Switching From Crisp: A Practical Migration Guide

First, be clear about why you are leaving

Nearly every team that looks at moving off Crisp is looking at one line on the invoice: the AI credits. The phrase hides the behaviour, so spell it out.

AI credits means you buy a dollar balance, and when it runs out mid-month the bot stops answering. It does not slow down. It stops, until you top up or the next billing period begins. Everything else on your plan keeps working. Here is how that plays out across the tiers.

In plain terms: on Mini, your ninety-first automated conversation of the month is the one that does not happen. On Essentials it is roughly your four hundred and fifty-first. A product launch, a shipping delay, a bad review going around: your busiest month is exactly the month the bot goes quiet.

The second reason people leave is the watermark. Removing Crisp branding from your widget requires the Plus plan at $295 a month. Until then your website advertises someone else's company.

What included AI looks like instead

EzyConn counts messages, not dollars, so you know your ceiling before the invoice arrives and there is nothing to top up. GPT-4o and Claude are on every plan, with no per-resolution and no per-seat fees. Annual billing takes 20% off.

Line the two up at the same price. Crisp Essentials at $95 carries about $25 of AI credit, roughly 450 AI conversations a month. The same $95 buys EzyConn Professional, which covers 25,000 messages a month and 11 people, with the AI included in the plan.

Full numbers sit on the pricing page and the Crisp alternative page.

When switching from Crisp pays off

If you want AI that answers from your own website pages and documents and shows the source, and a chat that goes to your team the moment a visitor asks for a person, the switch is worth a parallel run.

Step 1: List what you actually use (half a day)

Write down only what is live: the widget, the automations that fired this month, the saved replies people genuinely use, the connected integrations, the people who actually log in. Most teams find their real list is a third of what they configured over the years. That list is your scope. Moving the rest is how a two-day job becomes a three-week one.

Step 2: Export before you touch the subscription (half a day)

Do this while your account is fully active and paid, and open every file afterwards to check it is complete. Export first, cancel last, never the reverse.

Step 3: What breaks

Two things break in every migration between every pair of vendors. Plan for them.

Bot rules do not transfer. Scenarios, triggers and decision trees are written in each vendor's own format, and no competitor reads it. You rebuild them. The good news: AI trained on your help content replaces most hand-built branches, so teams typically rebuild five or six rules where they had forty.

Integrations need reconnecting. Every connected app has to be re-authorised on the new side. Boring rather than hard, but it is the step people forget.

Step 4: Run both side by side (two weeks)

This is the part that makes the switch safe, and the part everyone wants to skip.

Never put both widgets on the same page. Two chat bubbles confuse visitors and split your reporting. Split by page instead: the new widget on your blog, docs and pricing pages, the old one on everything else. Those pages get real questions from real people, but they are not your checkout.

Two weeks is enough to see a proper spread of questions, including the weird ones. Because the EzyConn free plan does not expire, the trial costs nothing and runs on live traffic rather than a demo dataset.

Step 5: The test list

Before you cut over, work through this. It takes an hour.

• Ask the new AI the twelve questions you get most often, taken from your real transcripts.
• Ask three questions it should refuse, and check it escalates rather than guesses.
• Check the handoff reaches a human, on a phone, out of hours.
• Reply to a test chat from Slack without opening a dashboard.
• Confirm the widget loads on mobile and does not cover your own buttons.
• Check what a visitor sees at 11pm on a Sunday.
• Have one non-technical colleague use the inbox for a day and complain freely.

The Slack test matters more than it sounds. If replies land in a Slack thread, your team answers from where they already are, and response times improve for reasons unrelated to AI.

Step 6: Cutover

Cutover is a ten-minute job on a quiet Tuesday morning, not a weekend project. Swap the script site-wide, remove the old one, update support links in your emails and footer, then tell your team in one sentence where conversations now arrive. Leave the old account alive until the term lapses. It costs nothing extra and it is your undo button.

Frequently asked questions

Crisp plan details reflect published pricing as checked in July 2026 and may have changed. Confirm any figure that matters to your budget directly with the vendor.