This is the follow-up promised in the review-replies guide. Inquiry responses automate on the same principle as review replies, with one difference — a wrong reply can become a promise that costs money. Which makes the structure matter more.
Conclusion first: unless you're fielding dozens of inquiries a day, you don't need a chatbot subscription. Search this topic and nearly every result is written by a chatbot company — of course they recommend chatbots. A general AI plus one document gets you the same outcome for close to free.
The target structure
inquiry arrives → AI drafts from the FAQ doc → human reviews → send
Inquiry handling fits the three automation criteria (recurring, describable, recoverable) only when the review step exists. Fully unattended sending will eventually misquote a price, promise unavailable stock, or confirm a booking you can't honor.
Step 1 — the FAQ doc is the asset (1 hour)
Open last month's inbox and pull the 15–20 questions that actually arrived. For each:
- The model answer — the best reply you've actually sent
- Never-promise list — discounts, exceptions, firm dates
- Escalation triggers — "if this word appears, no draft, straight to me" (refund, complaint, legal language, partnership pitches)
That document is the whole system. Same principle as the review-reply guideline doc — the document's quality is 90% of the output's quality. And it's valuable even without AI: it doubles as training material for your next hire.
Step 2 — implementation, by level
Level 1: copy-paste (start today) Paste the inquiry + your FAQ doc into the AI with one instruction: "Answer only from this document. If it's not covered, reply 'let me check and get back to you.'" That constraint is the core of the system — it's what stops the AI from improvising.
Level 2: morning batch With a tool like Claude Code, run "collect new inquiries and draft replies per the FAQ doc into a file" once each morning. Have the AI write that script too.
Level 3: triage + drafts + escalation Auto-classify inquiries by type, draft only what the FAQ covers, and flag escalation-trigger messages into a human queue with no draft. At this level, response handling drops from hours per week to minutes.
Three lines that keep you out of trouble
- Prices, stock, and schedules: only what's written. Let the AI infer and it will eventually sell inventory you don't have.
- Escalation words from day one. Refunds, complaints, legal language get a human, not a draft — same logic as the low-star rule in review replies.
- Review everything for two weeks. Fold the awkward drafts back into the FAQ doc, and by week three reviewing becomes approving.
Cost
Nothing beyond a general AI subscription (~$20/mo). Same principle as the small-business AI tools guide: consider a specialized service only after this structure hits its limits — not before.
Part of the AI Automation for Business series. Stuck setting it up? 1:1 coaching builds it with you over screen share. New posts via the newsletter.