Run the evaluation
- Count monthly conversations and the share a bot could plausibly resolve from your top ten question types.
- Find each vendor's billable AI unit (outcome, resolution, session or automated interaction).
- Multiply expected resolved volume by the per-unit price and add it to the seat cost.
- Model a 30 percent volume swing, since usage billing moves with demand.
- Ask what happens at the cap or when the bot hands off, then pilot with a billing limit if one exists.
How to read the result
Decide first how your bill should scale: per agent seat (Zendesk, Freshdesk, Help Scout, Front, Intercom), per workspace (Crisp), per ticket (Gorgias) or per AI-resolved conversation. Then price a realistic AI-agent volume, because vendors meter AI separately from seats. Chatwoot is the only product here you can self-host, and the self-hosted build is a different commercial arrangement from Chatwoot Cloud.