Explore Chatropic

What happens when the AI cannot help?

Keep a clear path to your support team when a request needs human judgment.

2 min read · Updated 7 September 2026

Human handoff gives customers a path to your team when a request needs judgment, additional access, or help beyond the agent’s configured responsibility. It should be designed as part of the support workflow from the beginning.

A handoff is only useful if someone owns the next step. Configure the behavior, decide who will respond, and set expectations that match your actual staffing policy.

Decide which requests need a person

Start with situations your team already treats differently: disputed payments, policy exceptions, identity problems, missing information, or a failed operation that the agent cannot safely retry.

Describe what the agent should do in each case. It may need to ask for clarification first, explain a limitation, or immediately follow the escalation path. Avoid vague instructions that leave both the customer and the receiving team uncertain about what happens next.

Preserve the work already done

In the Builder, Conversations provides the transcript and information such as channel, identity, routing, and tool evidence. Reviewers can see what was asked and what the agent attempted.

That context helps a person continue the request without unnecessarily repeating the same questions. Check the actual information available in your configured channel rather than promising that every external system will receive an identical record.

Walk through a disputed delivery

Imagine a fictional customer saying, “The order says delivered, but I have not received it.” A lookup can report the system status, but that status alone does not resolve the dispute.

The agent should follow the configured dispute policy, preserve the attempted lookup, and route the case to the appropriate person. Its customer-facing response should explain the next step without claiming that the customer is mistaken or promising an unsupported response time.

Make sure your team can reply

Human agents can reply when handoff is enabled. Before rollout, test assignment, the human response, and what the customer sees while waiting.

Use a real operating policy for availability. If the team is not staffed continuously, the agent should not imply that a person is already responding. Configure an accurate expectation and ensure the team knows how to find and continue the case.

Include handoff in your quality review

Test a request that should be handled by the agent and one that should be escalated. An unnecessary handoff creates work; a missed handoff can leave a sensitive issue unresolved. Review both outcomes when improving instructions.

Your next step

Read the handoff walkthrough