Saved-reply suggestions mined from resolved tickets
An agent reviews recently resolved tickets, finds questions your team answers repeatedly, and proposes new or improved saved replies for a lead to approve.
The use case
Small support teams retype the same answers about shipping times, password resets, or how to change a plan, and each agent words it slightly differently. Existing saved replies go stale as policies change. A mining agent periodically reviews resolved tickets, clusters repeated questions, and compares the replies that resolved them against your current macro list. It proposes new saved replies, updates to outdated ones, and macros nobody uses. A support lead reviews each suggestion against current policy, edits the wording, and adds approved replies to the help desk. Nothing changes in the live macro library without that approval.
Example workflow
- A monthly schedule exports resolved tickets and the current saved-reply list from the help desk.
- The agent strips customer names, emails, order numbers, and other personal details before analysis.
- It groups tickets by repeated question and identifies the replies that resolved them without follow-up.
- Claude drafts proposed saved replies, marks existing ones that conflict with recent answers, and lists unused macros.
- Suggestions go to a review sheet where the support lead approves, edits, or rejects each one.
- The lead adds approved replies in the help desk and records the change and date in the log.
Tools
- Zendesk, Help Scout, or Gorgias
- Google Sheets
- Claude
- Make or Zapier
- Notion
Risks and guardrails
- A popular past answer may reflect an old policy; check every suggested macro against the current written policy before approving.
- Ticket exports contain personal data; redact before sending to the AI tool and confirm your vendor's data handling.
- Macros can make replies sound robotic; keep placeholders for personalization and encourage agents to adapt them.
- One-off exceptions might get turned into standard answers; exclude tickets tagged as exceptions or escalations.
Related recipe
Support Ticket and Review Triage with Policy-Cited Draft Replies: An agent tags every helpdesk ticket and public review by intent, urgency, and sentiment, routes it, and drafts a reply that cites your help center. People approve refunds and all public replies.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Zendesk AI (official page): www.zendesk.com/service/ai (checked 2026-11-09)
- Help Scout (official page): www.helpscout.com (checked 2026-11-09)
- 中央社新聞 CNA TAIWAN NEWS (official page): ask.cna.com.tw/ask (checked 2026-11-19)
- Gorgias (official page): www.gorgias.com (checked 2026-11-09)
- Claude (official page): claude.ai (checked 2026-11-09)
- Make MCP server (official page): developers.make.com/mcp-server (checked 2026-11-09)
- Zapier MCP (official page): zapier.com/mcp (checked 2026-11-09)
- Notion MCP server (official page): developers.notion.com/guides/mcp/overview (checked 2026-11-09)