Per-Seat, Usage or Hybrid Pricing: How to Compare AI SaaS Quotes
Two AI SaaS quotes are rarely priced in the same unit. One vendor charges per user, another per credit, a third per resolved conversation. To compare them, convert every quote into the same question: what would this cost us in a month of our own real usage, and what is the worst month? This guide shows the method, using three public price pages as worked examples. Prices below were read on 12 November 2026 and change often, so check each page before you decide.
Key facts
- Intercom's pricing page lists its Fin AI agent from $0.99 per Fin outcome, with a minimum monthly commitment for use with an existing helpdesk (the FAQ gives 50 outcomes as an example).
- GitHub's Copilot plans page states "1 AI credit = $0.01 USD" and lets paid plans keep going past the monthly allowance by setting a dollar budget for extra usage.
- Salesforce lists Agentforce Flex Credits at $500 per 100k credits, with each Agentforce action costing 20 Flex Credits, and a separate Conversations option at $2 per conversation.
- Oliver Wyman's buyer research (dated October 2026 in its URL) reports that roughly nine in 10 buyers worry about unpredictable costs under usage pricing.
The three models in plain terms
Per-seat: you pay for each named user. The bill is predictable but does not follow how much work the AI does.
Usage: you pay for what the software consumes, counted in credits, tokens, actions or outcomes. The bill tracks activity, so it can outrun your budget.
Hybrid: a seat fee or platform fee for access, plus a usage allowance and a price for going over. Oliver Wyman's analysts describe hybrid models as combining seat minimums or platform fees for recurring revenue with usage mechanisms that grow with adoption. Our earlier piece on why per-seat pricing is breaking covers why vendors are moving this way.
Step 1: find the billing unit and what counts as one
Every quote hides a definition. Intercom's page says a Fin outcome is counted when a customer confirms the issue is resolved, when they do not ask for more help after Fin responds, or when Fin completes a workflow, including handoffs, and that you are charged once per conversation. That is not the same as "a resolved ticket", and the gap affects your bill.
Salesforce shows why the unit matters. Its page lists two ways to pay for the same agent: Flex Credits, or Conversations at $2 each. Our own arithmetic: at $500 per 100,000 credits, one credit costs $0.005, so a 20-credit action costs $0.10. A conversation that triggers more than 20 actions would cost more under Flex Credits than the $2 flat rate, and one with fewer would cost less. The page says the two cannot be used in the same org, so you choose one.
Step 2: price your own volume in each model
Take a month you know well and count the units. Then multiply.
- Intercom Fin, 1,000 outcomes at the listed starting rate: $990 (our own arithmetic, before any seat costs for the helpdesk and assuming the starting rate applies to your volume).
- Intercom helpdesk seats, 10 Advanced seats at the listed $85 per seat per month: $850 (our own arithmetic). The page shows no separate annual and monthly price for Advanced, so confirm that in writing. Intercom lists Fin on an existing helpdesk with no seat costs, so Advanced seats apply only if you also take its helpdesk.
- GitHub Copilot Pro lists $10 per user per month with $15 of monthly credits. At $0.01 per credit, that is 1,500 credits (our own arithmetic).
These figures cover different jobs, so do not add them up. They show the pattern: a flat seat bill in one tool, a bill that scales with tickets in another.
Step 3: ask what happens at the limit
For any usage or hybrid plan, get the answer to four questions in writing.
- When the allowance runs out, does the service stop, slow down or keep billing?
- Is there a spending cap you control? GitHub's page describes a dollar budget for additional usage on paid plans, and says that for Business and Enterprise, admins decide whether extra paid usage is allowed. That is the kind of control to look for.
- Is there a minimum commitment? Intercom's FAQ mentions one for Fin with an existing helpdesk.
- Do unused credits expire at month end or the contract year?
Step 4: stress-test the bill
Oliver Wyman recommends that vendors stress-test margins at three and five times current usage. Buyers can use the same idea from their side. Multiply your expected volume by three and by five, recompute each quote, and see which one still fits your budget. A seat plan will barely move. A per-outcome plan will not. A flat price protects you if usage explodes, while a usage price protects you if adoption stays low.
The same research found that nearly two-thirds of current customers view usage pricing favorably, so the aim is not to avoid it. Take it with a cap, a clear unit and a tested number.
Step 5: compare on cost per result
Divide each monthly figure by something you value: tickets handled, documents processed, pull requests merged. A cheaper quote that needs more human review can cost more per finished task. Pilot identical work before you commit, and use the AI tool directory to shortlist candidates. More buyer guides are on our articles page.
What to do this week
- Pull one recent month of real volume for the task.
- Ask each vendor to define its billing unit.
- Price your volume at 1x, 3x and 5x in every quote.
- Get the cap, minimum and expiry terms in writing.
- Divide each total by finished tasks, not by licenses.
Questions people ask
Is per-seat pricing always cheaper to predict?
It is easier to forecast, but it ignores how much AI work you use. Oliver Wyman notes that automation could shrink seat counts, which is why vendors add usage components.
What is a hybrid pricing model?
A base fee for access, usually per seat or per platform, plus a usage mechanism such as credits that you pay for as adoption grows.
How do I make a usage quote predictable?
Get a spending cap and a defined billing unit in writing, then test the bill at higher volumes. Oliver Wyman's authors recommend caps, true-ups and contractual flexibility.
Sources