Jewelia: An AI chatbot for ecommerce, built for Earth’s Treasury
Buying a sapphire or an engagement ring starts with a color, a style or a budget, not a product name. Zen Agency built Jewelia, an AI jewelry concierge on the Earth’s Treasury store that guides shoppers to the right pieces, answers from the store’s own content and brings in a person when it matters. The MVP went live in 4 months, and we’re still adding to it.

Estimated at 6 months. MVP live in 4.
Earth’s Treasury sells Montana and other sapphires, gemstones, engagement rings, wedding bands and custom jewelry. Each shopper starts somewhere different. One knows the color they want but not the words for shape or clarity. Another cares about sparkle and setting height. A third needs a band that works with a ring they already own, and some need a designer, not a product list.
Jewelia adds an AI concierge to the existing WooCommerce store instead of replacing it. Structured choices handle the things a catalog can filter; conversation handles everything else; and the store stays where people buy.
The MVP went live in 4 months against a 6 month estimate. Since then we’ve kept working with Earth’s Treasury to extend what Jewelia can do.
ClientEarth’s TreasuryIndustryGemstone and jewelry e-commerceServicesAI chatbot development, conversation design, catalog search, WooCommerce integration, admin toolsBuilt withReact, TypeScript, Express, PostgreSQL, OpenAI, Algolia, WooCommerce, LiveChatTimelineMVP live in 4 months, against a 6 month estimate; development ongoing
Time to MVP4 monthsZen Agency MVP, live6 monthsTraditional estimateLive in a third less time, and still growing.
Honest about being AI, from the first message
Jewelia says what it is before the chat starts. Shoppers see how the conversation is recorded and kept, agree in a checkbox, and can contact the team instead if they prefer. Then they pick a path: A loose sapphire, a gemstone engagement ring, wedding bands, or a necklace, bracelet or earrings.
Consent recorded separately from the chat, with withdrawalPrivacy policy and a non-chat contact route on the first screenA set retention period for chat history“Chat with a Human” one tap away on every screen


Find a sapphire by color, shape and carat
The sapphire path asks what matters most, then narrows by color, shape, carat and clarity. Every choice shows how many stones are available before the shopper continues, so they never end up at an empty result.
Multi-select color and shape tiles with live countsCarat range sliderGo back a step and change any earlier choiceFinish on a filtered store page or a single stone
Engagement rings chosen by picture
Most people can’t name a ring style, but they know it when they see it. Jewelia shows the store’s own product photos for each style, then refines by sparkle, band, setting height and budget.
Halo, solitaire, three-stone, cluster, vintage, nature and bezel stylesSparkle level and band treatmentBudget rangesBespoke requests routed to custom-design intake, not a stock result


Wedding bands that ask the right follow-up
Bands come in women’s, men’s and unisex, and couples often want both. Jewelia asks which, then whether the engagement ring came from Earth’s Treasury, so it can suggest bands that pair with it or start a custom band.
Women’s, men’s and unisex, or both in one conversationPairing suggestions for a ring bought from the storeGemstone or plain band preferenceCustom band intake when nothing in stock fits
The conversation follows the shopping journey
Jewelia is embedded in the WordPress store, so it knows where the shopper is and what they’ve done. Nobody has to repeat themselves.
1. Browse
Product views are shared with the chat, including the item, image and selected options.
2. Build
The ring builder passes its progress, so advice fits the setting being built.
3. Add to cart
A cart addition prompts a relevant next step, like choosing a setting for a loose stone.
4. Get help
A handoff to LiveChat carries the conversation, inside business hours.
Shopper contextPageProductRing builderCart
Answers grounded in the store’s own knowledge
Free-form questions are answered from Earth’s Treasury’s WordPress content using retrieval with OpenAI embeddings and PostgreSQL vector search, with links back to the source page. Questions it can’t answer go to a review queue, so the team can fill the gap.
Source links on answers, with visitor clicks trackedUnanswered questions queued for human reviewKnowledge refreshed when store content changesBusiness-hours-aware handoff and contact options
An admin to run the concierge
Behind the chat is a protected workspace where the team reviews conversations and keeps Jewelia accurate.
Conversation search, filters, transcripts, tags, notes and CSV exportShopper journey view: Products viewed and added to cartConsent records and lead notificationsCatalog health, knowledge refresh, business hours and appearance settings
The outcome
Jewelia gives Earth’s Treasury shoppers a guided way into a deep catalog: By color and shape for loose stones, by picture for rings and with the right follow-up for bands. It works inside the existing WooCommerce store, answers from the store’s own content and hands off to the team when a person is the better answer. The MVP went live in 4 months against a 6 month estimate, and new functionality keeps shipping.
For Zen Agency it shows how we build AI chatbots for e-commerce: Connected to the catalog, honest about being AI and designed around the way people actually shop.
Tell us the constraint. We’ll tell you whether it can be built, and what it costs.
Scope, price and schedule inside five business days.











