Skip to Content
<- Back to Blog
Published Feb 1, 20262 minUpdated Feb 1, 2026

Customer support chatbot checklist: 25 questions to answer before launch

A practical pre-launch checklist for AI support bots: data readiness, escalation, accuracy, and maintenance.

IAInqry AI
Customer support chatbot checklist: 25 questions to answer before launch

Most support bots fail for one reason: nobody defined what “good” looks like before launch. This checklist fixes that. Answer these 25 questions before you go live and you’ll avoid the biggest trust-killers.

1) Source of truth

  1. Do we have a single place for policies, pricing, and product details?
  2. Are the top 20 questions answered clearly in those docs?
  3. Who owns doc updates when something changes?
  4. How quickly do changes reach the bot (hours, days, weeks)?

2) Content coverage

  1. Are refunds, cancellations, and plan limits explicitly documented?
  2. Do we have onboarding answers (setup, integration, troubleshooting)?
  3. Are product names and feature labels consistent across docs?
  4. Are there “known gaps” we should block or escalate instead of guessing?

3) Answer quality

  1. Do we have a rule for “don’t know” responses?
  2. Is the bot allowed to estimate, or must it only use cited info?
  3. Do we require links to relevant docs in responses?
  4. What is an acceptable accuracy bar for launch (e.g., 80%+ on top 20)?

4) Escalation and handoff

  1. Where do escalations go (email, ticketing, Slack)?
  2. What keywords or intents trigger handoff?
  3. Do we collect the right context before handoff (email, order ID)?
  4. Are escalation hours clearly communicated?

5) UX and tone

  1. Is the greeting clear about what the bot can and cannot do?
  2. Does the tone match your brand voice?
  3. Are follow-up questions concise and specific?
  4. Is there a path to “talk to a human” in one or two clicks?

6) Analytics and ownership

  1. Who reviews conversations weekly?
  2. How are failed answers tagged or tracked?
  3. What metric defines success (deflection, CSAT, resolution time)?
  4. How often do we review and refresh sources?

7) Security and compliance

  1. Are we aligned on what data should never be collected (passwords, secrets)?

How to use this checklist

  • Answer each question with a single owner.
  • If you can’t answer a question quickly, that’s a pre-launch task.
  • Revisit monthly. What was true at launch usually breaks after the first product update.

Quick win: the top 5 that matter most

If you only have an hour, focus on these:

  • Source of truth for policies
  • Top 20 questions answered
  • Clear escalation path
  • “Don’t know” behavior defined
  • Weekly review owner assigned

If you want, copy this list into a Notion doc and assign owners. That alone gets you 80% of the way to a good launch.