Best fit
ArcLight Rain Shell
$168
Lightweight waterproof shell with sealed seams and a hood that stows in its own pocket.
- $168, within your $200 budget
- Waterproof
- Also packs down small
Back to demo library
Retail and ecommerce on AWS
Turn your product catalog into a guided shopping conversation.
Offer shoppers a brand-aware digital associate that interprets intent, asks useful questions, recommends relevant products, explains the fit, and keeps your store policies in view.
What the experience includes
From discovery to a confident shortlist.
Natural-language product search
Clarifying questions
Rich product recommendations
Cross-session preferences
Store policy answers
Guardrails and evaluations
Understand shopper intent
Turn natural-language requests, budgets, preferences, and follow-up answers into a structured shopping goal.
Ground every response
Retrieve current catalog, inventory, pricing, product attributes, and approved store policies before answering.
Recommend and explain
Rank relevant products, explain why they fit, and present clear cards that help the shopper compare options.
Measure and improve
Trace conversations, score recommendation quality, review failures, and test changes before release.
Guided demo
Follow one shopper from request to shortlist.
Step through a shopping session on a sample store. Every choice changes the result, and the panel beside the conversation shows what the assistant did to produce it.
What this demo shows
- Who it is for
- Retail and ecommerce teams deciding whether a conversational shopping assistant fits their catalog and brand.
- What you will see
- Five steps of one shopping session: understanding a request, filtering the catalog, comparing products, answering policy questions, and handing off to a person.
- How it runs
- A guided simulation with a fictional store, Northstar Outfitters, and an 8-product sample catalog. It does not call an AI model. A pilot connects Amazon Bedrock to your own catalog and policies.
Northstar Outfitters shopping assistant
Guided simulation with a fictional sample catalog
Step 1 of 5: Describe the need
Pick what the shopper asks for.
I need a rain-ready commute setup under $200.
I searched 7 products and found 4 matches for waterproof and up to $200. Here are the top 3 and 1 more. The ArcLight Rain Shell fits best.
Accessories
Stormline Compact Umbrella
$39
Wind-tested compact umbrella that fits in a jacket pocket.
- $39, within your $200 budget
- Waterproof
- Also packs down small
Bags
Transit Roll-Top Pack 22L
$124
Waterproof roll-top commuter pack with a padded laptop sleeve.
- $124, within your $200 budget
- Waterproof
Business value
Bring your product expertise into guided shopping sessions.
Guided product discovery
Help shoppers move from an open-ended need to a useful shortlist through natural conversation.
More useful catalog data
Ground conversations in your product attributes, merchandising knowledge, inventory, and brand guidance.
Consistent brand experience
Apply the retailer's tone, approved claims, policies, and escalation rules across conversations.
Controlled path to purchase
Keep the assistant inside approved actions while preserving human ownership of sensitive exceptions.
AWS implementation foundation
Connect to the commerce stack you already operate.
The assistant can sit above existing product information, search, inventory, profile, order, and policy systems. The architecture separates conversation from commerce actions so teams can control what the assistant may read, recommend, or change.
Review an implementation pathAmazon Bedrock
Model access, orchestration, and response generation
Amazon OpenSearch Service
Product and policy retrieval across catalog content
AWS Lambda and APIs
Catalog, inventory, profile, and commerce actions
Amazon DynamoDB
Consent-aware shopper preferences and session memory
Bedrock Guardrails
Content controls and brand-policy boundaries
CloudWatch and evaluations
Conversation tracing, quality checks, and operating metrics
Frequently asked questions
What retail teams ask before a pilot.
What does an agentic shopping assistant do?
It asks clarifying questions, retrieves relevant catalog and policy information, recommends products, explains the fit, remembers approved preferences, and helps a shopper move toward a purchase decision through conversation.
Can the assistant match a retailer's brand voice?
Yes. Brand language, product claims, policies, response patterns, and escalation rules can be configured and evaluated before a version is released.
How does the assistant avoid inventing product or policy details?
Responses are grounded in approved catalog, inventory, pricing, and policy sources. Guardrails, citations or source traces, automated evaluations, and human review can be applied according to the risk of the action.
Can this connect to an existing ecommerce platform?
Yes. The implementation can connect to existing product information, inventory, customer profile, order, search, and commerce APIs instead of requiring a catalog replacement.
Start with one catalog and one shopping journey
Turn this demo into a retailer-branded pilot.
Connect a focused product set, configure the brand and policy layer, test with real shopping questions, and define the quality gates required for launch.