Escalation Rules
Escalation Rules determine when a conversation should be handed off from the AI Resolution Agent to your support team.
Use these rules to ensure customers are connected to a human agent whenever the AI shouldn't continue the conversation—for example, when a customer requests an agent, expresses frustration, asks about complex topics, or when a conversation becomes too long.
These rules apply across all supported channels and take precedence over individual AI journeys and campaigns.
How Escalation Works?
During every conversation, the AI continuously checks whether any escalation rule is met.
If a rule is triggered:
The AI stops responding.
The conversation is handed off to your support team.
Your support team's assignment and routing rules continue as configured in Helpdesk.
Some escalation rules are enabled by default, while others can be customized to match your business needs.
Default system rules
These rules are built in and cannot be edited.
Customer asks to speak to a human
If a customer explicitly asks to speak with a person—for example:
"Can I talk to an agent?"
"Connect me to support."
"I want to speak with a human."
the conversation is immediately handed off to your support team.
Customer sentiment
The AI also monitors customer sentiment throughout the conversation.
If it detects signs of frustration, anger, or distress, the conversation is automatically escalated to a human agent.
Examples include:
Repeatedly asking the same question
Expressing dissatisfaction
Using aggressive or emotional language
Sending repeated all-caps messages
This helps ensure customers receive human assistance before their experience deteriorates.
Custom escalation rules
In addition to the default rules, you can define rules that are specific to your business.
Customer Intent
Route conversations based on what the customer wants to do.
Choose up to three intents that should always be handled by your support team instead of the AI.
Examples:
Returns & refunds
Order cancellations
Payment disputes
When the AI detects one of these intents, it immediately hands the conversation over.
Trigger Messages
Configure specific customer messages or phrases that should always trigger a handoff.
Examples:
"I need more help."
"This didn't solve my problem."
"I want a refund."
"Cancel my order."
This is useful when you know certain requests should always be handled by a human agent.
You can also choose which channels each trigger applies to.
Conversation Length
Set the maximum number of AI messages allowed before a conversation is escalated.
If the conversation reaches this limit without being resolved, it is automatically handed off to your support team.
This helps prevent customers from getting stuck in long conversations with the AI.
You can configure different limits for each supported channel.
Best practices
To provide the best customer experience:
Reserve Customer Intent for requests that always require human intervention.
Use Trigger Messages for phrases that customers commonly use when they expect human support.
Set a reasonable Conversation Length limit to avoid unnecessarily long AI conversations.
Frequently asked questions
Which rule is applied if multiple rules are triggered?
The conversation is escalated as soon as the first applicable rule is detected. The AI does not wait for additional rules to be triggered.
Can I disable the default system rules?
No. The default rules for Customer asks to speak to a human and Customer sentiment are built in to ensure customers can always reach a human agent when needed.
Do escalation rules apply to all channels?
Yes. Escalation rules apply across all supported channels. For rules that support channel-specific configuration (such as Trigger Messages and Conversation Length), you can choose which channels the rule applies to.
Can AI continue responding after an escalation?
No. Once a conversation is escalated, the AI stops responding and your support team takes over.