Inspiration
Fashion buyers face two fundamentally different questions: What is culturally relevant? and What is commercially viable?
A brand can have strong cultural connections while failing basic purchasing constraints. Conversely, a commercially viable product is not automatically culturally relevant.
I built VennShift to separate these questions and bring cultural intelligence into a more transparent, evidence-aware decision process.
What it does
VennShift is a cultural and commercial decision-support application for Fashion & Accessories.
It offers two distinct workflows:
Explore a Concept — Users define their market, audience, product direction and one to three reference brands. VennShift can request Qloo cultural discovery to investigate relevant directions, presenting validated evidence and its limitations.
Evaluate a Shortlist — Users compare two to ten brands using declared availability, minimum order quantities (MOQ), gross margins and lead times. VennShift first determines commercial eligibility, then attempts cultural comparison when at least two eligible candidates have comparable Qloo evidence.
The key difference: Cultural affinity is not presented as a sales forecast, market-demand estimate or probability of commercial success.
When evidence is insufficient, VennShift abstains rather than inventing a recommendation.
How I built it
VennShift uses Next.js, React, TypeScript and Zod, with a server-side Qloo API integration.
The application combines:
- Deterministic commercial eligibility checks.
- Bounded Qloo discovery and comparison workflows.
- Exact brand-entity resolution and typed provider-response validation.
- Source-derived explanations and explicit evidence limitations.
- Server-side request admission, execution journals and safeguards against invalid or repeated requests.
- Responsive Explore and Evaluate interfaces with Light/Dark themes.
The workflow is agentic through constrained orchestration: it selects permitted analysis steps, resolves entities, validates evidence and determines whether a supported conclusion is possible.
The implementation includes automated regression tests, TypeScript validation and production-build checks.
Challenges I ran into
The biggest challenge was maintaining a clear boundary between cultural affinity and commercial evidence.
Qloo cultural signals do not independently verify supplier availability, margins, audience demographics or local market demand. I designed the system to preserve those distinctions rather than create misleading conclusions.
Other challenges included resolving exact brand entities, handling incomplete provider responses, protecting API credentials, controlling request execution and ensuring that the interface accurately represents the available evidence.
Accomplishments that I'm proud of
I built two distinct but connected decision workflows instead of a generic AI recommendation chatbot.
VennShift can assess commercial constraints without making external cultural-intelligence requests, preserve user-declared context, and distinguish an evidence-supported comparison from a case where it must abstain.
I also implemented an explainability-oriented decision interface and automated tests covering commercial eligibility, evidence validation and failure conditions.
What I learned
A useful AI agent is not defined by how confidently it produces an answer.
It is defined by whether its actions are controlled, its evidence is traceable and its conclusions respect the limits of the available data.
The most valuable outcome of this project is an approach to combining cultural intelligence with deterministic business rules without confusing correlation, eligibility and proof.
What's next
The next priorities are hosted end-to-end validation, additional accessibility and cross-browser testing, user research with fashion buyers, and evaluation against real procurement scenarios.
Future work will focus on expanding verified Qloo context mappings and improving decision explanations while retaining the same evidence-first safeguards.
Built With
- next.js
- node.js
- qloo-api
- react
- typescript
- zod