From Intercom to Conveya in 3 weeks, without ticket leakage
A step-by-step migration plan with rollback checkpoints. For anyone who wants to switch without losing CSAT.
Vendor switches in customer support have a bad reputation, and rightly so. The biggest risks are not the tooling itself, but ticket leakage during the transition and the loss of macros, segments and routing you built up over years. The plan below is what we give customers who move from Intercom to Conveya. Three weeks, with explicit rollback moments and without the customer noticing a thing.
Week 1, Running in the shadows
Goal: Conveya configured, connected, trained. No customer sees any of it. Your team can test hands-on without pressure.
- Create a Conveya account and connect the same data sources you have in Intercom (Shopify, your CRM, your knowledge base).
- Import your Intercom macros. We have a script for this, export them as JSON from Intercom, hand them over during onboarding, and we convert them into agent instructions.
- Index your help center. Conveya reads your existing Intercom articles in directly, so you don't have to rewrite the content.
- Tag everyone on your team for access and run a first internal test week with fictional tickets. Aim for 50-100 test conversations to calibrate the tone.
Week 2, Running in parallel
Goal: Conveya alongside Intercom on your production channel, but with a safety net. Customers don't see Conveya, your team measures whether it does better than Intercom Fin.
- Turn Conveya on in 'shadow mode': for every incoming ticket Conveya generates a draft answer, but it is not sent. Your team sees both side by side, Intercom's proposal and Conveya's.
- For each category (order status, refund, product question) you keep a comparison. Which answer would you actually have sent? We give you a spreadsheet template for this.
- At the end of week 2 you have a performance score per category. Categories where Conveya reaches 80%+ acceptance can move on to week 3.
Week 3, Phased cutover
Goal: move the live routing from Intercom Fin to Conveya category by category, starting with the least risky.
- Day 15-16: order status routes over to Conveya. Low risk, high volume, the perfect starting segment.
- Day 17-18: pre-purchase / product questions.
- Day 19: refund + more complex cases. Here you let Conveya do only the preparatory work and human handling stays intact.
- Day 20-21: everything that's left. By the end of the working week 100% of production runs through Conveya, with Intercom as a read-only archive.
What you do NOT do
- Big-bang migration in a single evening. Customers notice immediately if CSAT drops and you have no time to tune.
- Move macros over literally. Conveya works differently from Intercom Fin: you describe the desired behaviour in natural language, not in step-by-step scripts. Our migration helper turns your macros into instructions, not one-for-one.
- Replace your telephony stack at the same time. Keep your migration scope strictly to the chat and email channels that are in Intercom now.
What it delivers (concretely)
Our first four customers who went through this route ended up at:
- Resolution rate +12 percentage points on average (from 71% to 83%) in the first month after cutover.
- Cost-per-resolution dropped 38% through per-message pricing instead of per-seat.
- Setup time 17 days on average (one team did it in 11), including all data connections.
- No noticeable CSAT dip during the migration itself, week 2 dampens the risk effectively.
Keep reading
Automating customer service: where do you start?
Automating customer service starts with your own inbox, not with the tool. Which questions to tackle first, how to measure success, and which beginner mistakes cost you time you did not need to lose.
ReadPlaybookChatbot or AI agent? The difference in plain language
A chatbot follows a decision tree. An AI agent reads along live and looks up the answer. Here is the difference in plain language, plus when you actually need which.
ReadPlaybook5 mistakes when deploying an AI chatbot (and how to avoid them)
Most AI chatbots don't fail because of the technology. They fail because of five choices around it. From giving the bot no access to your data to trying to automate everything at once, these are the traps and how to avoid them.
ReadReady to build this yourself?
Put your own AI agent live on your site, over email, WhatsApp and phone. Start today, no hassle.