Project benchmark for building a weekly basket.
02
Paco
Agentic Shopping Assistant · Mercadona
What if grocery shopping started with your goal, not a product search?
Paco is an agentic shopping assistant designed for Mercadona Online, helping customers plan weekly missions, stay within budget, manage substitutions and build a basket with less effort.
🙂
PacoMercadona shopping assistant
ProductPaco
Brand contextMercadona Online
FocusAgentic AI · E-grocery
PrototypeInteractive Figma Site
StatusConcept Case Study
01 · The challenge
Weekly grocery shopping is not only a search task.
It is a planning task.
Customers move between meal planning, budget decisions, product availability and repeated household purchases. Paco was designed to reduce that cognitive effort while preserving transparency and control.
Project benchmark for abandonment while building menu-based carts.
Project benchmark for abandonment related to out-of-stock products.
02 · The experience
Designed around the
real Mercadona flow.
The following mockups use the actual prototype screens, preserving Paco's green-and-orange visual language, Mercadona's shopping patterns and the interaction hierarchy of the live experience.
01 · ENTRY POINT
Paco lives inside the shopping journey
The assistant is visible from the Mercadona home experience without replacing familiar navigation. Customers can still browse categories, offers and their usual purchases.
Real prototype screen · Home entry point
02 · PLAN A MISSION
One request becomes an editable weekly plan
Paco understands the goal, proposes four balanced dinners using Mercadona products and makes the budget explicit before the customer adds everything to the basket.
Real prototype screen · Agentic meal planning
03 · SMART CART
The agent supports decisions inside the cart
Paco can identify a missing ingredient and detect an unavailable product. The interface proposes the next best action instead of silently modifying the order.
04 · CONFIDENCE & FEEDBACK
Automation ends with clear confirmation
Once the purchase is completed, the experience returns to a simple transactional state with delivery time, order details and an explicit confirmation that reinforces trust.
Real prototype screen · Order confirmation
Explore Paco in context.
The public prototype is the source of truth for the full Mercadona interaction and detailed transitions.
03 · Success framework
Measure whether the agent
actually reduces effort.
These are project targets and projected impacts, not measured production outcomes.
| Dimension | Metric | Target / projected impact |
|---|---|---|
| User efficiency | Time-to-Cart | 18 min → <5 min |
| Business | Average Order Value | +20% to +30% |
| Conversion | Checkout completion | +15% |
| Adoption | Carts interacting with Paco | 35% of MAU in 6 months |
04 · Strategic learnings
Agentic does not mean
autonomous by default.
Accelerate, don't interrupt
Paco structures the answer so customers can act instead of forcing a long conversation.
Stay inside the Mercadona mental model
AI complements familiar browsing, basket and checkout patterns rather than replacing them.
Control builds trust
Recommendations and substitutions remain visible, understandable and explicitly confirmable.
05 · Product thinking
The design challenge was not the chat.
It was deciding what Paco should do.
The experience uses AI where it can reduce repetitive work — planning, ingredient grouping, missing-item detection and substitutions — while keeping final purchase decisions in the customer's hands.