RAWF - El Alimento BARF Correcto
AI-AugmentedOps-First

Designing an
Ops-Tight,
AI-Augmented
E-commerce System

A premium raw dog food brand with weekly local delivery in Monterrey. The challenge wasn't selling dog food—it was designing a system where rules, automation, and AI work together to reduce operational load.

View Live Site
// project.config
role:
"Full-Stack Engineer & Product Architect"
focus:
"Ops-driven systems, AI workflows, rule-based automation"
stack:
"Remix, PostgreSQL, Prisma, Stripe, Mapbox, OpenAI"
market:
"Monterrey, Mexico — Premium pet food delivery"
Live Product

The Platform

Homepage

Homepage

Product catalog with formula selection and "How it works" flow

Homepage
Homepage
About
About
FAQ
FAQ
Contact & Zones
Contact & Zones
Context

Hard Constraints

RAWF operates under strict operational constraints that typical e-commerce solutions can't handle.

Weekly delivery cycle
Saturday only
Fixed order cutoff
Wednesday night
Delivery-zone enforcement
Monterrey metro area
Perishable inventory
Cold chain required
High education needs
BARF diet guidance
Typical Solutions
  • Manual support via WhatsApp/Instagram DMs
  • Repeated explanations of the same questions
  • Humans validating orders and addresses
Compounding Problems
Operational Fragility
Manual processes break under load
Support Fatigue
Same questions drain team energy

AI was introduced not as a novelty, but as a force multiplier for a rules-driven system.

The Problem

The Core System Problem

How do you design a platform where:

Customers can self-serve answersAI handles repetitive cognitive loadHumans intervene only for exceptionsAI does not bypass operational rules

AI hallucinating policies

Making up business rules that don't exist

Accepting invalid orders

Chatbot bypassing delivery zone or cutoff rules

"Support theater"

AI that still requires humans to double-check everything

Disconnected AI

Bolted on without access to real system state

Strategy

The Guiding Principle

"AI can explain, guide, and assist — but rules must still be enforced by the system."

01

Rules First, AI Second

System enforces constraints; AI operates within them

02

AI Grounded in Real Data

Connected to actual inventory, orders, and policies

03

Support Deflection Over Novelty

Reduce tickets, not impress with gimmicks

04

Human-in-the-Loop Escalation

Clear paths for AI to hand off complex cases

Execution

What Was Built

// core.stack
RemixFull-stack
PostgreSQLDatabase
PrismaORM
StripePayments
MapboxZones
RedisJobs
Fly.ioHosting
OpenAIAI

AI Capabilities

Product and ingredient explanations
BARF feeding and portion guidance
Policy explanations (cutoffs, zones)
Order status lookup
Checkout guidance

Explicit Non-Capabilities

Cannot place or modify orders
Cannot override cutoffs or zones
Cannot issue refunds

These boundaries are enforced at the system level, not just prompts. AI cannot bypass what the code doesn't allow.

Outcome

What Changed

Reduced Support Volume

AI + FAQ deflects repetitive questions before they reach humans

Clearer Customer Expectations

Self-serve answers about policies, ingredients, and timing

Lower Operational Load

Team focuses on exceptions, not routine inquiries

Extensible Architecture

Foundation ready for future automation and AI features

Lessons

Key Takeaway

AI is most valuable when it removes interruptions, not when it replaces judgment.

The goal isn't to make AI do everything—it's to let humans focus on what only humans can do.

Why This Matters

What this project demonstrates

Disciplined AI Integration

AI as a tool within a rule-based system, not a replacement for structure

Real-World Systems Thinking

Designing for constraints, not ideal scenarios

Ops-First Engineering

Building software that reduces operational load, not just ships features

RAWF

Rules, automation, and AI working together.

Ops-tight. AI-augmented. Human-centered.

Visit rawf.mx