Why the AI copilot is in the plan, not an add-on
How ChaosDesk can include reply drafts, summaries, translation and triage in every paid plan, and what fair use means in practice.
Generative AI is now table stakes in a helpdesk. In many products it is also a second subscription: an add-on per agent, on top of the plan. We took a different route and put the AI copilot into every paid ChaosDesk plan. Here is why that works, and what you get.
What the copilot does
- Reply drafts with sources. Suggestions grounded in your help-centre articles and similar solved tickets, with the sources shown so agents can check them.
- Summaries of long threads, so a ticket that changes hands does not need a re-read.
- Translation both ways. Customers write in their language, agents read and answer in theirs.
- Intent and sentiment tags on arrival, ready for triggers, views and reports.
- Suggested macros, the ones agents used on similar tickets, at the top of the menu.
Why it fits in the price
The cost of an AI action is small and falling: a typical copilot action (a couple of thousand tokens in, a few hundred out, on an efficient model) costs well under a cent. What makes add-on pricing expensive is not the model, it is the pricing model.
We size the allowance per plan so that every plan keeps a healthy margin even when a team uses it heavily:
| Plan | Copilot actions per agent per month |
|---|---|
| Starter | 50 |
| Growth | 300 |
| Pro | 450 |
| Enterprise | 700 |
Allowances as of 2026-11-09; the current numbers are always on the pricing page.
Fair use, pooled
The allowance is pooled across your paid seats, so a power user borrows from quieter colleagues instead of hitting a personal cap mid-afternoon. It is a pool, not an infinite one: if the whole workspace spends its month’s actions, copilot actions pause until the next month or until you upgrade. Admins see usage per agent under Admin Center, Integrations, AI & translation, in Usage and cost, so a pool running dry is something you see coming.
Privacy
Personal data is masked before text is sent to the AI provider, provider traffic is not used to train models, and admins can switch AI off for the whole workspace with one setting.
The AI agent is metered separately
The autonomous AI agent is a different thing from the copilot: it answers customers on the messenger, email and WhatsApp by itself and hands over to a person when it should. It is in beta, and it is billed per automated resolution, at $0.50 excluding tax (as of 2026-11-09; see the pricing page). Copilot actions and AI-agent resolutions are counted separately, so switching the agent on does not spend your copilot allowance.
