Managing Escalations

Escalations decide when an issue moves from the AI agent to a human team.

Good escalation setup keeps the customer experience smooth, makes ownership clear, and prevents your team from being flooded with work the AI could have resolved safely.

What an escalation does

An escalation moves a conversation into a human-handled workflow when the AI should not finish the issue on its own.

This usually happens because:

The customer asked for a person.
A policy or workflow requires human review.
The issue is sensitive, risky, or unclear.
A configured path tells the system to hand off.

When escalation should happen

Use escalation when the team wants human judgment, approval, or follow-through.

Common examples:

Refunds or billing exceptions
Account access issues
Complaints or high-risk cases
Edge cases with incomplete information
Cases that need another team to investigate

Avoid escalating too broadly. If too many routine issues escalate, your team loses the benefit of automation and the queue becomes harder to manage.

Escalation timing

Your workspace may support different timing approaches for escalation.

Immediate handoff

Use this when a case should go to a human as soon as the trigger is met.

Wait-based handoff

Use this when the AI should try first, then hand off if the issue still is not resolved after a defined wait.

SLA-aware handoff

Use this when human timing targets matter and escalation needs to work with the way your team measures response expectations.

What the customer experiences

During a good escalation, the customer should feel like the conversation continued naturally. They should not have to restate everything from the beginning.

That means your escalation setup should preserve:

The original question
The conversation so far
Relevant customer context
Any work the AI already attempted

What your team experiences

For the human team, escalation should make it obvious:

Why the case escalated
Who should take it
How urgent it is
What still needs to happen next

This is where routing, groups, and ticketing Connectors matter. Escalation works best when the handoff lands in the right place with the right context.

When to use a custom Flow

Use a custom Flow for escalation when the handoff needs more than a simple rule.

Examples:

Collect required details before handoff.
Route differently based on the issue type.
Send the case to a specific team under specific conditions.
Combine escalation with another operational step.

If the handoff logic is becoming more than a simple “if this, escalate,” a Flow is usually the cleaner option.

Best practices

Start with a small number of clear escalation triggers.
Escalate early for high-risk issues.
Keep ownership obvious after handoff.
Test real conversations before relying on new rules at scale.
Review escalation volume regularly to catch over-escalation.

Common mistakes

Too many cases escalate

This usually means the rules are too broad or the AI does not yet have enough Knowledge for the issue types it sees most often.

Escalations land in the wrong queue

Review routing groups, customer segments, and any related help desk setup.

Customers still repeat themselves

Review what context is visible at handoff and whether the escalation path is collecting the right information before transfer.