AI AUTOMATION · MODULE 01 OF 08 | breadcrumb: Home > Learn > AI automation > Modules > 01 What AI can and cannot do
Module 01: What AI can and cannot do in ecommerce
AI in ecommerce can draft, sort, forecast, match and alert at volume, and some tools can act within limits you set. It cannot know facts it was not given or own what it says. Knowing which is which before you buy a tool saves months on the wrong project.
Quick answer
In ecommerce, AI can draft, summarize, classify, forecast and alert across most repeat store work, and some tools can act within limits you set. It cannot know facts it was not given, take responsibility for what it says or make policy decisions. People keep prices, large refunds, appeals and policy.
What kinds of automation are there?
Three. Rules-based workflows follow a rule you wrote ("when an order over $300 arrives, tag it for review"). AI-assisted work drafts or sorts and a person approves (a reply to "where is my order?" with live tracking filled in). Agentic AI plans steps and acts across tools within limits (adjusting a price inside a floor and ceiling). Each fails differently: rules repeat a wrong rule perfectly, AI-assisted drafts can be confidently wrong and agents can act on bad data quickly.
What can AI do well in ecommerce today?
Language and pattern work at volume. Drafting listings from your product data: Amazon's tools draft from an image, a URL or a spreadsheet, and sellers review, edit and approve (sell.amazon.com, Nov 3, 2025). Drafting support replies from order data. Sorting tickets and reviews by topic. Forecasting demand from history and seasonality. Spotting mismatches between two data sets, such as a 3PL invoice and your orders. Watching dozens of metrics and alerting a person when one moves.
Each platform now builds some of this into its own seller tools. In every case, the seller stays in control:
| Channel | What its own AI does | Who decides |
|---|---|---|
| Amazon | Drafts listings from an image, a URL or a spreadsheet (sell.amazon.com, Nov 3, 2025); Seller Assistant surfaces insights, drafts actions or acts within limits the seller sets (Amazon Accelerate 2026 recap, Sep 29, 2026) | The seller approves listings and chooses the assistant's mode |
| Walmart | AI-powered single and bulk item setup and a 24/7 Smart Assistant for sellers (Walmart Marketplace, Aug 25, 2025) | The seller reviews every attribute |
| Shopify | Sidekick analyzes data, edits products and drafts content, and "presents changes for your review before applying them"; workflows it drafts in Flow start inactive (Shopify Help Center, read Oct 20, 2026) | The merchant applies changes and turns workflows on |
The pattern matters more than any single feature: the platforms build AI that drafts and suggests, and they leave the decision with you. Plan your own automation the same way.
What can't AI do?
Four things matter most.
- It cannot know what it was not given. If your stock count is wrong, it will sell stock you do not have.
- It cannot take responsibility. What your AI says, you said. A Canadian tribunal held Air Canada responsible for misleading bereavement-fare refund information its chatbot gave a customer (Moffatt v. Air Canada, 2024 BCCRT 149, Feb 14, 2024).
- It cannot decide policy. Whether you refund a late order, accept a return after 30 days or appeal a marketplace decision is a business decision.
- It cannot create demand. Automation saves time on work you already have. It does not make a product sell.
Why do so many AI projects disappoint?
In McKinsey's latest survey, nearly nine in ten respondents report regular use of AI in at least one business function, but AI high performers are just 6 percent of respondents (McKinsey, "The state of AI," Aug 25, 2026). In stores the gap is usually data, hand-offs and ownership, not the model. That is why the next modules start with mapping and data, not tools.
Where automation stops
Let AI do the busywork. Three actions always wait for a person:
- 01Price changes outside the limits you set.
A pricing mistake can sell your stock at a loss in minutes.
- 02Refunds above a set amount.
Small ones can be automatic. Big ones get a second look.
- 03Anything touching account health.
Appeals and policy replies are written and checked by people.
We add three more to every build: account appeals and reinstatement, policy decisions (what your store promises customers) and any situation nobody has written a rule for. Everything else: automate it, log it and review it every week.
Five guardrails sit under every workflow: limits on every action, approval for anything risky, only the access each tool needs, a log of every step and one person who owns the outcome. What your AI says, you said.
Key takeaways
- Three types: rules, AI-assisted and agentic. Use rules wherever you can write the rule.
- AI is strong at drafts, sorting, forecasting, matching and alerts.
- AI cannot know missing facts, own its words, set policy or create demand.
- Prices outside limits, refunds above a set amount and anything touching account health always wait for a person.
Do this now
- List three tasks you do every day that follow a rule you could write in one sentence.
- List three tasks that involve writing or sorting text (replies, listings, reviews).
- Write down your refund limit and your price floor for your top product. If you do not have them, that is your first job.
- Read Module 02.
Tool for this lesson: Shopify Flow (free on Basic, Grow, Advanced and Plus; Shopify Help Center, Oct 20, 2026) as the simplest place to see a rule-based workflow; Sidekick to see AI-assisted drafting that "presents changes for your review before applying them."
Ecom Relays is not paid by and not affiliated with any tool listed. We name widely used tools so you know your options. Amazon, Walmart and Shopify are trademarks of their owners; Ecom Relays is not affiliated with them.
Most new stores take months to find steady sales, and some never do. Inventory, platform fees, freight and ad spend are paid by you. Automation saves time on repeat work. It does not create demand for a product.
A note on AI claims: the FTC has warned businesses not to exaggerate what their AI can do (FTC business guidance, "Keep your AI claims in check," Feb 27, 2023). We describe each tool from its own documentation, label vendor figures as vendor estimates and never say that AI will produce sales for you. AI here means software that drafts, sorts, checks and alerts. People still own the decisions.
The problem this lesson fixes
-
In owners' words:
"We keep hearing AI will fix this."
-
Cause:
nobody has written down the rule the AI is supposed to follow.
-
Fix:
write the rule first, in one sentence, before choosing any tool (Module 02 shows how).
-
Measure:
hours a week spent on the task, before and four weeks after.
Talk it through with a real person
Book a free 30-minute session with an ops lead. Bring one workflow you want off your plate, and we will tell you plainly whether to automate it, how and where it should stop. A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday. No pitch deck.
A written plan within 3 business days. No sales call required.
Free
Ask a real expert. Free.
A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday.
Request received.
That did not send.
You can reach an ops lead directly, and nothing you typed is lost.
Can AI take over running my store?
No. It can draft, check and alert across most daily work, but prices outside your rules, large refunds, appeals and policy decisions need a person. Marketplaces hold you responsible for what your tools do, and customers hold you to what your chatbot says, so someone in your business has to own every automated decision.
Is AI the same as automation?
No. Most store automation is rules with no AI at all: when an order ships, send tracking; when stock falls below a number, send an alert. AI adds drafting, sorting and pattern work on top. Start with rules wherever you can write the rule down, and add AI where the work is language.
Which is safer, rules or AI?
Rules are more predictable: the same input always gives the same output, and the log shows exactly what happened. AI is useful where the work is language or judgment, with a reviewer, because its output varies and can be confidently wrong. Use both, each for the work it suits.
Will AI increase my sales?
We cannot promise that, and you should be wary of anyone who does. AI reduces time on repeat work and catches errors sooner. Whether that shows up in sales depends on your product, price, costs and demand. Measure the task you automated, before and after, with the same method.
What is agentic AI?
AI that plans steps and takes actions across tools within limits you set, such as Amazon's Seller Assistant when it acts within seller-set limits. It is the most powerful kind and the riskiest, because it can act on bad data quickly. Use it only where the limits can be written as hard numbers and every action is logged.
Where should I start?
With Module 02: list your repeat tasks for one normal week before choosing any tool. Most stores find their first automations are simple rules, not AI. The full path, from first automations to governance, is in our free AI automation guide.
Sources8 sources
- 01help.shopify.comchecked Oct 20, 2026Primary
- 02help.shopify.comchecked Oct 20, 2026Primary
- 03sellingpartners.aboutamazon.comchecked Oct 20, 2026Primary
- 04sell.amazon.comchecked Oct 20, 2026Primary
- 05decisions.civilresolutionbc.cachecked Oct 20, 2026Primary
- 06marketplace.walmart.comchecked Oct 20, 2026Primary
- 07ftc.govchecked Oct 20, 2026Primary
- 08mckinsey.comchecked Oct 20, 2026Primary
Free and ungated. No sign-up wall. Ecom Relays is not a course or coaching program. We do not sell stores and we never promise income.
Next step
Ask a real expert. Free.
A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday.