Skip to main content
Escalation is the process by which your AI employee recognizes it has reached a decision point that requires a human and hands the conversation or task off accordingly. It is one of the most important parts of how your AI employee operates. Without clear escalation rules, an AI could attempt to handle situations it isn’t equipped for — and in a business context, that creates risk. With well-defined triggers, you get the best of both worlds: the AI handles the volume, and your team steps in only when it genuinely matters. You control what those triggers are.

​
How escalation works

When your AI employee encounters a situation that matches one of your escalation triggers, it follows a fixed sequence:
  1. Recognizes the trigger — The AI identifies that the conversation or task has hit a defined threshold or condition.
  2. Pauses the interaction — Rather than continuing, the AI stops and holds the conversation in a safe state. The customer is acknowledged and told that someone will follow up.
  3. Notifies the right person or team — An alert is sent to the designated contact — whether that’s a specific person, a team inbox, or a support queue — based on the rules you set.
  4. Provides full context — The handoff includes everything the receiving person needs: a summary of the conversation, what triggered the escalation, and a suggested next step so no time is wasted getting up to speed.

​
Common escalation triggers

Specific financial limits that require human approval. For example: escalate any refund request over 200,orflaganyinvoiceabove200, or flag any invoice above 10,000 for manual review before processing. You set the thresholds — the AI enforces them automatically.
When a customer’s tone indicates frustration, distress, or explicit dissatisfaction, the AI recognizes it and escalates rather than continuing to engage. Examples include the use of complaint language, expressions of anger, or phrases that suggest the customer is close to churning.
Certain subjects are always handled by a human, regardless of how they arise. Legal matters, regulatory questions, formal complaints, and anything that could carry liability are common examples. When these topics appear, the AI routes the conversation immediately without attempting to respond.
If your AI employee’s confidence in a response falls below a defined threshold — meaning it isn’t sure it has the right answer — it escalates rather than risk giving incorrect information. This is especially useful for edge cases or unusual requests that fall outside the AI’s normal operating range.
If a conversation or task hasn’t reached a resolution after a set number of interactions, the AI flags it for human review. For example, if a customer issue hasn’t been resolved after three exchanges, it is automatically escalated so it doesn’t get stuck in a loop.

​
What you receive during an escalation

When your AI employee triggers an escalation, the person receiving it gets everything they need to take over immediately:
  • Context summary — A plain-language summary of what the conversation or task was about and why it was escalated.
  • Full conversation history — Every message or action that took place before the escalation, so the receiving person can read the full picture in seconds.
  • Suggested next step — Based on the trigger type and the context, the AI provides a recommended action to help your team respond quickly and consistently.

​
How to update escalation rules

Escalation rules are set during the role mapping phase, but they are not permanent. As your business evolves — or as you see how your AI employee performs in practice — you may want to adjust the triggers. To update your escalation rules, contact your Jobr account manager or submit a request through the standard update process. Changes are documented and applied to your AI employee’s configuration before they go live.
Start with broader escalation triggers and tighten them over time as you see what the AI handles well. It’s always safer to escalate more at the beginning and reduce it as confidence builds — rather than the other way around.
Escalations are real-time — make sure the right person is available to receive them during business hours. An escalation that reaches an unmanned inbox defeats the purpose of the handoff. Review your notification routing before your AI employee goes live.