Glossary
Co-pilot mode
Definition
Co-pilot mode is a human-in-the-loop configuration where the AI agent drafts replies and actions, but a human reviews and approves each one before it reaches the customer.
In co-pilot mode the agent does all the work - classifies the ticket, pulls the order, drafts the grounded reply - but nothing is sent until a person clicks approve. The human can edit the draft first; rejects go back to the team like a normal escalation.
Teams use co-pilot as the on-ramp to autonomy: run it for a week or two, watch the approval rate, and flip to auto-resolve once the drafts prove themselves. It is also the permanent right answer for merchants whose policy requires a human on every send.
Why it matters
The scariest moment in support automation is the first week. Co-pilot removes the leap of faith: you get the speed of AI investigation with zero chance of an unreviewed reply reaching a customer.
How Resolvo HQ approaches it
Resolvo HQ ships co-pilot as a first-class mode: drafts land in an Approvals queue with the customer question, confidence, and an editable reply; approve posts it back into the original channel (widget, email, Gorgias, Zendesk) and only then bills a resolution.
Frequently asked
Does co-pilot mode still save time if a human reviews everything?+
Yes - reviewing a correct, pre-investigated draft takes seconds; writing one takes minutes. The lookup and drafting work is where the time goes, and that part is automated.
How do I know when to switch from co-pilot to auto-resolve?+
Watch your approval rate. When you are approving the overwhelming majority of drafts without edits, the agent has earned the confidence threshold - switch WISMO-style intents first and keep sensitive ones gated.
See it on your own tickets
Replay real tickets in a sandbox and watch the agent resolve them before you pay.