Apparel
Fit, size, style, occasion, material, color, and availability may drive the choice.
Ecommerce & Online Retail/ Product discovery
Turn shopper intent into a short list that makes sense
Shoppers often know the outcome they want before they know the product name, category, model, or specification.
What the shopper says is rarely what the catalog stores.
THE SHOPPING LANGUAGE BRIDGEFROM INTENT TO SEARCH CRITERIA
WHAT THE SHOPPER SAYS
“I need a chair for long workdays.”
I need something waterproof for commuting.
Which one works with my camera?
I want a gift under $100.
Which of these is better for a small apartment?
WHAT THE CATALOG STORES
01dimensions02compatibility03materials04variants05weight06price07use case08technical specifications09inventory
DATA SEARCH
The job of the AI agent is to connect those two worlds. Data Search provides the product attributes. The conversation turns the shopper’s language into the criteria needed to search them.
This use case helps translate that intent into relevant products, useful comparisons, and a clearer buying decision without making the customer work through the catalog alone.
Guided shopping should change with the product.
Different categories need different decision logic.
A useful AI agent should follow the buying logic of the category instead of forcing every shopper through the same questions.
Electronics
Compatibility, specifications, existing devices, performance requirements, and price often matter more.
Furniture
Dimensions, room size, material, delivery constraints, assembly, and intended use can be decisive.
Beauty
The shopper may describe an outcome, routine, texture preference, ingredient requirement, or product type.
Parts and accessories
Compatibility may need to be confirmed before any recommendation is shown.
Ask only the question that changes the result
The conversation should not become a product quiz.
If the shopper has already provided enough information, search. If one missing detail changes the recommendation, ask for it.
EXAMPLE CONVERSATIONONE USEFUL FOLLOW-UP
At that point, another five questions would create friction rather than improve the result.
When several structured preferences genuinely matter, the AI Form node can collect them together.Intent · Constraints · Evidence
A useful recommendation needs three things.
Intent
What is the customer trying to accomplish?
This may be a task, occasion, problem, style, or desired outcome.Constraints
What limits the choice?
Budget, size, compatibility, color, availability, dimensions, materials, delivery timing, or other requirements can remove products that technically match but are not realistic choices.Evidence
Why does a product fit?
The answer should point back to actual product information rather than simply saying something is “best.”AI Answer can use the retrieved catalog data to explain that evidence in customer-friendly language.
Show the decision, not just the products
A shortlist is more useful when each option has a role.
This lets the shopper understand the tradeoff immediately.
SHOPPER CONTEXT
Business-focused backpack. 15-inch laptop. Daily train commute. Under $150.
OPTION 01
Best for compact commuting
Smaller profile and dedicated 15-inch laptop compartment.
OPTION 02
Best for organization
More internal storage and accessory pockets, but slightly larger.
OPTION 03
Best for wet-weather use
Water-resistant exterior with the strongest weather protection of the three.
AI ANSWER WITH PRODUCT CARDS ENABLED
With AI Answer with Product Cards enabled, those products can be presented directly in the conversation. For categories where appearance matters, AI Answer with Image Answers enabled or AI Answer with Albums enabled can support visual comparison.
Recommendation confidence should drop when the data is weak
The agent should not force a recommendation when a required fact is missing.
In those cases, the better response may be another question, a narrower statement of what is known, or Human Handoff when specialist judgment is needed. Trust is more valuable than always producing an answer.
FACTS THAT CAN CHANGE THE RECOMMENDATION
01unknown compatibility02unavailable dimensions03uncertain stock04missing variant information05unclear technical requirements
01Ask another question
02State what is known
03Human Handoff
Inventory can change the shortlist
A product that is ideal but unavailable is not a useful recommendation.
When pricing, stock, or variant availability is connected through product data or APIs, the AI agent can account for what the customer can actually buy now.
SHORTLIST CONDITIONSCONNECTED PRODUCT DATA
This matters especially for:
01size-dependent products
02limited inventory
03fast-moving electronics
04seasonal collections
05color or finish variants
The shortlist should remain grounded in current sellable options.
AskHandle building blocks
Components that power guided shopping.
Data Search
Provides the product attributes and searches the criteria against the store’s available data.
AI Answer
Uses retrieved catalog data to explain why a product fits, using customer-friendly language.
AI Answer with Product Cards enabled
Presents the shortlist directly in the conversation.
AI Answer with Image Answers enabled or AI Answer with Albums enabled
Supports visual comparison in categories where appearance matters.
AI Form
Collects several structured preferences together when they genuinely matter.
Human Handoff
Brings in specialist judgment when a required product fact is missing or uncertain.
Questions stores usually ask
Product discovery & guided shopping.
Can the AI agent recommend from our own catalog?
Yes. Data Search can retrieve product information from the store’s available data so recommendations are grounded in the actual catalog.
Can it compare products?
Yes. AI Answer can explain differences using the shopper’s stated priorities instead of only listing specifications.
Can it display products directly?
Yes. AI Answer with Product Cards enabled can surface products in the conversation. Image Answers and Albums can support more visual categories.
Can it collect several preferences at once?
Yes. The AI Form node can be used when multiple structured inputs materially improve the product search.
What happens when compatibility is uncertain?
The agent should avoid guessing. It can ask for the missing information or move the conversation to Human Handoff when specialist judgment is required.
Product discovery should end in the right next action.
Not every shopper needs the same ending.
The conversation may lead to:
- opening a product page
- comparing two finalists
- selecting a variant
- checking compatibility
- viewing another category
- speaking with a specialist
The useful outcome is not “the AI answered.” It is that the shopper has fewer, better choices and knows what to do next.