🛍️ Shoppr - Agentic Marketplace

📋 TL;DR

Shoppr is a personal concept for a marketplace built around AI agents from the ground up. In a 2-week exploration, I designed 22 screens covering every core flow, from onboarding to checkout: natural-language and image search, transparent reasoning on every pick, and an agent that finds and applies coupons for you. The outcome is a purposeful shopping experience that respects the user's intent instead of exploiting impulse.

🎯 The problem

Online marketplaces have barely changed in decades. You type a keyword, scroll through hundreds of results, compare tabs, hunt for coupons somewhere else, and either buy something you didn't need or give up. Most "AI" in shopping today is a chatbot layered on top of that same default flow.

Shoppr asks what a marketplace looks like when agentic intelligence is the core of the experience, and how to make that intelligence feel trustworthy rather than just another way of selling things you didn't need to begin with.

👨‍💻 Who is it for?

Mindful shoppers, typified by our persona Alex (29), who usually knows exactly what he wants before opening the app, is trying to cut unnecessary spending, and is tired of opening five tabs and googling coupons just to buy a pair of running shoes.

🔬 What competitors taught us:

I studied eBay, Amazon, Shopee and others, from the maximalist, stimulus-heavy environment of Shopee to the more utilitarian flows of eBay. Three gaps showed up across the spectrum:

• Built for impulse. Heavy visuals, endless scrolling and promotional noise keep users being sold to rather than helped.

• Filters don't get us. People think about what they want in plain words, but have to translate it into keywords, categories and filters.

• Deals and logic live elsewhere. Coupons are hunted mostly off-platform, and system recommendations feel like ads rather than a tailored experience.

🔍 Our defined edges:

1. Agentic search in plain language

The Problem: Users know what they need but have to express it through keywords and filters.

The Decision: Let users describe what they want in their own words. The agent interprets the request, curates the best results and goes beyond, for example by suggesting comparable brands the user hadn't considered.

The Trade-off: An agent-first experience could alienate people who just want to browse. So the familiar foundation stays with category browsing, best prices and best-sellers sitting alongside the agent, which adds an extra mile rather than replacing the basics.

"I want to just describe what I'm looking for in my own words and have the app find the best options for me, without needing to set up a dozen filters."

2. Transparency as a feature

The Problem: Agentic tools live and die by user trust, and black-box recommendations feel like ads.

The Decision: Every agent result carries a "Why this?" affordance that shows the reasoning behind the pick.

The Trade-off: Explaining every result up front would bury the products in text. So the reasoning is on demand: one tap away, sitting inside a convenient dropdown, never in the way.

"If an agent is picking products for me, I want to understand why it picked them. Just show me the reasoning."

3. Coupons, handled

The Problem: The coupon hunt is one of the most universally painful parts of shopping online.

The Decision: The agent proactively finds and applies available deals, both while searching and at checkout.

The Trade-off: An agent changing your cart on its own can feel like a loss of control. So every applied deal is made visible, with a clear "Your coupon has been applied!" signifier, and a discount line in the order summary. The best price available becomes yours by default, not through the usual hassle.

"I would love for the app to just tell me when there's a coupon available, rather than making me go and find it myself."

4. Visual search

The Problem: Sometimes the product is easier to show than to describe, or the user doesn't know exactly a brand name, comparative term or such.

The Decision: Users can upload a picture, and the agent identifies the product and matches it to available listings, closing the loop between inspiration and purchase.

The Trade-off: Image matches won’t always be exact. When they miss, the agent invites the user to refine with text, and offers close alternatives based on the image, each with its reasoning shown.

"Sometimes I see something I like and I just want to find it online. I shouldn't have to describe it in words if I already have a picture of it."

🎨 Visual system

A calm, purposeful environment that keeps the focus on products and agent output, never on the interface. A warm off-white background with high-contrast black type stays approachable without being loud. A restrained green accent does double duty: it's the brand color and the language of value, shared by discounts, agent highlights and interactive elements. DM Sans handles body text, with Montserrat for impact headlines, and a consistent rounded radius across cards, buttons, inputs and chips keeps even complex screens friendly.

🖥️ Key screens

1. An onboarding that tailors the experience

Users shape how the agents work for them during account creation, and can adjust it later from their profile.

2. Agentic search that understands you

Plain-language and image search, with curated results and comparable brands the user hadn't considered. Sometimes you're looking for silver and end up finding gold!

3. A clean foundational experience

Familiar category browsing, best prices and best-sellers, with the agentic tools as an extra layer on top.

4. Coupons, handled

The agent finds and applies deals from search to checkout, ending in a clear payoff in the order summary.

💭 Outcome & reflections

Shoppr is a concept, so there are no launch metrics. Instead, this is how I'd measure whether it works:

• Agentic adoption rate: the share of sessions with at least one agent query, showing users not only choose, but trust the agent as their main way to shop.

• Query-to-purchase conversion: agent searches vs. manual browsing, testing the core bet that intent-driven shopping converts better.

• Coupon engagement: how often agent-found deals are applied, and their effect on cart abandonment.

• Intent-driven retention: return visits tied to specific needs, compared with competitor benchmarks.

If I could change something about how this project came to be, that would be getting the chance to expand on edge-cases. Since this was a quick exploration, meant to become a portfolio piece, it doesn't consider a few interesting edge cases, like how the search elements handle vague terms, some more traditional flows like payment method setup, etc.

Below, you'll find all the core screens, and if you liked reading through this project, or have ideas on where to take it next, let me know in our soon-to-be scheduled interview 😉