How e-commerce stores are using Khub to power AI product support

E-commerce businesses have a unique AI challenge: they often have hundreds or thousands of products, each with specific details, sizes, materials, shipping times, return policies, care instructions. When an AI customer support tool doesn't have access to accurate product data, the results range from unhelpful to actively harmful.

We spoke with three Khub users in the e-commerce space who solved this problem in different ways, all using the same platform.

Luma Home, handmade candles (42 SKUs)

Luma Home sells artisan candles online and through their Claiq-powered website. Before Khub, their AI chatbot frequently gave wrong burn times, confused fragrance families, and couldn't answer questions about ingredients, critical for customers with allergies.

"A customer asked if our Lavender Fields candle contained soy, and the chatbot said yes. It's 100% coconut wax. That was the moment we knew we needed a better knowledge source."

Amy L., Founder

Luma created a Khub knowledge item for each of their 42 candles. Each item includes the name, description, wax type, fragrance notes, burn time, price, and care instructions. They also added 15 FAQs covering shipping, returns, and wholesale enquiries.

Result: zero incorrect product responses in three months of operation. Their chatbot now handles 60% of customer queries without human intervention.

ThreadCraft, custom apparel (200+ SKUs)

ThreadCraft sells custom-printed t-shirts, hoodies, and accessories. With over 200 SKUs and frequent seasonal drops, keeping AI tools up to date was a full-time job for their operations manager.

Their approach: they export their product catalogue as a CSV, attach it to a Khub knowledge item, and let Khub's file indexing extract the text. Each product also has a dedicated FAQ item for common questions ('Does this shrink in the wash?', 'What's the weight of the hoodie fabric?').

• Product updates: change the CSV, re-upload, and Khub re-indexes automatically.
• Seasonal drops: add new items to the CSV and attach seasonal FAQs.
• Discontinued items: archive the knowledge items and they're removed from search instantly.

"We used to spend 2 hours per product drop updating all our AI tools. Now we update the CSV, re-upload it to Khub, and everything stays in sync."

Marcus T., Operations Manager

GreenLeaf Supplements, regulated products (85 SKUs)

GreenLeaf sells health supplements online. Their biggest concern was compliance, their AI could not make health claims, recommend dosages, or provide medical advice. The risk of an AI chatbot making an unsupported health claim was significant.

They used Khub's Behaviour Rules section to set clear constraints on what their AI can and cannot say. Combined with precise product items (listing only the approved claims from their label), they created a knowledge base that's both informative and compliant.

• Behaviour Rule: 'Never recommend a specific dosage. Always direct the customer to the product label or a healthcare professional.'
• Behaviour Rule: 'Do not compare our products to competitors or make claims about curing, treating, or preventing any condition.'
• Product items include only the claims that appear on the approved label, nothing more.

Result: zero compliance incidents since launch. Their AI accurately describes products, answers ingredient questions, and routes medical queries to their support team.

Common patterns

Across all three businesses, the same principles emerged:

• One item per product. Don't bundle everything into a single 'catalogue' item.
• Supplement with FAQs. For every product, add 2 to 3 FAQs covering the most common questions.
• Use tags consistently. All three businesses tag items by category, season, and availability status.
• Set behaviour rules. Tell the AI what not to say, not just what to say.
• Review monthly. Check for discontinued products, outdated prices, and new questions that aren't covered yet.

If you have more than 50 products, consider using the bulk import approach: export your catalogue as a CSV, attach it to a knowledge item, and let Khub's file indexing handle the extraction. Then create dedicated items for your top 10 to 20 products that get the most customer questions.