What it does

Copilot helps agents inside Customer Service workspace: it answers questions from your knowledge and trusted websites, drafts email and chat replies, and summarises long cases and conversations so a handover doesn't mean re-reading the whole thread.

Key facts

• Core features: Ask a question (knowledge-grounded answers), Write an email, Draft a chat response, and case/conversation summaries.
• Grounding sources are things you configure — internal knowledge articles and specified trusted external sites. Point it at rubbish and it answers with rubbish.
• Included with Customer Service Enterprise; the newer autonomous agents (see Service Agent) are separate and metered.
• Admin controls live under Copilot / Agent experience in the Customer Service admin centre, and some features need explicit opt-in.

When to use / skip

Switch on summaries and knowledge answers on day one — they're the lowest-risk, highest-thanks features. Hold back on customer-facing draft replies until you've watched agents review them for a while.

Configuration decisions

• Which knowledge sources and trusted websites Copilot may cite.
• Which features are on per queue/team, and whether answers can be sent without edits.
• Data and regional settings, since some capabilities are geo-limited.

Gotchas

• If knowledge is stale or thin, "Ask a question" quietly disappoints and agents stop using it. Fix the knowledge base first.
• Summarisation tone can read oddly on emotional cases. Coach agents to sanity-check before pasting.

Consultant notes

• Treat knowledge readiness as a prerequisite workstream, not a nice-to-have. It's the single biggest predictor of adoption here.
• Turn features on gradually and measure handle time and CSAT, so you can show the win.
• Keep Service Agent (autonomous, metered) mentally separate from this agent-assist Copilot when scoping and costing.

---

Review after each Customer Service wave.