A voice shopping assistant for e-commerce

Product Q&A: real answers from JSON-LD data

Most e-commerce sites already embed schema.org/Product markup (JSON-LD) on product pages: name, brand, price, currency, stock status and category. When RAQS crawls your site it parses this structured data and stores it as a separate fact layer alongside the page text. So when a visitor asks 'is this kettle in stock?' or 'how much is that cast-iron pot?', the assistant uses the price and availability that are actually on the page — it does not guess.

An example: picture a store selling home and kitchenware in Turkey — say a site called 'Ev & Sofra' with thousands of SKUs across steel cookware sets, cast-iron pans, kettles, coffee makers and serving sets. If a visitor asks 'does your 24 cm granite pan work on an induction hob?', the assistant combines the compatibility detail from that product's description text with the price and 'in stock' fact from the JSON-LD into a single clear answer. When a price changes, re-crawling the site updates the answer too; there is no separate 'assistant catalog' to maintain by hand.

If the data is not on the site the assistant does not invent it: it says 'that information isn't on the page' or guides the visitor to the relevant category. In e-commerce this is a critical trust factor — a wrong price, or claiming stock that doesn't exist, means returns and complaints.

'Show me this' and 'is X available' — seeing the page and guiding

Product Q&A is only half the conversation. RAQS also 'sees' the live page and can move the visitor. When a customer says 'show me your induction-safe pans' the assistant guides them to the relevant category or filter page; when they say 'open that red cast-iron pot' it takes them to exactly that product page. This is far more natural than a classic search box: the visitor doesn't need to know the exact product name, they describe it in their own words.

In the Ev & Sofra scenario a typical flow looks like this: a visitor taps the orb on the homepage and says 'I'm looking for a lightweight, scratch-resistant pan set for my mum, mid-range budget.' The assistant finds suitable products in your catalog via RAG, summarizes two or three options aloud ('there's a 3-piece granite-coated set and a 2-piece ceramic set'), and when they say 'show me the granite one' it opens that page. Here the voice shopping assistant behaves like a sales advisor — but available 24/7 and to thousands of visitors at once.

Add-to-cart and the confirmation shield

RAQS doesn't just guide — with agent actions it gets work done on the page: when a visitor says 'add this to my cart' the assistant triggers the add action. Sensitive operations — payment, account changes, deletion — always require explicit user confirmation; a smart safety layer enforces this step. So the assistant never completes a purchase on the visitor's behalf; it brings them to the cart and leaves the decision to the user.

In practice this reduces shopping friction: instead of mousing around to find an item, pick a variant and add it to the cart, the customer progresses by speaking. In Ev & Sofra a sentence like 'add that coffee maker and two mugs to my cart' collapses several clicks into a single conversation turn — a big difference especially on mobile, where the user is tapping on a small screen.

Impact on conversion: don't let the lost visitor leave

Much of the abandonment and exit in e-commerce comes down to one thing: the visitor can't find what they want, or can't get a quick answer to a question. When 'is this right for me, is it in stock, when does it ship?' goes unanswered, the user closes the tab. A voice shopping assistant fills that gap at the exact moment of decision — without leaving the page, without hunting for a chat window, the user gets the right answer spoken back.

RAQS targets the three blockers that change a purchase decision: discovery (finding the right product), trust (removing uncertainty about compatibility/price/stock) and friction (shortening the add-to-cart step). Each of these is a measurable drop-off point in the funnel. Because the assistant answers verbally, instantly and from your site's real data, it is far more effective than a classic FAQ page; the user asks in their own words rather than having to search.

Reducing support load

A large share of customer-service messages are repetitive, simple questions: 'do you have size X', 'when does this arrive', 'how many pieces in this set', 'how do I return it'. The answers are usually already on the site — on the product page, in the shipping or returns policy — but the customer prefers to message rather than search. Because RAQS takes all of your site content (product pages + policy/FAQ pages) into its knowledge base, it answers the bulk of these questions on the page, instantly, without ever reaching a human agent.

You manage the knowledge base from the dashboard: crawl your site (sitemap + pages; a headless browser is used for JS-rendered product listings), paste your returns/shipping text or upload your PDF catalog. Content is chunked, embedded with Azure and retrieved with hybrid search (semantic + keyword + RRF fusion). Your support team then focuses on the genuinely human, complex cases rather than the repetitive ones.

Setup and data security

You add it to your store with a single <script> line — Shopify, WooCommerce, Magento or a custom stack, it doesn't matter; RAQS runs in front of the site, in the browser. From the dashboard you add your site, crawl the knowledge base, set allowed domains, and choose the mascot/persona and the assistant's name. Instead of the classic orb, a 3D mascot that roams the page (the flagship deer, 'Spark', or 30+ characters) becomes a face for your brand.

On the data side, the multi-tenant architecture isolates each store's data with row-level security; the crawler is SSRF-guarded; an origin allowlist and Cloudflare Turnstile block abuse. Infrastructure is hosted in Azure's EU region and designed with KVKK/GDPR awareness. Pricing is pay-as-you-go usage plus a base plan; you start with a free trial.

<script async src="https://raqs.ai/v1/raqs.js"
  data-raqs="YOUR_SITE_KEY"></script>