E-commerce
June 28, 2026
An e-commerce chatbot can resolve many simple questions, but it should not automate everything. Certain requests require human verification, a business decision, strong empathy, or a secure procedure.
The right chatbot knows how to answer when the source is reliable and transfer when the risk increases: payment, fraud, personal data, dispute, health, complex warranty, or a very unhappy customer.
This guide shows what is best not to automate in order to protect the customer experience.
Summary
Why define limits before launching?
Without limits, the chatbot can give an overly confident response on a subject it does not master. An error regarding a deadline is annoying; an error regarding a payment, personal data, or a guarantee can become a real dispute.
Defining limits allows for confidently automating simple cases and escalating sensitive ones.
A reliable chatbot is not one that answers everything; it is one that knows when to stop.
Convert over 2,000 customers on average per month with Qstomy.
The world’s 1st Shopify AI dedicated to customer conversion
Empowering 200+ e-commerce merchants
Which topics should we be cautious about?
High-caution subjects include disputed payments, double charges, fraud, hacked accounts, data deletion, complex warranties, carrier disputes, out-of-policy commercial gestures, health, product safety, and highly personalized advice.
The bot can collect context and explain the procedure, but it must not decide on its own.
Which requests can be automated?
Simple and well-sourced inquiries are suitable: order tracking, return status, support hours, general policy, invoice, availability, size guide, frequently asked product questions, or simple checkout help.
Even then, transfer must remain possible if the customer does not understand or if their case falls outside the scope.
How to detect when someone is out of frame?
The chatbot must identify signals such as human request, frustration, repetition, contradictory evidence, high amount, urgency, sensitive data, or missing source.
These signals must interrupt the automated path before it worsens the situation.
How to prepare for the transfer?
The transfer must contain a useful summary: request, data collected, source consulted, response already given, blockage, urgency, and customer expectation. The customer must not have to start over from scratch.
A good transfer is part of automation, not its failure.
This preparation also reduces frustration on the agent side: the person taking over already knows what the customer asked for, what the bot attempted, and why the case must leave the automated path.
Which flow to follow?
The flow must separate simple, sensitive, and unknown.
Classify reasons by risk: low, medium, sensitive, regulated, or emotional.
Automate only requests with a reliable source, clear rule, and low consequence.
Collect context without deciding for payment, fraud, data, dispute, and health.
Trigger transfer if frustration, repetition, contradictory proof, or missing source appears.
Regularly review conversations to adjust limits, sources, and escalations.
Which messages should be used?
For limit: "This request requires human verification before a decision."
For security: "I can collect the context, but I will not ask you for passwords or sensitive banking data."
For transfer: "I am forwarding the summary to avoid making you repeat yourself."
When to transfer?
Transfer is necessary as soon as a financial, legal, privacy, security, health, complex warranty, or strong emotional decision is detected.
The bot must transmit context, evidence, status, source consulted, risk, emotion, and requested action.
Which KPIs should be monitored?
Track sensitive transfers, avoided errors, reopenings, requests for human intervention, satisfaction after transfer, incidents on automated topics, and time saved on simple cases.
These KPIs show whether the limits are truly protecting the customer.
Which mistakes should be avoided?
Avoid aiming for 100% automation, hiding the handoff, responding without a source, dealing with strong emotion using a script, or deciding alone on a sensitive topic.
The chatbot's limit must be visible and assumed.
How can Qstomy help?
Qstomy can connect the chatbot to Shopify, conversations, orders, carts, contact pages, support KPIs, conversion data, and escalation rules to answer clearly, then transfer sensitive cases with an actionable summary.
The chatbot helps the customer move forward without inventing a result, an ROI, a resolution, a Shopify action, or an automated decision that still needs to be confirmed by a reliable source.
Explore AI support, the AI sales agent, or request a demo.
Key takeaways
Key takeaways
A chatbot should not automate high-risk topics on its own: payment, fraud, privacy, litigation, health, security, and major exceptions.
What the customer must understand
The customer must receive quick help when it is simple and a human takeover when necessary.
The right limit of the chatbot
The chatbot can collect and summarize, but it must hand over as soon as the risk or uncertainty increases.
Enzo
June 28, 2026
