Escalate negative sentiment users to live chat
Last updated: 09 April 2025
Delivering a great chatbot experience also means knowing when to automation should step aside and let a human step in.
In this doc, you will learn to set a sentiment-based live chat fallback to ensure highly negative users get the help they need from a human.
This functionality requires the sentiment analysis and live chat add-ons. Speak to your account manager to enable.
Step 1: Enable live chat
Make sure live chat is enabled and your live chat open hours are set.
Step 2: Enable negative sentiment fallback
Go to Service → Live chat and open the Settings.
In the side panel, turn Auto escalation on.
In the new field that appears, enter the number of consecutive negative sentiment messages a user must send before they are escalated to live chat.
Auto-escalation does not trigger if the user is in the middle of a workflow. Workflow transitions take priority over any new automation.
In this example, we’ve set two. This means on the second negative sentiment message, the chatbot will not respond (e.g. trigger an FAQ or a workflow) and immediately escalate to live chat.
Step 3: Create your live chat fall back workflow
For this step, you may either create a new workflow from scratch or simply use the existing default Live chat fallback workflow.
For the purpose of this doc, we will simply edit the existing workflow.
Go to Builder → Live chat fallback.
Add a New trigger to the first step of the conversation.
In the conditional, select Negative sentiment limit is reached.
Voila! As soon as a user sends, in this case, two consecutive negative sentiment messages, they will automatically be escalated to our live chat fallback workflow.
Frequently asked questions
My fallback is not triggering, what could I have done wrong?
If you have followed all the steps above and your fallback isn’t triggering, check the following:
Have you set your live chat working hours? The fallback will only work during these hours.
Are you sure the user has sent X consecutive negative messages? A neutral or positive message will break the streak and fallback will not trigger.
What constitutes a ‘negative sentiment message'?
Our sentiment analysis feature identifies negative sentiment through AI by analysing the intent behind each message sent by your users. To a human, negative sentiment should be fairly obvious to the naked eye, too:
“I hate this t-shirt” is negative.
“I like this song” is positive.
“Trees are tall” is neutral.