How to Set Up Email Escalation for Your Shopify Store

Well-designed customer support automation isn't about replacing human judgment entirely. It's about removing the need for human judgment on the 90% of emails that don't require it, so you can focus on the 10% that do. Escalation is the mechanism that keeps those two categories correctly separated.

Without escalation, automation becomes a liability — the bot handles everything, including the things it shouldn't, and you only find out when a customer has already filed a chargeback or posted a public complaint. With escalation configured correctly, automation becomes safe: the routine is handled, and the hard stuff reaches you fast.

What should always be escalated

Some emails should never be auto-replied regardless of how good the bot is:

• Legal threats. "I'm going to contact my lawyer," "I'll be filing a complaint with the FTC," "I'm reporting this to my credit card company." These need a human response, and the wrong automated reply to a legal threat can make things significantly worse.
• Fraud reports or accusations. Customers reporting unauthorized charges, claiming they never placed the order, or accusing the store of scamming them require immediate human attention and often need coordination with your payment processor.
• Chargeback notices. If a customer mentions they've already initiated a chargeback or dispute with their bank, the window to respond is short and the response needs to be human and specific.
• Reports of completely missing orders. An order that never arrived after the full expected window is a potential dispute. These need investigation — checking with the carrier, deciding on a replacement or refund — not an automated reply.
• Requests that require manual Shopify actions. Address changes after an order has already been fulfilled, cancellation requests for orders that have shipped, requests to combine orders. These need someone to actually do something in the backend.
• Extreme complaints. A customer who is genuinely furious — using all-caps, multiple exclamation points, threatening public posts — needs a human to de-escalate. An automated reply to someone at that level of frustration almost always makes it worse.

What should NOT be escalated

Over-escalation is as much a problem as under-escalation. If everything gets forwarded to you, you've just built a more complicated version of checking your email. Common situations that don't need escalation:

• Order status and tracking questions (the bot can look these up and reply accurately)
• Return or exchange requests that fall within your stated policy
• Product questions the bot can answer from your instructions
• Shipping delay complaints where the order is still within the normal window
• Requests for discount codes or goodwill gestures (the bot can generate and send these)

If your bot is escalating things in these categories, the fix is usually to add more specific instructions — not to lower the escalation threshold further.

Setting up escalation in Chepa

Chepa has a dedicated escalation email field in the connection settings. When you enter an email address there, the bot will forward any email it judges as out of its depth to that address — with the full original email and any Shopify order context attached — instead of replying automatically.

The escalation address doesn't have to be the same as your support inbox. Many merchants forward escalations to a personal email or a shared inbox that only gets checked for urgent issues. The key is that it reaches someone who can respond quickly.

Beyond the escalation field, you can make the bot's escalation behavior more precise through your instructions. Examples of useful escalation rules to write explicitly:

Escalation rules written this precisely give the bot clear criteria for when to forward and what to say (or not say) before doing so. Vague rules — "escalate anything serious" — produce inconsistent results because "serious" is undefined.

The right mental model for escalation

Think of your escalation setup as a triage system, not a fallback. The bot handles everything it confidently can. Anything at the boundary of its confidence gets escalated with context. You get a curated inbox of genuinely difficult situations instead of a mixed pile of routine and urgent.

Merchants who are hesitant to automate customer support usually worry about the bot handling something badly. Escalation addresses that concern directly — the bot doesn't handle things badly, it escalates them. The setup work is defining exactly where that boundary is.

Frequently asked questions