15 AI agent prompts for Ecommerce
Copy-paste, beginner-friendly prompts to help you recommend products, recover carts, answer sizing questions and write listings. Each one is based on a real-world example and works with ChatGPT, Claude or an AI agent. 8 are free below, and 7 more are in AgentPlaybook Pro.
Free Ecommerce prompts
Size & fit advisor
Answer sizing and fit questions to cut returns, like Hang Eleven's assistant.
Size chart: [paste]. Fabric/fit notes: [paste]. A shopper says: '[usually wears M in X brand, height, preference]'. Recommend a size, explain why, and mention the return policy briefly.
Abandoned cart follow-up
Write friendly follow-ups that offer help, not just discounts.
Write 3 abandoned-cart messages for [product] for WhatsApp/email: (1) 1 hour later: helpful reminder + fit/question offer, (2) 24h: social proof, (3) 72h: small incentive. Keep each under 60 words.
Review collection & routing
Collect reviews after delivery; publish happy ones, route unhappy ones to a human.
Draft a post-delivery review request message that asks every customer for an honest review and includes our public review link. Then write follow-up rules: everyone gets the same review link, whatever their rating; if a customer rates us 3 or lower or mentions a problem, also apologize, ask what went wrong, and create a task for the team with the order details. Don't offer discounts or gifts for reviews, and don't send the link only to happy customers (Google's review policy forbids selectively asking for positive reviews).
Proactive order updates
Tell customers about shipping and delays before they ask.
Write message templates for: order confirmed, shipped (with tracking), delayed (with apology and new date), delivered, and return received. Friendly tone, under 40 words each.
Move stock between stores and clear slow sellers
Have an agent read sales and stock across your locations, then prepare transfers toward the stores where items sell, clearance tags for items that don't move, and draft purchase orders, all waiting for your approval. Lightspeed added this to Lightspeed AI for retailers in an opt-in beta, with every action previewed first and logged.
You are my retail inventory planner. Data: stock on hand by location [paste], sales by item and location for the last [8] weeks [paste], and open purchase orders [paste]. Do three things and show each as a proposed action for me to approve, not a done deal: (1) transfers: for each item that is overstocked in one store and selling fast in another, propose how many units to move and why; (2) clearance: list items with no sales in [X] weeks and more than [Y] units on hand, with a suggested markdown; (3) reorders: draft a purchase order for items that will run out within [Z] days at current sales, grouped by supplier. Keep it to a table I can scan in five minutes, and list any data you think is wrong.
Let a shopper's AI agent find products and check out with Shop Pay
Connect an AI agent to Shopify's Shop MCP so it can search a billion-product catalog, build a cart, and finish checkout with the shopper's saved Shop Pay details after they confirm the total. Shopify launched Shop MCP so agents no longer have to bounce buyers to a storefront just to pay.
You are my shopping agent. I want: [describe product, budget, size/color, must-haves]. Search results from the catalog: [paste search hits with store, price, stock, product id]. Step 1: shortlist up to 3 options that match my must-haves and explain why in one line each. Step 2: ask me which one to buy and confirm size/options before any cart. Step 3: once I pick, draft the cart/checkout summary you will create (merchant, items, quantity, price, currency) and wait for me to say 'approve'. Step 4: after I approve, outline the checkout steps (create cart → create checkout → complete_checkout with my chosen payment) and remind me that you will not pay until I confirm the final total and payment method. Never invent stock, prices, or complete a purchase without my explicit OK.
Get a monthly keyword ranking report for each Amazon product in minutes instead of hours of exports
Amazon and Walmart sellers and the agencies that manage them spend days each month exporting keyword rankings and search volumes into spreadsheets to see what's slipping. An AI agent can pull those numbers together for each product and point out what changed and what to do next. Helium 10 reported on Oct 21, 2026 that ecommerce agency CoreTrex now asks its Helium agent for a ranking-shift and search-volume report per product (ASIN), cutting each report from 30 to 45 minutes down to 5 to 10, about 80% less time, so the team reviews keywords far more often and adjusts targeting and ad spend faster.
I sell [product] on [Amazon / Walmart], product ID [ASIN or item ID]. Here's my keyword data for the last [3] months: [paste export with keyword, rank by date and monthly search volume]. Here's my ad spend by keyword if I have it: [paste or 'none']. 1) Show the top 20 keywords by search volume with my rank then and now, and mark each as rising, falling or steady. 2) Call out the 5 biggest drops and the 5 biggest gains and a likely reason for each, labeled as a guess. 3) List high-volume keywords where I rank poorly or not at all that fit my product. 4) Recommend what to change in my title, bullets and backend keywords, and where to raise, lower or pause ad spend. 5) Finish with a 5-line summary I can send a client or partner. Use only the numbers I gave you.
Count your stock from photos and get a reorder plan: snap your shelves or bins, let AI build the inventory sheet and suggest what to reorder
Counting merchandise by hand and working out what to reorder is the job small sellers dread, so it gets put off until something sells out. You can photograph your stock, have an AI assistant identify and count each item and size, turn it into a spreadsheet, compare it with what sold, and suggest a reorder list you check before sending anything to a supplier. OpenAI reported (Oct 2026) that Larissa Guetter of ATV Big Air Tour, a two-person touring motorsports show, uploads photos of the tour's merchandise to ChatGPT Work, which organizes the inventory, builds a spreadsheet and reorder recommendations in under 15 minutes; with her review and adjustments, inventory-to-reorder now takes two to three hours instead of two to three days.
You are my inventory assistant. I'm uploading photos of my stock: [upload clear photos of each shelf, bin or rack; one photo per area, with labels or tags visible where possible]. My product list: [paste item names, sizes or variants, SKU if any, cost, supplier, supplier minimum order and lead time]. Sales since my last count: [paste or upload, or say 'none']. Upcoming events or busy periods: [e.g. 6 shows in the next 8 weeks / holiday sale]. Do this: 1) From the photos, list every item and variant with your count, and mark any count you're unsure of (stacked, hidden or blurry) so I can recount it by hand. 2) Build an inventory table: item, variant, counted quantity, confidence, photo it came from. 3) Using the sales and upcoming periods, estimate how many weeks each item will last. 4) Suggest a reorder list per supplier with quantities, respecting minimums and lead times, and point out slow sellers I should not reorder. 5) Give me the total reorder cost. Don't place any order; I'll review and send it myself.
More in AgentPlaybook Pro
- Product recommendation assistantGuide shoppers to the right product with clarifying questions, like SharkNinja's agent.
- Product description writerWrite product listings that sell and rank.
- Turn product photos into a priced marketplace draftDrop photos of items into a watched folder and have an agent research comps, propose a price, and draft the listing and social posts, like Tommy Stark's Grok Bot eBay sports-card workflow (NovaBay) that picks up iCloud photos, prices, lists, and drafts reels so selling is basically photograph then ship.
- Give store associates an aisle-side project advisorEquip floor staff with a conversational project advisor so any associate can answer cross-department questions in minutes, like Lowe's Mylow Companion on OpenAI, used by roughly 300,000 associates and linked to 3x online conversion for customers using Mylow versus those who don't.
- Clean up your product catalog with a small team of AI checkers and a final human sign-offProduct names, categories and specs drift into a mess as a catalog grows. Split the cleanup into small jobs (naming, category, spec extraction, banned-character check), have each step check the others, and approve the final version yourself. On Oct 22, 2026, Fujitsu announced it built this for MISUMI's e-commerce catalog of tens of thousands of products: nine AI agents with separate roles review each other's output, a person gives final approval, and the system went live in October after about 2.5 months of work.
- Every morning, compare your store's order list against what your automations actually received, so no order silently misses fulfillment or follow-upMany stores pass each new order to automations (fulfillment, invoices, thank-you emails, spreadsheets) through webhooks, and when a webhook drops or the automation tool restarts or times out, the order simply never arrives, with no alert or error. An AI assistant or a small scheduled workflow can pull yesterday's orders straight from the store, compare them with your automation's log, and list any order that never made it through so you can push it manually. TrenoxaHQ shared on X (Oct 22, 2026) a free n8n workflow that does exactly this every morning, checking Shopify's own order list against the workflow's log because Shopify webhooks can fail silently.
- Set a new-customer ad target that actually makes money: have an AI work out your break-even ROAS from real margins and repeat purchases, then brief your agencyMany online brands judge ads by a ROAS number their agency picked, without checking it against real product margins, shipping and fees, or how often new customers come back, so they can grow revenue while losing money on every new buyer. An AI assistant connected to your reconciled financial data can work out what return on ad spend you need on first orders to be profitable over the next 90 days, write the brief for your agency, and then check each week whether you're hitting it. Lioranpi shared on X (Oct 22, 2026) a weekly 'Analyze. Act. Lock In.' playbook using Finaloop's financial data connected to Claude: ask what new-customer ROAS target maximizes profit over 90 days, send that target to the agency with the reasoning, and then check whether you're hitting it or spending profits away.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free