Our read
Treat iAdvize as a sales-first shopping assistant that also reaches into support through ticketing and order-tracking integrations. Start with a product-page pilot where shoppers hesitate over choice, such as Caudalie, Mon Lit Cabane or Payne Glasses. Before you set a target, ask which method sits behind any conversion figure, because the published ones use different bases.
Our read: a sales assistant with support hooks
iAdvize describes itself as an AI shopping assistant for e-commerce sites, and its homepage frames it as shopping-focused rather than a traditional support desk. It uses engagement widgets placed at moments of hesitation and a shopping panel for product discovery, quick answers and cross-sell, styled to match the site.
Support is still in reach. Through its integrations, a conversation can open a ticket, answer an order-tracking question or update a customer record. So the choice is about what you want to measure first: revenue from guided shoppers, or deflected tickets. Start from the first and treat the second as a secondary effect.
Question before purchase: three product-page deployments
Caudalie, a French cosmetics brand selling in 23 e-commerce markets, wanted to reproduce the advice of a boutique adviser, who builds a full routine instead of selling one product. It launched the assistant in France and the United States in late 2025, then extended it to Italy, Spain and Turkey, with conversation starters on product pages and a tone adjusted to the brand.
Mon Lit Cabane sells customisable children's furniture. Choices of bed type, wood finish, safety options and storage overwhelmed shoppers, and support was thin at peak hours, so the assistant sits on product pages to narrow choices. Payne Glasses had a small team available in business hours and shoppers who struggled with filters and product discovery when picking frames and lenses; the assistant runs on product and listing pages around the clock.
The common pattern is a complex catalogue where the shopper needs a conversation, not a search box. If your products are simple and well filtered, the case studies give you less to learn from.
Catalogue context and the connected stack
The integrations page says Shopify and PrestaShop have a native one-click install with automatic catalogue synchronisation and knowledge-base ingestion. WooCommerce and Salesforce Commerce use a universal tag instead. Product feeds can come through Lengow, Channable, SmartFeeds or Shoppingfeed, and Smart Tribune feeds FAQ content.
Order questions rely on Shipup, Shopify, PrestaShop, OneStock or Welcome Track for live shipment status. Consent platforms such as Didomi, OneTrust, Axeptio and Trust Commander are listed so visitor preferences stay in sync, and Klaviyo, Attentive and Postscript capture contacts for email or SMS. For data handling, the vendor's trust centre lists ISO 27001 and says a data processing agreement is available; neither is proof of your own compliance.
Handing over to a person
Caudalie's case describes escalation to human beauty advisers when the assistant should step back, and it reports that 81.8% of conversations were handled by the AI. That figure tells you how often a conversation was handled by the assistant, not whether the answers were good.
On the service side, Zendesk, Gorgias, Salesforce, Microsoft Dynamics, Helpscout and Kustomer integrations transfer the conversation into a ticket with its history. Payne Glasses reports a 50% reduction in human support requests. Kendra Scott's case page text does not detail a handoff, so check that path yourself.
Assisted, influenced, incremental: why the numbers differ
Several published methods sit side by side. Caudalie compares the first half of 2026 with the first half of 2025 and reports 10x ROI on the French market and a 9.7% conversion rate. Kendra Scott reports 6% of total sales influenced by the assistant, a 6x conversion lift for assisted shoppers and 160% year-over-year growth in influenced purchases. The homepage shows a Samsonite figure of 2.2x incremental ROI, labelled as proven by A/B test, and separately a standalone +15% incremental revenue line with no scope or method given beside it.
A before-and-after comparison mixes in season and traffic. An influenced sale counts any purchase where the assistant took part, including shoppers who would have bought anyway. Only a controlled A/B test approaches incremental gain, and the +15% line cannot be placed on that scale without its method. Mon Lit Cabane (8% conversion among assistant users, over three months) and Payne Glasses (8x lift among assisted shoppers) are also vendor-published, and the case pages do not explain how the comparison group was built. These are vendor and customer statements, not independent tests, and they are not comparable with each other.
Editorial recommendation: run a pilot on a few product categories, with a share of visitors who never see the widget, over a period that spans your usual traffic swings. Then ask the vendor which of these definitions its reporting uses, and agree on that before you set a target.
Method and limits
2026-10-18
This dated snapshot counts distinct companies in public customer evidence. CONFIRMED means a case study, named testimonial or press source; DECLARED means the company itself told usedby; OBSERVED means a vendor listing; SIGNAL means a job post. All four count, but their strength differs. Removed and unlevelled lines are excluded. Examples are a capped selection verified within 90 days. Vendor facts are sourced records checked within 60 days. Coverage is uneven and missing evidence is unknown. These figures measure neither market share nor product performance; capture dates are not adoption dates.
