Set up human support from escalation to resolution
Configure support channels and handoff rules, prepare operators, test takeover, and return completed conversations to AI.
Human support in Humind connects Agent escalation settings with the operational Inbox. A shopper may be offered configured support channels, enter the human queue, provide information before handoff, receive an operator reply, and later return to AI after the operator completes the exchange.
This guide covers the entire chain so a configuration does not promise support the team cannot deliver. It also explains operator step-in, internal notes, satisfaction collection, and ticketing boundaries.
How this fits into Humind
Inbox is the operational record of customer conversations. It combines message history, shopper and product context, internal collaboration, human-handoff state, tags, and tickets. Customer replies and internal notes have deliberately different visibility.
A consistent team routine matters more than any individual filter. Agree when to take over, when to leave an internal note, when to create a ticket, and when to return a conversation to the AI so ownership remains clear to every operator.
Before you start
Access: Agent configuration access is required for Escalation. Operators need Inbox and Inbox handoff permissions; team and ticketing administration may require additional permissions.
- Define staffed channels, expected response behavior, and ownership hours outside Humind.
- Invite operators and confirm their Inbox and Inbox handoff permissions.
- Prepare a shopper test case that should escalate and one that should remain with AI.
Step-by-step workflow
Configure support channels
Open AI Agent, then Escalation. Add the support channels the company actually operates, give each a correct destination, enable only active channels, and arrange their order. The customer-facing preview helps review the choices before testing.
A channel is a promise to the shopper. Remove or disable stale contact details instead of leaving a path that no team monitors.
Define human-handoff rules
Enable human handoff when operators should handle live requests. Add natural-language include criteria for cases that should escalate and exclude criteria for cases that should remain automated. When several include criteria exist, choose whether any or all must match.
Add only the pre-handoff questions operators need. The collected answers appear in conversation context, so they should reduce repetition without asking for unnecessary information.
Choose operator and satisfaction behavior
Operator step-in controls whether a teammate may reply to an AI-active conversation and take it over. If it is disabled, the backend rejects that step-in rather than silently changing the policy. Satisfaction can be collected after AI conversations and, when handoff is enabled, after human support.
Operator profile name, image, online state, and availability affect the live-support experience. Ask each operator to review their own profile before launch.
Prepare Inbox and ticketing
Agree who watches queued conversations and how tags, notes, mentions, canned responses, and tickets are used. A customer reply can take over a queued conversation. An internal note stays invisible to the shopper and does not take over the thread.
Humind ticketing can be enabled by an administrator. Gorgias and Zendesk credentials are also shown in the administrator-only Escalation integration area, while the general Integrations marketplace labels those integrations as early access. Confirm the intended provider and tested behavior before relying on external ticket creation.
Test escalation, takeover, and hand-back
As a shopper, trigger the include case and confirm pre-handoff questions and queue state. From Inbox, review the collected context, send a customer reply, and confirm the thread becomes human-active. Add an internal note separately to verify it is not customer-facing.
When the exchange is complete, use the hand-back action and confirm the dialog. The thread returns to AI-active for future messages. Also test the exclude case and any offline or availability expectation the team communicates.
Permissions and important caveats
- Human handoff should not be enabled until operators, permissions, availability, and queue ownership are ready.
- Operator step-in from AI-active follows the configured policy and can be rejected when disabled.
- Internal notes and mentions are teammate-only; customer replies are visible to the shopper.
- External ticketing providers require separately verified credentials and operational ownership.
Verify the result
Use this checklist before considering the work complete:
- Enabled support channels point to staffed, current destinations.
- Include, exclude, and pre-handoff behavior match the written support policy.
- An authorized operator can take over, collaborate internally, and return the thread to AI.
- Queue ownership, availability, satisfaction, and ticketing expectations are documented for the team.
Troubleshooting
The shopper cannot request a human
Confirm human handoff and a valid support path are enabled, the scenario meets include criteria and not exclude criteria, and the active Agent configuration is the one being tested.
An operator reply is rejected
Check Inbox and Inbox handoff permissions. If the thread is AI-active, confirm operator step-in is allowed. Refresh after a permission change so the session receives current access.
The conversation stays with the operator
Use the explicit hand-back action and confirm it. Sending a reply alone keeps the thread human-active unless the send menu action also resolves or hands it back.