AI copilot and AI agent
The AI agent converses with the customer and resolves without anyone stepping in.
The copilot works for the person on the team and never sends anything by itself. They complement each other. The agent takes what it can close alone, and the copilot speeds up what reaches a person.
Why it matters
What reaches a person is the hard part, and the customer who gets there has already waited.
The copilot takes the mechanical work out of that moment, such as reading a long thread or finding the article, and leaves the judgement to the person. It is also the lowest-risk way to start with AI, because nothing goes out unreviewed.
What a copilot does
An AI copilot does five things, all at the person’s request or in their view.
- Summarise. Boils a long conversation or a customer’s history down to a few lines.
- Search. Finds the answer in the documentation and in earlier cases.
- Draft. Proposes a reply in the company’s tone.
- Rewrite. Shortens, corrects or translates what the person wrote.
- Document. Turns a useful reply into a help article.
Copilot versus agent
The difference is who talks to the customer and who decides.
| AI copilot | AI agent | |
|---|---|---|
| Works for | The person on the team | The customer |
| Who sends | The person | The agent |
| Where it helps | On what reaches the team | On what is resolved without the team |
| Measured by | Handle time and quality | Resolution rate |
What its usefulness depends on
A copilot is only as good as what it can read.
With out-of-date documentation it proposes out-of-date replies, just as confidently. Before switching it on, review the most-used articles.
The copilot in Intake
Intake Copilot works inside the inbox.
It drafts the reply from what the team knows, sums up the thread and rewrites a draft, and sending is the person’s call.
