Shoppers get an answer on the product page in under a second, grounded in documents the brand controls.
Solution Studio Product Insights Agent
Solution Studio
Solution Studio Shopping Greeter
A stated goal becomes a finished look, so a shopper who cannot phrase a query still gets a usable answer.
A shopper states a goal, and a short guided conversation ends in a curated collage rather than a results page.
What the agent does
Turns a stated goal into a structured intent, chooses its next question from what is still unknown, then composes coherent sets from what it finds.
- Contributed by
- MACH Alliance (Reference implementation)
- Industry
- Retail, Apparel & Footwear
- Audience
- B2C
01
The problem
Why this exists
A search box answers the question it was given. A shopper who is furnishing a room, dressing for an occasion, or replacing one piece of a set is not asking a question yet, and the interface offers no way to arrive at one.
Read more
The usual response is more filters, which moves the work rather than removing it. The shopper still has to translate a goal into the retailer's vocabulary before anything useful comes back, and the goal is the part they can actually articulate.
02
Outcomes
What this demonstrates
- A shopper describes a goal in plain language and receives a curated set of products and matched content, rather than a ranked list to sort through.
- The conversation is short because each question is chosen from what the goal still needs. A fashion goal is asked about fit before style, a home project skips questions that do not apply, and nothing is asked that an earlier answer already settled.
- A thin category does not silently degrade the result. When a search comes back empty the constraints relax in a fixed order, and anything the shopper stated explicitly is preserved to the end.
Read more
- Something the shopper already owns is dropped from search rather than filled with a lookalike, and still reads as part of the finished look.
03
How it works
Architecture
The widget is a chromeless iframe embedded in a mockup retailer storefront, so it owns its own chrome and does not inherit the host page's styling. The retailer pages are deliberately not MACH Alliance branded; the operator surfaces are.
Read more
Intent resolution runs as a guided conversation that builds a structured model rather than a query string. Each turn selects the next question from what the model still lacks. When the model is complete the categories are queried against translated tags, with a retry ladder that drops the budget filter, then the style tags, then the hardest constraints, relaxing fully only on genuine emptiness.
Curation is a separate step. Once the products are final, a model composes two coherent sets from the results, and a third when the products support a different combination rather than a swapped item.
Search and CMS sit behind a single adapter interface. The same code runs against local fixtures or live services, switched by one flag, so the prototype behaves identically either way. Operator routes are off when the environment is production, and the two endpoints that spend money are rate limited.
Open standards used
REST APIs
Tech stack
HonoReactViteVercel AI SDKAlgoliaContentstack
Everyone credited on this solution
Next step
Take the code
This one is published in full. Read it, run it, fork it. Nothing here is behind a conversation.
Request an intro
This one is the MACH Alliance’s own work. Ask about turning stated goals into curated results, and the people who built it will answer.
Not ready to talk to anyone? MACH Insider is the Alliance mailing list. New solutions, program openings, and the research behind them, once a month.
Related solutions
Gift shoppers who cannot phrase a keyword search get a curated recommendation out of a conversation instead.
Kate Spade AI Gift Concierge
A shopping assistant that lifted conversion 25%, taken from a split test to all traffic.
Ask Frasers Shopping Assistant
Verification
Public code
The implementation is published in full, so anything on this page can be checked against the repository. Anything the repository does not show is not established here.
Where this came from
Written from the public repository README and the accompanying implementation article.
Status: Reference prototypeAdded Sep 10, 2026A listing is an editorial decision, not a MACH Alliance certification.
Built something worth showing?
Put your agentic solution in front of the enterprise leaders who use the hub to decide what to build next. Submissions are open to MACH Alliance Members, Collaborators, or Ambassadors.