Work05 · Multi-sided platformPlatform
Cargo
A two-sided marketplace run by a team of agents
A wholesale marketplace where independent makers sell to retail buyers. The hard part was never the storefront — it's the operations. We designed the two-sided product and the multi-agent system that runs the back office, so a five-person team can operate at marketplace scale.
Try the product
Try the buyer concierge
Act as a retail buyer — ask for products, build a sample order, set a reorder cadence — like the real marketplace concierge.
A working agent, running live — not a mockup
Cargo
Buyer concierge · live demo
Cargo — operations console
4 agents liveCatalog agent
1,284 listings enriched today
Pricing agent
7 anomalies flagged
Demand agent
23 reorders nudged
Concierge agent
16 buyer carts assisted
Needs a human decision
3 pendingPricing agent: TERRA mugs priced 40% below category — approve override?
Catalog agent: New maker 'Alder' — 3 listings missing photos. Publish anyway?
Demand agent: Maple St. Goods low on ALDER — send reorder nudge?
Agents do the work; you decide the edge cases.
Interactive prototype — click around, it actually works
Strategy deliverable
The journey we mapped before designing anything.
Every stage from first spark to loyal return — what the customer does, feels, and exactly where the experience leaks. The dip in the curve is the moment we built the work around.
Customer journey map · concept
Renée — stocking the shelves
Owns a gift boutique; a retail buyer who needs fresh stock her customers can't get on Amazon.
Customer journey map for Renée (stocking the shelves), across 6 stages. Sentiment dips lowest at the Reorder stage (the biggest drop-off) and peaks at the Sell-through stage.
- Awareness: Shelves looking tired. The customer Needs unique, on-trend stock and starts hunting for independent makers. Pain point: Sourcing vetted independent makers is slow, scattered, and risky. Our move: A marketplace that puts vetted makers in one trustworthy place.
- Discovery: So many makers — where to start?. The customer Browses categories, unsure about minimums and lead times. Pain point: Opaque wholesale terms (MOQs, lead times) scare buyers off the first order. Our move: A buyer concierge that explains the terms in plain language and builds a starter order.
- First order: A small order to test. The customer Assembles a sample order spanning a few makers and checks out. Pain point: A clunky multi-maker checkout loses the cart before it's placed. Our move: One cart across every maker, with clear per-maker terms and one confirmation.
- Sell-through: These are flying off the shelf. The customer The products land with her customers and sell fast. Pain point: She's too busy running the shop to notice she's about to run out. Our move: Sales signals surfaced back to her, so success doesn't quietly become a stockout.
- Reorder: Sold out — and forgot to restock. The customer Discovers an empty shelf days after it sold through. Pain point: Manual reordering means stockouts, lost revenue, and buyers drifting to whoever's easier. Our move: The demand agent forecasts her cadence and nudges the reorder before she runs out.
- Advocacy: It basically restocks itself. The customer Reorders on autopilot, adds new makers, and recommends Cargo. Pain point: Without automation, even happy buyers churn back to manual chaos. Our move: Automated reorders and easy expansion turn a first-time buyer into a power buyer.
Scroll the map sideways →
Engagement
Discovery + ongoing build
Timeline
10 weeks to first release
Surfaces
Maker portal, buyer marketplace, ops console
Agent role
4-agent operations system + concierge
The problem
Marketplaces don't die from ugly design; they die from operational drag. Cargo connected makers and retail buyers, but every new maker meant hundreds of messy product listings to clean and categorize, every buyer expected accurate stock and effortless reorders, and a tiny team couldn't keep up. Manual operations, not demand, was the ceiling on their growth.
The approach
We mapped two journeys, not one: the maker onboarding their catalog and the buyer placing a first order and — the real prize — reordering. The product had to serve both audiences from the same data without feeling like one was an afterthought.
The interface design solves the two-front-doors problem: a maker portal built for bulk catalog work, and a buyer marketplace built for fast, confident wholesale ordering, sharing one clean product model underneath.
Then the operations: a multi-agent system acts as the back-office team a five-person startup can't hire. A catalog agent enriches and categorizes new listings, a pricing agent flags anomalies, a demand agent forecasts and nudges reorders before stock runs out, and a buyer-concierge agent helps retailers find products and build carts — with humans supervising the edge cases.
Key moves
01
One product, two front doors
A maker portal for bulk catalog management and a buyer marketplace for fast ordering, built on a single shared product model.
02
An ops team made of agents
Four coordinated agents — catalog, pricing, demand, concierge — handle the listing cleanup, anomaly checks, and reorder nudges that used to need a back office.
03
Reorder before they run out
The demand agent forecasts each buyer's cadence and nudges the reorder at the right moment — turning one-off buyers into recurring revenue.
Design targets
- operating at marketplace scale
- 1 small team
- coordinating the back office
- 4 agents
- reorder nudges from demand forecasts
- Automatic
A self-initiated concept study. Cargo is a fictional marketplace; the agent below is a working demonstration of the buyer concierge.
Next study · 01 Direct-to-consumer product
Anvil →