Agentic AI Shopping Assistant: Conversational Commerce

Overview

A Fortune 100 technology company’s online store serves approximately 120 million annual visitors in the US alone, targeting consumers and businesses across thousands of products and configurations. It set out to replace browse-and-search navigation with a personalized, conversational shopping experience that helps every shopper find the right products, services, and information, aimed at lifting engagement and conversion.

Business Complexity

Traditional navigation and static pages made personalized product discovery hard to deliver at scale, leaving shoppers to sift through large catalogs on their own. The client wanted an AI-native experience that understands shopper intent through text and voice, remembers preferences across visits, and adapts as shopper needs evolve across interactions.

Action

  • Built an agentic AI assistant that interprets shopper intent through text and voice conversations, orchestrating specialized agents that understand needs, compose pages, and guide shoppers.
  • Grounded every recommendation in the live product catalog, pricing and stock, giving shoppers guidance that accurately reflects availability, specifications, and options rather than generic or outdated information.
  • Generated each shopping page live from the conversation, assembling recommendations, comparisons, and direct answers around the shopper’s stated need, replacing static templates and click-through navigation.
  • Introduced opt-in memory across visits, letting the assistant build on past interactions, and refine future recommendations and permissions managed through the site’s existing consent platform.
  • Delivered concept to limited production beta in 60 days, clearing the enterprise’s architecture, security, privacy, and release approvals while aligning marketing, legal, and platform teams.

Tech Stack

  • HCL Commerce
  • Adobe analytics
  • Dynamic Yield
  • One Trust
  • LLMs: GPT Realtime 2.1 (voice) and GPT-5.6 Luna (text)

Impact

  • Concept to production beta in 60 days, through full architecture, security, and privacy validations
  • Consent-based personalization live across shopper visits
  • One voice-and-text experience serving both business and consumer shoppers
  • Publicly launched by the client as a new way to experience its brand online, with global expansion planned