AI Chatbot for Retail

Retail support splits into two kinds of business with overlapping questions. An online retailer gets product, shipping, order and returns questions. A multi-location retailer with a website gets those plus "are you open on Sunday", "which branch is nearest" and "do you have this in the Leeds store". A retail chatbot is worth having when it can answer most of both lists from data the retailer already has.

The mechanism behind "grounded" is retrieval-augmented generation: for each question the bot searches your synced catalogue and knowledge base first and writes only from what it finds, so it does not invent a branch, a price or a returns window. What is a RAG chatbot explains it; the customer-service chatbot ranking compares tools built for support desks.

• Store hours and locations. The bot answers "are you open on bank holiday Monday" and "which store is closest to SW1" from your location pages. Keep one page per store with address, hours, phone and services (click-and-collect, returns desk, fitting), and the bot has what it needs.
• Product availability. "Do you have the 32-inch in stock" is answered from the catalogue sync with a product card showing price, sale price during a sale, and a link. Stock appears on the card only if you turn it on and your platform maintains it.
• Order status. The shopper gives the order number plus the email or phone on the order and the bot reads the status live from the platform. Available through the plugins for WooCommerce, OpenCart, PrestaShop, Magento, Shopware, Joomla and Drupal Commerce; not on Shopify. How to track orders with a chatbot has the flow.
• Returns and exchanges. The bot quotes your policy as written: window, condition, who pays postage, refund timing, whether online orders can be returned in store. Exceptions go to a person.
• Sizing and fit. Size charts per category and fit notes ("this brand runs small") turn "will a medium fit me" into an answer instead of a return.
• Gift cards and loyalty. Balance checks are usually a link to your gift card page; terms, expiry, how points are earned and redeemed are answered directly from the FAQ.
• Coupons. When a shopper hesitates on price, the bot can mention a coupon you have configured, with its conditions, when it fits the conversation.

Everything above works in the shopper's language: 95+ languages are detected automatically, with a fallback you choose. How the bot chooses which products to show is covered in conversational product recommendations.

The catalogue arrives through the platform plugin (or a crawl and CSV upload on other sites). The rest is a one-afternoon job. Sources can be a website crawl, pasted URLs, PDF, DOC, DOCX, CSV, TXT or Markdown uploads up to 10 MB each, manual text, FAQ import from CSV or Excel, and AI-generated FAQ suggestions you approve.

• One page per store: address, opening hours including holidays, phone, parking, services offered.
• Shipping: zones, costs, delivery windows, cut-off times, click-and-collect timing.
• Returns and exchanges: window, condition, exclusions, refund method and timing, in-store return rules for online orders.
• Sizing: charts per category, measuring guide, brand-specific fit notes.
• Product care and warranty: washing, assembly, warranty length and claim process.
• Payment: methods, instalments, gift cards (purchase, redemption, expiry), price-match policy if you have one.
• Loyalty programme: how to join, how points accrue, how to redeem, tier rules.
• Accounts and orders: changing an address before dispatch, cancelling, applying a coupon.
• Accessibility and services: wheelchair access, personal shopping, alterations, repairs.
• Top 30 questions from last month's email inbox, as an FAQ CSV.

Knowledge item caps are 100 on Free, 500 on Starter and unlimited on Grow. Load policies first, store pages second, FAQs third; policies remove the most handovers per item. How to train a chatbot on your data covers file formats and common mistakes.

Two weeks is enough because there are no flows to build; the bot answers from content, so the work is content review. Details of each setting are in the docs.

Add-to-cart and order value for chat-assisted sessions are measured in your store analytics, not in the chatbot; compare sessions that opened the widget with sessions that did not. Reports on Vatdi (overview, leads, feedback, insights, usage forecast) are on every plan, including Free; see features.

Retail demand is not flat, and neither is the question volume. Three things to do before the peak rather than during it:

1. Move to a plan without a conversation cap. Free stops at 15 conversations a month and Starter at 150. Grow at $7.49/mo has no cap, and AI replies inside a conversation are never capped on any plan. Switch two weeks before the campaign.
2. Load the seasonal content. Extended returns windows, order-by dates for delivery before a holiday, gift-wrap options, gift card delivery times, and extended store hours. Remove them afterwards so January shoppers are not told December rules.
3. Cover the handover hours. Extend agent hours for the peak, and check the missed-handoff count daily. Outside agent hours the widget tells shoppers honestly that nobody is available and captures their details.

The Black Friday chatbot strategy is a step-by-step version of this with a timeline. The Baymard Institute's running average for cart abandonment is about 70%, and unanswered pre-purchase questions are one of the causes a chatbot can actually remove; it is not a cart-recovery tool and does not send follow-up emails.

• No POS integration. The bot does not connect to till systems. If a question needs the POS, it is a handover.
• In-store stock only if your e-commerce platform holds it. Per-branch availability can be answered only when the platform exposes it to the catalogue sync and you enable stock. Otherwise the bot answers "check with the store" and offers the branch phone number from your location page.
• Website widget and WhatsApp. No SMS, Messenger, Instagram or email channel; WhatsApp works through your own WhatsApp Business number. Retailers who need those alongside the website pair Vatdi with a messaging tool, or choose an omnichannel product such as Tidio.
• Order lookup is not available on Shopify. Shopify stores get the widget and answers from pages, collections and policies; live order status is on the roadmap.
• No attachments in the widget, no canned responses or internal notes for agents.
• No upsell engine or cart-recovery automation. The bot answers what is asked and can mention a configured coupon when it fits.
• One store per plan. A retailer with three separate websites needs three plans.

Retail chatbot solutions fall into three categories: a website widget grounded in your own data (Vatdi), an omnichannel inbox (Tidio), and a helpdesk built for a contact centre. If your retail operation is mostly a contact centre handling phone, email and social channels, a helpdesk such as Zendesk (Suite Team from $55 per agent per month on annual billing, AI agents as add-ons) or Intercom (from $29 per seat per month plus $0.99 per Fin AI resolution) is the better category, at a very different price. Prices are as of September 2026 and change; check the vendors' pricing pages. Vatdi vs Intercom sets out that trade-off.

Retail chat collects personal data: names, emails, order numbers, sometimes addresses. What Vatdi does with it, on every plan:

• HTTPS everywhere; store credentials encrypted at rest; per-store data isolation so one retailer's data is never retrieved for another.
• Chat answers are generated by OpenAI GPT-4o-mini via API, with zero-retention API mode where enabled; conversations are not used to train models. LLM call logs are kept for 90 days.
• Order lookup requires the order number plus the matching email or phone, so a shopper cannot read someone else's order.
• GDPR and CCPA rights: JSON data export, account and data deletion, and a DPA on request. Subprocessors are listed on the Trust page.
• Encrypted backups with a 10-day rolling retention.

Vatdi does not hold SOC 2 or ISO 27001 certification; if your procurement requires one, say so early. AI chatbot with GDPR compliance covers what to put in your privacy notice when you add a chat widget, and security lists the technical controls.

Every plan has every feature: order lookup, handover, leads, languages, reports, unlimited teammates. The only paid-only capability is removing the badge. No setup fee, no contract, cancel from the billing page, annual saves about 17%. A retailer with a catalogue under 2,000 products and steady traffic fits Starter; anyone with a seasonal peak or a larger catalogue should be on Grow. See pricing and how much an AI chatbot costs for the comparison with per-seat and per-resolution pricing.