charged only when AI work is completed (CMSWire).
Salesforce announced Flex Credits in May 2025: $0.10 per action, replacing $2 per conversation (Salesforce). Analysts at Futurum reported packs of 100,000 credits for $500, and an agreement that lets customers convert user licences into credits.
Good for: paying only for what agents actually do. Bad for: forecasting, because usage can spike, and a pre-committed pool you do not use is still paid for.
Outcomes
You pay when the AI achieves something. HubSpot moved its Customer Agent and Prospecting Agent to outcome-based pricing in 2026: a Customer Agent conversation only uses credits once it is judged resolved after a 72-hour window (Metronome).
Good for: aligning cost with value. Bad for: understanding the bill, because the vendor defines what counts as an outcome.
Per workspace
You pay for the account, and capacity is metered on the workspace rather than per person. Empiraa Signal works this way: every plan has unlimited users, and limits sit on things like records held and prospects found each month.
Good for: growing teams, and teams moving work between people and agents, because neither changes the bill. Bad for: very large organisations that need per-team cost allocation.
How to choose
- Write down your headcount today and in 12 months.
- Estimate how much of the admin agents will do by then.
- Price each model at both points.
- Ask each vendor what happens when you exceed a limit or a credit pool.
If headcount grows faster than usage, a per-workspace price usually wins. If usage is low and occasional, credits can. What this means for HubSpot customers specifically is in will HubSpot's pricing change after the refounding.
Sources: HubSpot seats migration FAQ (read 18 November 2026), CMSWire (15 October 2026), Metronome (2 June 2026), Salesforce on Flex Credits (23 May 2025), Futurum (19 May 2025).
Related reading: HubSpot credits and Breeze agents: what you actually pay for, and what is a headless CRM.



