Claude Code for Customer Service Draft Review
hermanschutte · x · 2026-07-17
The author explains that their system first checks if their own chatbot gave customers wrong answers, then prompts them to send out a correction the same day.
This workflow relies on a few key rules:
- Drafts only, no auto-sending. Everything requires final human confirmation to prevent mistakes.
- If the system detects a customer has already received a reply, it automatically skips them, avoiding duplicate responses that feel robotic.
- Core principle: never write unverified runtime info into drafts. Otherwise, the LLM might easily hallucinate non-existent features.
- The system first categorizes conversations into urgent / prospect / noise, then cross-references them:
- Bug reports against real production errors and code;
- Usage issues against documentation;
- Support threads against tickets and history.
A follow-up adds that the setup is actually quite lightweight: Claude Code runs on a cron job at 6:00 AM daily, connecting to 4 MCP servers (support inbox, Sentry, Nightwatch, and the author's email client), with all instructions written in a single markdown file.
Related event: Customer Service Agents Prevent Hallucinations via Draft Reviews and MCP(4 posts)→
More from coding & agent
- AI agents are starting to strain code hosting platforms — craigsdennis · 2026-07-21
- Async OPD distillation doubles throughput while matching synchronous math accuracy — _lewtun · 2026-07-21
- Omnigent 0.6.0 adds Claude Code imports, Slack approvals and desktop apps — matei_zaharia · 2026-07-21
- Google appears to have quietly shipped Gemini 3.6 Flash, with lower pricing and better agentic scores — xiaohu · 2026-07-21
- Open-source CLI audits AI tools, MCP configs, and agent skills on local machines — Initial-Copy332 · 2026-07-21
- Coding agents feel less stressful when the 5-hour limits are temporarily removed — iamrobotbear · 2026-07-21