Customer Support Questions You Should Automate With AI — and the Ones You Should Not

Good automation category 1: published FAQs

Examples:

• What are your opening hours?
• Do you ship internationally?
• What payment methods do you accept?
• Where can I find your size guide?
• How do I book?

These work well when the answer is already documented and does not change from customer to customer.

Good automation category 2: policy explanations

Examples:

• What is the return window?
• Are sale items returnable?
• How do exchanges work?
• What is the shipping policy?

The chatbot should explain the policy, not approve exceptions to it.

Good automation category 3: product and service discovery

Examples:

• Which product is suitable for this use?
• What is the difference between these two options?
• Which service should I read about?

This works when the underlying product or service information is complete and current.

Good automation category 4: routine status lookups

On platforms with a secure integration, order-status questions can be a strong automation use case.

Farnsla’s Shopify and Wix ecommerce flows can require email verification before returning available order status and tracking information.

That is very different from a chatbot simply guessing where an order might be.

Good automation category 5: navigation

Many support questions are really navigation problems.

“Where is pricing?” “How do I book?” “Where can I see the return policy?”

A chatbot can answer briefly and direct the visitor to the right page.

Questions that should usually stay with humans

Refund or replacement decisions

The chatbot can explain the standard policy. A person should approve an exception or decide what remedy is appropriate.

Complaints

AI can acknowledge and collect context, but emotionally charged or relationship-sensitive situations often need human judgment.

Custom pricing

Do not let the assistant invent a quote unless you built an explicit pricing workflow with approved rules.

Manual account or order changes

The chatbot should not claim it changed an address, canceled an order, or modified an account unless the system actually performed that action.

Anything with missing source information

If the answer is not in your reviewed business knowledge, guessing is not automation — it is risk.

A simple test

Before automating a question, ask:

1. Is the answer documented?
2. Is the answer the same for most customers?
3. Can the chatbot retrieve the required information securely?
4. Does the answer require business authority?
5. What happens if the AI is wrong?
6. Is there a clear human path?

If the question fails those tests, keep it with a person or redesign the workflow first.

How Farnsla fits the model

Farnsla can automate FAQs and policy answers from merchant knowledge, recommend products on supported ecommerce platforms, handle secure routine order-status questions, and preserve conversation context for follow-up.

Shopify also supports Customer Messaging for direct messaging or AI-to-human handoff through supported configured apps.

The objective is not to maximize the percentage of conversations answered by AI. It is to reduce repetitive work while keeping important customer decisions with the people responsible for them.