Live chat vs chatbot for ecommerce: which questions should stay human?

The short version

• Automate stable facts such as published policies, basic product details, and authenticated order status.
• Keep judgment-heavy recommendations, exceptions, complaints, and money-changing actions human.
• Require a visible handoff when confidence is low or the customer asks for a person.
• Review the transcript, outcome, and fulfillment before claiming automation resolved the issue.

Good automation is narrow and observable

A chatbot is strongest when the answer comes from a trusted source, the request is easy to classify, the action is reversible, and failure routes cleanly to a person. Shipping-policy basics and simple order-status lookups often fit.

Do not let a fluent answer hide weak evidence. The bot should identify the source or limitation, avoid inventing availability or delivery promises, and preserve the conversation when a teammate takes over.

Keep these conversations human

• Product advice where fit depends on personal context
• Refunds, cancellations, partial refunds, and policy exceptions
• Damaged, missing, unsafe, or emotionally charged orders
• Fraud-sensitive account or address changes
• Any case where the system data conflicts with the customer's evidence

These requests combine judgment, authority, and trust. Automation can gather context, but a named person should own the decision.

The small-team hybrid

Let self-service answer the stable question, let chat collect the missing context, and let a human approve the consequential action. That is more useful than pretending every conversation should be fully automated.

WooCommerce's own marketplace guidance for AI chat tools describes a fallback to a ticket or live chat when the answer is outside the system's knowledge. That fallback is part of the product, not an admission of failure.

FAQ

Recommended next steps

FAQ