Going live with a chatbot without a proper pre-launch review is how you end up with a bot that confidently tells customers the wrong price, loops them in circles, or gives out information you never intended to share. We have seen all three and every one was avoidable.

This is the checklist we run through before launching any chatbot for a Mumbai client. Twelve steps. Do not skip any of them.

Why a pre-launch checklist matters

The good news: pre-launch review is fast once you know what to check. Most items on this list take minutes.

The 12-step chatbot pre-launch checklist

Step 1 Knowledge base audit

Read through every answer your bot is trained on. Are the prices current? Are the policies accurate? Is the tone consistent? Update anything that has changed since you first wrote the content. This is the most common miss.

Step 2 Edge-case testing

Ask the bot your ten most common questions. Then ask the five questions your team dreads most the complicated, edge-case ones. See what it does. A good bot says "I am not sure let me connect you to the team" rather than inventing an answer.

Step 3 Off-script testing

Type in something completely off-topic. "What is the weather today?" "Tell me a joke." The bot should have a graceful response that steers back to your business, not a broken loop or a bizarre answer.

Step 4 Human handoff trigger test

Whatever your handoff trigger is "speak to a human," "I want to talk to someone," typing in frustration test every variation you can think of. The escalation must work every time, without exception.

Step 5 Mobile testing

Most of your users are on mobile, probably on WhatsApp. Test the full conversation on a real phone. Check that messages are a readable length and that any links open correctly on mobile.

Step 6 Channel-specific testing

Test on every channel you are deploying. A response that looks fine on website chat may have formatting issues on WhatsApp or Instagram DMs. Test each one separately.

Step 7 Privacy and compliance check

Does the bot collect personal data (name, phone, email)? If yes, does your website have a Privacy Policy that covers chatbot data? In India, data collection rules are tightening. At minimum, tell users their data is being collected and why. See privacy policy basics.

Step 8 Branded tone check

Read three conversations aloud. Does it sound like your business? Is it too formal, too casual, or using words your team would never use? The bot's voice should match your brand the same standard applies as any customer-facing communication.

Step 9 Integration test (if applicable)

If the bot connects to your CRM, calendar, or inventory, test every integration end-to-end. Place a test booking. Log a test lead. Confirm it shows up exactly where it should.

Step 10 Load test

Send 10 15 test messages in quick succession. Does the bot handle simultaneous conversations? Does it slow down or give partial answers under load? This matters especially on WhatsApp during peak hours.

Step 11 Team awareness

Your support and sales team must know the bot is live. They need to know what it covers, what it escalates, and how to find the transcript when a customer tells them "I already spoke to your bot." Surprises here create internal friction.

Step 12 First-week review schedule

Plan a review for Day 3 and Day 7 after launch. Read real transcripts. Find where customers got stuck. Patch those gaps immediately. The first week of real conversations teaches you more than any testing session.

After launch: the two weekly habits

Once live, two habits keep your bot sharp. First, read five transcripts a week not all of them, just five. You will see patterns. Second, update the knowledge base whenever your prices, policies, or products change. A stale chatbot erodes trust slowly and silently.

If you want us to run the pre-launch review for you, book a session with Perceptra. It is part of every chatbot project we deliver.

Frequently asked questions