E-Commerce Lifestyle Agent
Creating a personal shopping assistant that embodies the brand's avant-garde aesthetic.

The Challenge
Product discovery relied on generic recommendation logic. Results were accurate but interchangeable, lacking the brand’s editorial point of view. Users experienced the AI as a utility—closer to a search engine than a curator.
This created a gap between brand identity and digital experience. The brand was known for its distinctive taste, yet the AI surfaced items without narrative, context, or conviction. Without a point of view, recommendations felt safe and forgettable.
The core problem was not data quality; it was the absence of designed judgment. The system optimized for relevance, not taste.
The Solution
We reframed the AI from a retrieval tool into an editorial system. The goal was to encode the brand’s taste—how it sees, selects, and explains—so recommendations feel intentional rather than computed.
The foundation was a highly opinionated persona trained on the brand’s editorial archives, lookbooks, and past campaigns. This persona governs not just what is recommended, but why—and how that reasoning is communicated.
To make this operational, we designed a cognitive framework that enables the AI to express judgment with clarity and restraint. Each recommendation includes a concise rationale grounded in design principles, not surface-level attributes.
- Core Identity: The "Avant-Garde Curator" — confident, inspiring, and deeply literate in design history. It leads with perspective, not apology.
- Opinionated Logic: Recommendations are justified using silhouette, proportion, texture contrast, materiality, and cultural references. The system deprioritizes shallow signals like simple color matching.
- Conversational Style: Short, punchy sentences. Precise fashion vocabulary. Helpful, but never overly eager. The tone communicates authority without friction.
We also introduced visual generation guidelines to ensure coherence between language and imagery. The AI pairs recommendations with visuals that reinforce the narrative—consistent lighting, composition, and styling cues aligned with the brand’s aesthetic.
Finally, the interaction model favors guided exploration over exhaustive lists. Users receive a small set of high-conviction options, each with a clear point of view, and can refine through conversational prompts.
The Result
Encoding taste transformed both engagement and commercial performance. The experience shifted from transactional browsing to editorial discovery.
- A 25% increase in average order value (AOV) driven by high-conviction, contextual recommendations.
- Users spent 3× longer interacting with the AI compared to the traditional search bar.
- The brand’s intangible "vibe" was translated into a consistent, scalable digital experience.
Qualitatively, users described the AI as having “taste” and “perspective.” Recommendations felt curated, not computed. The system didn’t just surface products—it framed them.
This case shows that in discovery-heavy contexts, differentiation comes from judgment. When AI can articulate why something is right—not just that it is relevant—it becomes a curator users return to.
"Tarek didn’t just refine our AI—he redefined how it connects with people. What we had before was efficient but transactional. What we have now is an experience that feels composed, intelligent, and reassuring, even in high-pressure situations....Read more"