# Inventory Checks and Product Recommendations with a Chatbot > Learn how to connect an inventory and product recommendation chatbot to your store, covering catalog data, variants, recommendation rules and channel tests. _Mustafa Kuru — Kıdemli Yazılım Geliştirici · 2026-09-21 · https://palmate.ai/blog/inventory-check-product-recommendation-chatbot_ An inventory and product recommendation chatbot answers product questions using a store's stock and catalog data. When an item is unavailable, it can assess alternatives under defined category and price rules. Setup covers the store connection, data freshness, variant matching and channel testing. Order tracking is a separate workflow and needs its own record access rules. ## Key Takeaways - Stock questions use product information; order questions use order records. - Freshness depends on the source system and synchronization interval. - Recommendations should follow category, price and store-defined rules. - Test sold-out items, variants and connection failures before launch. ## What should be ready before setup? You need access to the store platform, product catalog and stock data. When connecting a system such as Shopify, iKAS or T-Soft, define the fields and permissions in scope. Consistent product codes, variants, warehouses and quantities make matching easier. Prepare management access for the website, WhatsApp or Instagram channel where the assistant will operate. Connecting a messaging channel does not automatically provide access to inventory. ## How do you build the workflow? ### 1. Verify the store connection Check that the integration can read the required product and stock information. Identify supported fields and action boundaries. The [integration options](/integrations) are a starting point for that assessment. ### 2. Define the stock source and refresh interval Decide where names, SKUs, variants and warehouse quantities come from. Whether updates are live or periodic affects what the assistant can tell a customer. > **Note:** Do not present periodically updated information as guaranteed live stock. More frequent queries may reduce delay, but the connection and failure scenarios still need testing. ### 3. Organize the catalog Prepare consistent product names, categories, prices and images. Match sizes and colors correctly. Inconsistent category labels can lead to irrelevant alternatives. ### 4. Separate stock and order questions Whether a size is available is a product-data question. Where a customer's order is depends on the order record. These workflows use different information; configuring one does not mean the other is ready. ### 5. Define recommendation rules Category, price range and permitted customer information can shape recommendations. An alternative should be relevant and available. When no suitable match exists, an explanation or agent handover is better than an unrelated suggestion. ### 6. Test the channels Use products that are available, sold out and offered in several variants. Ask the same question across channels and check that the same source data is used. Also test what happens when a record cannot be found or the source system does not respond. ### 7. Monitor results Track answer accuracy, response time and actions following recommendations. Revisit rules that perform poorly or product matches that repeatedly cause confusion. Assess the outcome using the store's own data. ## How do you know the setup works? - Stock replies agree with the source, allowing for the defined refresh interval. - Variants such as size and color match correctly. - Suggested alternatives fit the customer's need and configured rules. - Channels use the same information source. - Missing data and connection errors lead to an explanation or handover, not a guess. ## Common mistakes Hiding data-freshness limits, leaving category labels inconsistent, making recommendation rules too narrow and skipping tests are common problems. Another is treating order tracking and product availability as one operation. Validate each with the information and permissions it needs. ## Where does human support fit? Standard stock and product questions suit automation, while unusual needs and uncertain matches may require an agent. Palmate's shared Manuka example reports a 40% improvement in support efficiency. That customer result does not mean every inventory project will achieve the same improvement. Explore the [Palmate e-commerce chatbot](/ecommerce-chatbot) approach and review [usage plans](/pricing). [Book a demo](/contact) to assess a workflow using your catalog. ## Frequently Asked Questions ### Which store platforms can it connect to? Connections such as Shopify, iKAS and T-Soft can be assessed. Verify the required inventory fields and access conditions before setup. ### What does it recommend when an item sells out? It assesses alternatives that match category, price and other configured rules. If nothing suitable exists, it should explain that instead of suggesting an unrelated product. ### What should we check if stock information is wrong? Check source stock, variant matching, synchronization time and connection errors. Make freshness limits clear when updates are periodic. ### How long does setup take? It depends on the store connection, catalog quality, channels and recommendation rules. Do not assume it is ready before testing is complete. ### Which messaging channels can we use? Depending on configured connections, websites, WhatsApp and Instagram can be used. Test consistent information across channels. ### How is it priced? Usage, connections and service scope are considered. Review current plans together with any additional integration or service charges.