Most help desks now send two bills in one: a per-seat charge for every agent who logs in, and a usage charge for each conversation the AI agent handles. Seats grow with headcount; usage grows with customer volume. Which one dominates depends on your team size and how much the AI does, and the exact definition of the usage unit can change the usage bill more than the headline price.
The short version
- Seat charges scale with the number of agents, and per-agent AI add-ons scale with them.
- Usage charges scale with conversations the AI handles, whatever your headcount.
- A resolution, an outcome and a session are different units. A session can be charged even when nothing is resolved.
- Free allowances differ: some are per seat per month, some once per account.
- Run your own numbers in the true-cost calculator rather than comparing headline prices.
How a seat bill behaves
A seat is usually any person who can reply to customers, including admins; customers are never billed. The bill rises in steps as you hire, and falls only when you remove seats, often at renewal on annual contracts. Watch for three things: minimum seat counts (Help Scout's Pro plan has a 10-user minimum), maximums that force an upgrade (its Plus plan stops at 50 users), and per-agent add-ons that must cover every seat. Zendesk Copilot costs $50 per agent a month and cannot be bought for only some agents.
How a usage bill behaves
A usage bill rises with the number of conversations the AI handles, not with headcount. That makes it attractive to a small team with high volume, and a risk for any team whose traffic spikes, for example after a product incident or during a sales season. Ask whether you can set a monthly cap; Help Scout documents one for AI Answers.
Resolution, outcome, session: the definitions
| Unit | What is counted | Example |
|---|---|---|
| Resolution | A conversation the AI closes without the customer asking for a person or needing more help | Help Scout $0.75 per resolution, at most once per conversation |
| Verified or automated resolution | A resolution the vendor's own system judges to be resolved | Zendesk lists $1.50 committed or $2 pay-as-you-go |
| Outcome | A broader unit Intercom moved to from resolutions; defined in its help center | Intercom Fin $0.99 per outcome |
| Session | A conversation window with the AI agent, charged whether or not the issue is solved | Freshdesk $49 per 100 sessions |
| Tokens and credits | The amount of text the model processes | Zoho Desk: 30 million tokens free a month, credit price not published |
Why this matters: a session is consumed when the AI starts working, so at the same volume of solved questions a session-priced product needs more units than a resolution-priced one. Freshdesk defines a session as 24 hours on chat and 72 hours on email. A resolution is only billed when the AI succeeds, but the vendor's software decides what success is. Put the definition in the contract.
A worked comparison at 10 agents
Take our standard mid-size scenario: 10 agents on each vendor's mid tier, about 250 tickets per agent a month, and one ticket in five resolved by AI, so 500 AI resolutions a month, with agent assist switched on for every agent. The bill for each product is built like this:
- Help Scout Plus: 10 × $45 in seats; AI Drafts is included, so no assist add-on; 500 × $0.75 for AI Answers.
- Zendesk Suite Professional: 10 × $115 in seats; 10 × $50 for Copilot; 10 resolutions per agent a month are included, so 400 are billed at $1.50.
- Intercom Advanced: 10 × $85 in seats; 10 × $29 for Copilot; 500 × $0.99 for Fin.
- Freshdesk Pro: 10 × $55 in seats; 10 × $29 for Freddy AI Copilot; five packs at $49 if every session ends in a resolution, more if some do not.
- Zoho Desk Professional: 10 × $23 in seats; Zia is included; AI agent usage beyond the free tokens has no public price, so it cannot be totaled.
At this size the seat line is the larger part of every bill, and the per-agent assist add-on can move a product's rank more than its AI agent price. At 50 agents the seat line grows fivefold, and usage grows with it only if ticket volume does. With the same headcount and more automation, the balance shifts toward the usage line. The calculator runs this arithmetic for every product, and the pricing pages show the totals at 3, 10 and 50 agents.
Which model suits which team
- Small team, high volume of simple questions: usage pricing works in your favor, as long as the unit is a strict resolution and you can cap spending.
- Large team, complex tickets: the seat line and per-agent add-ons dominate. Negotiate those first; Zendesk vs Intercom shows two different approaches.
- Unpredictable traffic: prefer a plan with a cap or a committed volume, and avoid session units that charge for failed attempts.
Frequently asked questions
Is per-resolution pricing cheaper than per-seat pricing?
Not in itself. It is cheaper when the AI resolves questions that would otherwise need extra agents, and more expensive when the vendor counts conversations that did not really need or get resolving. Model both parts in the true-cost calculator.
What is the difference between a resolution and an outcome?
A resolution is a conversation the AI closed without a person. Intercom's outcome is a broader unit it adopted in place of resolutions for Fin; read its help center definition before you estimate volume, because broader units mean more billable events.
Do free AI allowances reset every month?
It depends. Zendesk's pricing page includes 10 resolutions per agent a month on Suite Professional, capped at 10,000 a year. Freshdesk's 500 free sessions are once per account.
Can I buy the AI agent without paying for more seats?
Usually yes. The AI agent is charged by usage and does not need a seat, although it often needs a paid plan underneath.
Keep reading
- Help desk vs shared inboxguide
- What is an AI help desk?guide
- Do you need the AI add-on?guide
- Help desk migration checklistguide
- Best help desk softwarebest list
- Help desk true-cost calculator
This page was researched and drafted with AI assistance from the sources listed on it. We have not run hands-on tests of these products. Method: How we review