AI chatbot launch checklist: 30 checks before going live

A practical pre-launch checklist for AI assistants covering content, accuracy, safety, security, handoff, compliance, accessibility, performance and monitoring.

By the UK Technology Hub teamPublished 3 minute read

The short answer

Before launching an AI chatbot, check that its knowledge is accurate and current, it has been tested with real and adversarial questions, guardrails and handoff work, security and privacy controls are in place, users are told they're talking to AI, the widget is accessible and fast, analytics are set up and staff know how to support it. Launch gradually and review conversations closely in the first weeks.

A rushed chatbot launch can create embarrassing answers, frustrated customers and compliance problems. Use this checklist to launch confidently.

Knowledge and content

  1. Knowledge sources reviewed for accuracy and currency.
  2. Contradictions between sources resolved.
  3. Policies and edge cases documented.
  4. Topics the bot must not answer defined.
  5. Owners assigned for content updates.

Testing

  1. Tested with 50 to 100 real customer questions.
  2. Tested with typos, slang and multi-part questions.
  3. Tested with out-of-scope questions.
  4. Tested with adversarial attempts to break rules. Our guide to prompt injection explains what to try.
  5. Answers scored for accuracy, and fixes made.

Guardrails

  1. Bot admits when it doesn't know.
  2. No promises of discounts, refunds or terms beyond policy.
  3. Sensitive topics trigger appropriate responses and escalation.
  4. Emergency situations handled with clear guidance.

Handoff

  1. Users can reach a person easily.
  2. Conversation history passes to agents.
  3. Out-of-hours handoff collects details and sets expectations.

Actions and integrations

  1. Actions require confirmation.
  2. Permissions limited to what's needed.
  3. Integrations tested end to end.

Security and privacy

  1. Data processing agreements in place.
  2. Training data opt-outs configured where needed.
  3. Retention periods set.
  4. Access to transcripts restricted.
  5. Privacy notice updated.

Transparency and accessibility

  1. Clear disclosure that users are talking to AI.
  2. Chat widget accessible by keyboard and screen readers.
  3. Widget doesn't slow page loading or cover key content on mobile.

Measurement and operations

  1. Analytics set up for resolution, satisfaction, handoff and unanswered questions.
  2. Staff trained on monitoring, handoff and updating content.

Launch gradually

Start with a subset of pages, channels or users. Review conversations daily in the first week, then weekly. Fix issues quickly.

After launch

  • Review unanswered questions and add content.
  • Monitor costs and usage.
  • Update content as products and policies change.
  • Revisit guardrails and security regularly.

Choosing the right platform first

If you haven't selected a platform, our guide to AI chatbot platforms explains how to compare options.

A worked example

A regional bus operator prepared to launch a chatbot to answer timetable, ticket and lost property questions. Its team used a launch checklist covering knowledge, testing, handoff, security, accessibility and measurement. Testing revealed that the chatbot gave outdated timetable information for routes recently changed, so the team connected it to the live timetable feed. They launched to a small share of website visitors first, reviewed conversations daily for two weeks and fixed problems before full launch. The chatbot handled common questions reliably, and complaints about it were rare.

Launch day plan

  • Confirm all integrations are working.
  • Make sure support staff know the chatbot is live.
  • Monitor conversations closely.
  • Have a quick way to switch off features if needed.
  • Collect early feedback from customers and staff.

First month review

After a month, review:

  • Resolution and handoff rates.
  • Customer satisfaction.
  • Common unanswered questions.
  • Accuracy issues found.
  • Security or privacy incidents.
  • Costs compared with expectations.

Use the review to plan improvements and decide whether to expand.

Communicating the launch

Tell customers what the chatbot can help with, and remind them they can still reach a person. Clear expectations reduce frustration.

Documentation

Document the chatbot's purpose, sources, settings, integrations and owners. Good documentation helps new team members and supports compliance.

Planning the next phase

Once the basics work, consider adding actions, new channels or languages, one at a time, with testing for each.

Who should sign off

Before launch, get sign-off from the people responsible for customer service, data protection, security and the brand. Each will spot different risks, and shared sign-off creates shared ownership for the assistant after launch.

Keeping the checklist

Store the checklist with your documentation and reuse it for every major update.

Fallback plans

Prepare for outages. If the chatbot or a key integration fails, visitors should see a clear message with other ways to get help, such as a phone number, email address or contact form, rather than an error or a silent chat window.

How we help

Our AI chatbot development team runs pre-launch testing and staged rollouts, our AI solutions team sets up analytics and security and our maintenance and support plans include ongoing monitoring and content updates.

Frequently asked questions

How much testing does a chatbot need before launch?

At least 50 to 100 realistic questions, plus edge cases and adversarial tests.

Should I launch a chatbot on my whole website at once?

A gradual rollout lets you catch problems before they affect everyone.

What's the most common launch mistake?

Launching without clean, current knowledge and without an easy route to people.

This guide is general information, not legal, tax or financial advice, and details such as prices, features and rules can change. Product and company names mentioned are trademarks of their respective owners and are used only to identify those products; UK Technology Hub is not affiliated with or endorsed by them.

Related guides

All AI chatbots & assistants guides