Escalations start as chat conversations. Sometimes the agent asks a human for help because the issue needs judgment. Sometimes the shopper clicks “I'd prefer email instead” in the widget and turns the chat into a human follow-up. Merchants use this page to understand why the chat escalated, read the transcript, decide whether to reply, and turn the issue into tracked work if it needs more than one response. Even after the handoff, Garrio still uses the store's AI assistant to draft the follow-up for review.
The first question is whether the customer still needs an answer. The second is whether the answer should be an email, a task, or a manual operational follow-up outside the conversation.
This page should feel different from Email Thread because the work begins in chat, not in an email inbox. It is the slower, higher-judgment handoff the agent creates when it decides a person should step in, or when the customer explicitly asks to finish the conversation by email instead. The merchant is not dropped into a blank composer, though. Garrio still uses the same store-level AI assistant, knowledge, and settings to prepare the draft response.
michael@fieldsupply.co
Customer wants one order split across two shipments and asked for a manual delivery guarantee.
I need one part of order #1882 to go to our office and the other part to my home. Can you split it before it ships?
I can help with standard shipping questions, but splitting one order into two shipments needs a person to confirm the exact addresses and delivery promise. I’m bringing in a human teammate to follow up.
That works. I also need to know whether the office shipment can still arrive by Friday because it is for a trade-show setup.
To: michael@fieldsupply.co
Subject: We can help with that split shipment
For a one-time cancellation outside the normal policy window, Garrio can queue the cancel action and pause the composer on Review 1 pending action until the merchant decides whether to execute or skip it.
The shopper clicked 'I ’d prefer email instead' in the chat widget after describing a damaged gift order and wants a person to confirm how the return will be handled.
Jordan K. received a policy exception reply on Mar 9, 2026 10:05 AM.
Approved a one-time cancellation outside the policy window.
Repeated escalation reasons usually point to the next improvement: a missing Knowledge answer, a policy that needs clearer wording, or a queue flow that still needs a human checkpoint.
Start with the reason badge and transcript, not the draft. That tells you whether the real problem is a policy exception, a sensitive request, or an answer gap Garrio could not safely fill.
The handoff begins because the chat agent wanted a human's help, or because the shopper explicitly chose an email follow-up. Send from your configured support inbox (support@garr.io) only when the customer truly needs a reply outside chat. If the issue was handled elsewhere, use Mark as resolved. If the issue belongs to another owner or will take more than one reply, create or assign a task.
The draft is still produced by your store's AI assistant, not by a separate escalation bot. That matters because the response keeps the same brand voice, knowledge, and store context the assistant already uses in chat and in queued email drafts.
The best escalations are feedback for the system. If the same kind of issue keeps landing here, that is usually a sign to improve Knowledge, Settings, or the queue rules so fewer conversations need human rescue next time.
Once the human follow-up is sent from support@garr.io, later replies continue as email work instead of staying frozen in the escalations queue.