Convert all three models to the same denominator — annual dollars at your actual volume — because the headline unit price is not comparable across them. Per-seat pricing bills named users. Usage-based pricing bills the work the software does, metered in credits or actions. Outcome-based pricing bills per result the vendor agrees happened: a resolved ticket, a qualified lead, a booked meeting. The seat line is the number you can forecast from your org chart; the meter is the number that decides your bill.
This is not a discount question. Usage-based is not “pay less” — it moves the cost driver off your headcount and onto your customers’ behavior, your inbound volume, or your data spend, none of which your budget owner controls. It is also not a token comparison. A credit is a vendor-defined currency with a vendor-set exchange rate, re-priceable without changing the dollar figure on your contract.
The four models in market
Per seat. Named user × price × 12. Intercom publishes $29/seat/mo (Essential), $85 (Advanced), and $132 (Expert). This is the only model where next year’s cost falls out of next year’s hiring plan.
Flat platform fee plus a credit allowance. Gumloop charges $37/mo for Pro, carrying 20,000 credits and unlimited seats, with no free tier behind it — only a 14-day trial. Seats stopped being the meter; run volume became it. Common Room sits at the enterprise end of the same shape with a published $2,500/mo floor.
Pure consumption. Clay ships unlimited seats on every plan and bills two separate meters: actions (running a table, calling an AI model, exporting data) and data credits (the email address or phone number itself, bought from its data partners). Paid plans start near $60/mo billed monthly, and Enterprise starts at 100,000+ data credits and 200,000+ actions a month.
Outcome, or per result. Intercom’s Fin bills from $0.99 per resolution and charges once per conversation even when it answers several questions. HubSpot’s Customer Agent bills $0.50 per resolved conversation — 50 credits at HubSpot’s $0.010/credit rate — after moving off $1.00 per conversation on 2026-04-14; its Prospecting Agent bills $1.00 per lead recommended for outreach.
Every vendor named above actually bills hybrid: a seat or platform fee that gates access, plus a meter that produces the bill. Lindy is the clearest specimen — $29.99, $99.99, and $199.99 per user per month carrying 3,000, 15,000, and 35,000 credits respectively, pooled across the workspace, with credit-consuming actions pausing when the pool empties. You buy seats and you buy credits, and the second number is the one that moves.
Modeling true annual cost
Run the two lines separately, then add them:
Annual = (seats × seat price × 12) + (monthly volume × unit price × 12)Take a 12-person support team on Intercom Advanced handling 4,000 conversations a month, with Fin resolving half:
- Seats: 12 × $85 × 12 = $12,240
- Fin: 2,000 × $0.99 × 12 = $23,760
- Total: $36,000, two-thirds of it on the meter
Doubling support volume adds roughly $23,700 a year while the seat line does not move at all. That inversion is the point. Under per-seat pricing your cost driver sits inside your hiring plan, where you control it. Under usage pricing it sits inside your customers’ behavior, where you do not.
Then there is the comparison teams get wrong. A $0.99 resolution looks expensive next to an $85 seat until you notice those are not the same denominator. Divide the meter price by your fully-loaded cost per contact — salary, benefits, tooling, management overhead — never by your software seat price. The meter is priced against labor, not against licenses. Benchmarking it against a license line makes every usage-based product look like a markup and kills pilots that would have paid back.
Seven traps, each with a guard
Credit allowances are not additive. Holding several HubSpot subscriptions gets you the highest single allotment, not the sum — credits are not combined across subscriptions. Guard: ask which single allowance applies to the whole account, and get that number into the order form rather than the sales deck.
The default overage is a tier upgrade, not a surcharge. HubSpot’s default behavior on exceeding your credit limit is an automatic upgrade to the next credit tier for the remainder of the contract term. The opt-in alternative, pay-as-you-go at $0.010 per credit, resets to your original limit the following cycle. Guard: pick the overage behavior before go-live and set an alert at 80% of allowance.
One product, two meters. Clay’s actions reset monthly while its data credits roll over up to 2× the monthly amount on paid plans. Forecasting one and not the other under-buys whichever does not carry. Guard: ask which meter each workflow step hits, and forecast both.
Unlimited seats is a pricing signal, not a discount. Gumloop and Clay both include unlimited seats because the bill moved to volume. Rolling the tool out to 40 people costs nothing in licenses and everything in runs. Guard: model the volume 40 people generate before treating the seat count as a saving.
The vendor defines the outcome. HubSpot counts a resolution when the agent shares a source or performs an action with no human handoff inside 72 hours. Fin counts one when the customer confirms resolution, stops asking for help, or Fin completes a workflow. Identical traffic, two different bills. Guard: get the definition in writing, ask who adjudicates a disputed outcome, and require an exportable per-outcome log you can audit against your own helpdesk data.
A unit-price cut can be a price increase. HubSpot’s Customer Agent went from $1.00 per conversation to $0.50 per resolved conversation — a 50% cut on a base that changed underneath it, against a reported 65% resolution rate across 8,000+ customers. Guard: recompute against last quarter’s actual traffic under both definitions before accepting the headline.
No rollover means you size for peak, not average. HubSpot credits expire at the end of each usage period; Lindy’s refresh monthly with no carry-forward. Sizing on an average month buys a shortfall in every above-average one. Guard: size the allowance on your busiest month, or negotiate rollover explicitly.
Which model to take
Take per-seat when volume per person is stable and headcount is the variable you control. You are buying forecastability, and it is worth a premium at budget time — a finance team that can predict the line will approve a higher one.
Take usage or outcome when volume is spiky or seasonal, when you are running a time-boxed pilot, or when the tool replaces labor instead of assisting it. Paying $0.99 per resolution beats buying twelve months of seats to survive a Q4 spike.
Refuse hybrid contracts that meter something you cannot instrument. If you cannot count the billable unit in a system you own — your helpdesk, your CRM, your ATS — you cannot audit the invoice, and disputes become the vendor’s word against your recollection.
When you cannot decide, model twelve months at your peak month × 12 rather than your average month × 12, and buy the model that wins under that number. If a vendor’s pricing only works at average volume, the vendor has moved seasonality risk onto your budget. For the prior question of whether the thing you are metering is an agent or a feature bundle, see what makes an AI agent for ops; for the CRM-side version of the same buying decision, see agentic CRM.