# How to switch live chat software without losing a conversation

> A vendor-neutral migration plan for live chat: what to inventory, how to export and import, what you rebuild, and how to run cutover day so no customer gets dropped.

Source: https://pengo.chat/blog/switch-live-chat-software/
Topic: Buying guides
Updated: 2026-10-02

## The short version

- Switching chat tools is mostly **inventory and timing**, not data. List everything you use in the old tool before you export anything.
- **Contacts and conversation history** can usually be exported and imported. Saved replies, routing, triggers and bot flows almost always have to be **rebuilt by hand**.
- Build the new setup **alongside the old one** for a week, but never run two widgets on the same page.
- On cutover day, **swap the snippet everywhere in one release** and keep the old inbox open, so replies already in flight still reach customers.
- Cancel the old tool only after a **final export** and two weeks of watching both inboxes. Check the renewal date first, so you don't pay twice.

Switching live chat tools sounds risky because the chat is where customers talk to you. If something breaks, nobody files a bug report. They just leave, or they reply to a conversation that no longer reaches anyone.

The risk is real, but it's concentrated in a few predictable places: settings nobody wrote down, replies landing in the old inbox after the switch, and a snippet left running on one forgotten page. This guide takes you through a migration that works with any pair of tools, in the order we'd do it, with the parts that usually go wrong called out.

## Why teams switch live chat software

Most teams don't switch because of one feature. They switch because the tool's cost or shape stopped fitting the team. The common reasons:

- **The pricing model.** Per-seat pricing gets expensive as more people help out in chat, and per-resolution AI pricing makes the bill rise as the AI gets better. Many teams move to a tool with simpler, predictable pricing.
- **AI that costs extra or does too little.** Some tools charge separately for AI, others bundle an assistant that can't hand off well. Our guide to [setting up an AI agent that actually resolves questions](/blog/ai-agent-for-customer-support/) covers what good looks like.
- **Site speed.** A widget that loads a lot of code on every page view can drag down your Core Web Vitals. You can [measure your current widget](/blog/chat-widget-site-speed/) before deciding.
- **Missing pieces.** No mobile app, a weak shared inbox, no way to see who's on the site, or reports that don't answer your questions.

Pick your timing well. Don't move tools in your busiest weeks; an outdoor clothing store like Kettlewell Outdoor shouldn't cut over in the run-up to Christmas. And if you're locked into an annual contract with months left, do the preparation now but schedule cutover close to the renewal date, so you're not paying for two tools for long.

> **Cutover**
>
Cutover is the moment the new chat widget replaces the old one on your live site. Everything before it is preparation that customers never see. Everything after it is cleanup: catching stray replies, filling gaps and closing the old account. A good migration makes cutover the shortest and least eventful part.

## The migration at a glance

The plan below assumes a small team, a few hundred to a few thousand past conversations, and someone who can spend a few hours a week on it. Bigger setups take longer in week −1; the shape stays the same.

_Figure: A four-week migration around a single cutover day. Most of the work happens before customers see any change, and the old tool stays available for two weeks afterwards._

## Step 1: Inventory everything you use

Before you export a single file, make a list of everything your current tool does for you. This is the step people skip, and it's where most migration problems come from: a routing rule nobody remembered, a proactive message on the pricing page, a "Chat with us" button in the footer that calls the old widget directly.

Go through your current tool's settings page by page and note:

| What | What to write down | Why it matters |
|---|---|---|
| **Contacts** | Count, and which custom fields you rely on | Fields that don't map get lost on import |
| **Conversation history** | Rough count, and how far back you need it | Decides how many imports you'll run, or whether you keep an export instead |
| **Saved replies / macros** | Every one, with any variables | Almost never exportable; you'll recreate them |
| **Tags** | The ones you actually use | A good moment to prune the list |
| **Help articles** | Where they live, and whether they're public | Your AI and help center depend on them |
| **Routing rules** | Who gets which chats, and fallbacks | Mistakes here mean missed chats |
| **Team and roles** | Who has access, and who's an admin | You'll invite people again |
| **Business hours** | Hours, time zone, away message, offline form | Out-of-hours visitors are the easiest to lose |
| **Proactive messages** | Page, timing or condition, text, frequency | They're often driving real conversations |
| **Bot flows and automations** | Each flow as plain steps | Rarely portable; they get redesigned |
| **Integrations** | Slack, CRM, calendars, anything connected | Each needs reconnecting and testing |
| **Widget customization** | Colors, position, welcome text, launcher | So the new widget feels familiar |
| **Where the snippet lives** | Every site, subdomain, template, plugin and tag | The one you forget keeps the old widget alive |

Two more things belong on the list: the date your current plan renews and its notice period, and any code on your site that talks to the widget's JavaScript API, such as custom buttons that open the chat or a call that identifies logged-in users.

For Kettlewell Outdoor, Owen Pryce's inventory came to 1,284 contacts, around 2,140 past conversations, 37 saved replies, a dozen tags, two routing rules (walking club group orders to Hannah, everything else round robin), weekday business hours, and one proactive message on the cart page. That fits on a single page, and it turned into the plan below.

_Figure: An interactive migration planner. Tick what you use in your current tool and the plan updates to match, grouped by week. It lists tasks rather than guessing hours, because those depend on your setup._

## Step 2: Export from the old tool

Most chat tools let an admin export contacts as **CSV** and conversations as **JSON** or CSV. The option usually lives under settings, data or account, and large exports are often emailed to you when they're ready. Ask your vendor's support if you can't find it; they get the question all the time.

When the files arrive, check them before you need them:

- **Open the contacts file** and confirm the columns you rely on are there: email, name, company, and any custom attributes like plan or order count.
- **Spot-check a few conversations** in the history export. Are messages in the right order? Are internal notes included, and do you want them?
- **Note what's missing.** Attachments, ratings and bot transcripts are sometimes left out of exports. Better to know now than after you've cancelled.

These files contain personal data, so treat them that way. Keep them somewhere access-controlled, not in a personal downloads folder, and delete local copies once the import is done and you've stored one master export securely.

## Step 3: Import what moves, rebuild what doesn't

Here's what a migration carries across. Contacts move. Conversation history moves when the new tool has an importer for the old one. Almost everything that describes *how* you work, such as saved replies, routing and automations, gets rebuilt by hand.

_Figure: What carries across in a typical live chat migration. Data moves; the way you work gets rebuilt; reports and old automations stay behind in an export you keep._

Rebuilding sounds like a chore, but it's the best part of the move. Every team accumulates saved replies nobody uses, tags with three conversations each and routing rules for a teammate who left. Recreate the twenty saved replies you use most and the tags that appear in your reports; let the rest go. If something turns out to be missing, someone will ask for it within a week.

Bot flows deserve their own thought. Decision-tree flows ("Sales or support?") rarely transfer between vendors, and if you're moving to a tool with an AI agent, you may not need most of them. Write each one down as plain steps, then ask whether an AI answer from your help content, a saved reply or a handoff rule does the same job.

Pengo has **built-in importers for the most popular live chat tools** in Settings → Workspace → Import. Each takes a contacts CSV and a conversations JSON export, and imported threads arrive in the inbox tagged as imported, so your team has the full customer history from day one. From any other tool, contacts come in through the generic CSV option. Each upload takes one file of up to 5 MB and up to 300 conversations, so Kettlewell's 2,140 conversations go in 8 batches (2,140 ÷ 300 = 7.1, rounded up).

While you're rebuilding, set up the parts customers will notice first:

- **The widget.** Match your colors, position and welcome text closely enough that regular visitors don't wonder whether they're on the right site.
- **Business hours and the away message.** Then test the offline form by visiting outside hours, or by temporarily changing the hours.
- **Routing.** Send a test chat down every route, including the fallback when the assigned person is away.
- **Notifications.** Every teammate should get a desktop or mobile notification for a new chat before cutover, not after.
- **The AI, if you're adding one.** Point it at your help content and test it against your most common questions before it talks to anyone.

## Should you run both tools in parallel?

Yes, but run both **inboxes** in parallel, not both widgets on the same page. Two widgets on one page means two bubbles, two sets of scripts loading, and conversations split between two inboxes that each look half-empty.

What works instead:

- **A hidden test page.** Install the new snippet on a page that isn't linked from anywhere, and use it for a week of practice chats. Send one from your phone, one from a colleague's laptop, one outside business hours.
- **A team rehearsal.** Spend 30 minutes with everyone who'll answer chats. One person plays the customer; the others find the saved replies, assign, add a note, hand off. People learn a new inbox faster by using it than by reading about it.
- **One section first, if you must.** On a large site you can move one section, such as the help center, a week early. Keep it to clearly separate pages so no visitor ever sees both widgets.

## Cutover day: swap the snippet without dropping anyone

Pick a quiet weekday morning when the person doing the swap and at least one person answering chats are both around. Look at your chat volume by weekday and avoid the busiest; you also want a full working day afterwards to catch problems.

The swap itself is small. Remove the old snippet, add the new one, in a single release, so there's no window with no chat and no window with two. Where that happens depends on how the old one was installed:

- **In the site's code or theme**: replace the script tag before `</body>` in the shared layout.
- **Through a plugin or app** (WordPress, Shopify, a site builder): deactivate and remove the old vendor's plugin or app, then add the new snippet or plugin.
- **Through Google Tag Manager**: pause the old tag, add the new one and publish both changes in the same container version.

Then go looking for the places people forget: a help center on a subdomain, a separate checkout template, landing pages built in another tool, your web app if customers chat from inside it. A quick way to check each is to open it with DevTools' Network panel filtered for the old vendor's domain. If anything still loads, there's a snippet left somewhere.

Three technical details catch people out:

1. **Caching.** If your site uses page caching or a CDN, purge it after the swap, or some visitors will see the old widget for hours.
2. **Content Security Policy.** If your site has one, add the new widget's domains or it simply won't appear.
3. **Code that calls the old widget.** Custom "Chat with us" buttons, links that open the chat, and identify calls for logged-in users all reference the old vendor's JavaScript API. They fail silently after the swap unless you update them.

Pengo installs with one script tag, and there are install guides for [HTML, WordPress, Shopify, Webflow, Next.js and Google Tag Manager](/docs/install/). Buttons that opened your old widget can call `pengo('open')` instead, and logged-in users are identified with `identify`, with an optional HMAC user hash. Pengo's [importers and tool-specific switching guides](/integrations/#importers) cover each tool's export steps and field mapping.

### Keep the old inbox open

Conversations don't end on your schedule. On cutover day there will be chats that started in the old tool and haven't finished, and customers who reply days later to a transcript or follow-up email the old tool sent. Those replies go back to the old tool, not the new one.

So don't cancel or lock the old account on cutover day. Keep at least one person logged in, set it to send no new proactive messages, and have someone check it every day for the first week. When Ruth Adebayo replied on day two to a week-old email about a leaking Ridgeline Rain Shell, it landed in Kettlewell's old inbox, and Owen picked it up there because he was still checking it.

Do the same for email. If your support address forwards into the old tool, point it at the new one the same day.

**Cutover day checklist**

- [ ] Tell the team the cutover time and who’s answering chats today
- [ ] Remove the old snippet and add the new one in the same release
- [ ] Check subdomains and other templates
- [ ] Purge the page cache and CDN
- [ ] Add the new widget’s domains to your Content Security Policy, if you have one
- [ ] Update custom chat buttons and identify calls
- [ ] Send test chats from a phone and a private window
- [ ] Confirm every teammate gets a notification for a new chat
- [ ] Turn off proactive messages in the old tool, but keep its inbox open
- [ ] Point your support email forwarding at the new inbox
- [ ] Note today’s numbers

## Tell your team, and mostly don't tell customers

Your team needs more notice than your customers. A week before cutover, tell everyone who answers chats what's changing, when, and where things moved: saved replies, tags, how to assign, how to hand off. The rehearsal covers most of it. A one-page "where things are now" note covers the rest.

Customers mostly don't need an announcement. The bubble is in the same corner and looks much the same. If your old widget showed returning visitors their previous conversations, or customers are used to a particular email address for replies, say so in a short line where they'll see it. Otherwise, the best migration is one they never notice.

Two pieces of housekeeping belong here too. Add the new vendor to your privacy policy's list of processors if you keep one, and check that your cookie or consent banner describes the new widget correctly.

## The first two weeks after cutover

The first two weeks are about catching what the plan missed. Spend ten minutes a day on it:

- **Check the old inbox** for replies that landed there. They should tail off within days.
- **Compare missed chats and first response time** with the numbers you noted on cutover day. A jump usually means a routing rule or a notification isn't working, not that the team got slower.
- **Ask the team what's missing.** Saved replies people reach for and can't find, tags they want back, a view they used daily. Add them as they come up.
- **Spot-check imported history.** Open a few imported conversations and check names, dates and message order.
- **If you've added an AI agent, read its conversations.** Twenty a week is enough to catch content gaps early.

At the end of week two, close out. Take one final full export from the old tool and store it with the others. Revoke its API keys and app connections. Cancel or downgrade before the renewal date, then ask the old vendor to delete your data. Most vendors have a process for this, and keeping a copy of the confirmation is sensible under GDPR.

## Mistakes we see most often

- **Skipping the inventory.** The forgotten routing rule or proactive message is always the one that mattered.
- **Running two widgets on the same page.** It looks broken to visitors and splits your conversations.
- **Missing a snippet.** The help center on a subdomain keeps sending chats to an inbox nobody's watching.
- **Cancelling on cutover day.** Replies keep arriving in the old tool for days. Give it two weeks.
- **Rebuilding everything exactly as it was.** A migration is the cheapest moment you'll ever get to drop saved replies, tags and bot flows that no longer earn their place.

A chat migration done this way is mostly quiet work in the two weeks before anyone notices: a list, a few exports, an afternoon of rebuilding and a rehearsal. Cutover day is a ten-minute change and a few hours of watching. If you're still deciding where to move to, see [everything Pengo includes](/features/).

## Questions people ask

### Can I import my chat history into new live chat software?

Often, yes, but it depends on the pair of tools. Most vendors let you export conversations, and many newer tools have importers for the common ones. Pengo has built-in importers that bring over contacts and full conversation history from the most popular live chat tools, plus a CSV importer for contacts from anything else.

### How long does it take to switch live chat software?

The swap itself takes minutes, and setting up the new tool fits in a day or two: install the widget, import contacts and history, rebuild saved replies and routing, and train the AI. Plan the whole move over a few weeks so you can rehearse before cutover and keep the old tool for two weeks afterwards to catch stray replies.

### Can I run two chat widgets on my site at the same time?

Not on the same page. Visitors see two bubbles, you load two sets of scripts, and conversations split between two inboxes. Run the new tool on a hidden test page or a single section while you set it up, then swap everything in one go.

### What happens to conversations that are open when I switch?

They stay in the old tool. If a customer replies to an old chat by email, the reply goes to the old inbox, so keep it open and check it daily for a week or two. Close out what you can before cutover, and pick a quiet day so there are fewer open chats to begin with.

### Do I need a developer to switch chat tools?

Usually not. Most widgets install with one script tag, a plugin or a Tag Manager tag, and swapping it is a small change. You may want a developer if your site has custom buttons that open the old widget, identifies logged-in users to it, or has a Content Security Policy that needs updating.

### Will switching chat software affect my website?

Only the widget changes, but it's worth measuring before and after, because chat widgets vary a lot in how much they load. Our guide to [measuring a chat widget's effect on site speed](/blog/chat-widget-site-speed/) shows how to compare the two in about an hour.
