Current · Design Lead AI-native ERP Foodservice eCommerce Oct 2024 to now

A custom eCommerce ERP for a 30,000-SKU foodservice operation.

I lead product design across the eight modules. UX strategy, the design system, and the one AI pattern every AI feature here is built on.

Lead
PRODUCT DESIGN LEAD
OCT 2024 → NOW
250+
COMPONENTS
IN PRODUCTION
8
AI USE CASES
LIVE
1 PM · 12 eng
PARTNERED WITH
VUE.JS SQUAD

This is the overview. Deeper case studies cover the design system, the KPI dashboard builder, the AI RFQ experience, and vendor certificate management.

01 · The Product & The Problem

One system, replacing a stack that didn't talk.

HikeOn is a multi-module ERP for a US foodservice and sustainable-packaging brand with a 30,000-SKU catalog.

Catalog Inventory B2B & retail orders Custom-print workflow Fulfilment Returns Suppliers & POs Finance Reporting

When I joined the back office ran on disconnected tools: a separate PIM, spreadsheets for forecasting, a custom queue for print orders, supplier POs in email threads.

There was no shared design system, and AI features were being bolted on one modal at a time. The business was scaling. The tooling wasn't.

02 · What I Owned

Strategy, system, pattern, and the hardest module.

UX strategy across eight modules

Design direction for the whole product, reviewed monthly with the PM and founder.

The design system, end to end

Sole owner: architecture, tokens, governance and the Figma ↔ Vue.js handoff.

The shared AI pattern

The interaction model all eight AI use cases are built from.

The custom-print module

The highest-margin category and the messiest one operationally. I led that module myself.

03 · How I Sequenced It

Foundation first, so engineering never waited.

1
Audit
Every component, pattern and AI touchpoint across eight modules
2
Design system
The shared foundation, before any screen redesign
3
AI pattern
One canonical model, agreed before the next use case shipped
4
Features
Modules built on top of both

I resisted the pull to start redesigning screens. Twelve engineers were already building. Start with screens and they end up rebuilding behind you.

04 · The Foundation

A design system and an AI pattern, both built once.

Design system

250+ components on a three-layer token architecture. Maps 1:1 to Vue.js. Light, dark and density resolve through the same tokens, so nothing forks.

Full case study →

The AI cell

One pattern: suggestion, confidence, "why this?", accept / edit / dismiss. All eight use cases are built from those four parts, so an operator learns it once and knows what they are looking at everywhere else.

See it applied →
ERP sales overview dashboard built on the shared design system
In production Same components, same AI pattern, on every surface in the product.
05 · The Hardest Module

Custom-print orders, where four teams meet.

A single custom-print order runs through artwork upload, MOQ and material checks, a proof cycle, supplier handoff, press scheduling and sample approval. Which suppliers are involved depends on the substrate.

It was tracked in Slack, not the system. The product was heading toward a stack of modals. I built one ordered surface per order instead.

One timeline, not five tabs

Operators were jumping between artwork, supplier and customer threads. The timeline puts all of it in one order.

Proof status is a real state, with an SLA

"Awaiting customer approval" is a state the system holds, and it has an SLA on it. Nobody infers it from an email thread.

Supplier handoff became one action

Every order used to need a brief assembled by hand from three places. Now it is one action.

06 · Outcomes

In production, into year two.

250+
COMPONENTS SHIPPED
FIGMA → VUE.JS
8
AI USE CASES
LIVE IN PRODUCT

The system is in production and still growing. Specific outcomes, like the 85% of vendor certificates that auto-validate without manual re-key, sit in the deeper case studies.

07 · Reflection

The slow start was the point.

Spending the first months on a system and a pattern instead of screens felt slow to everyone except the twelve engineers who stopped rebuilding the same things.
Principle carried forward · HikeOn ERP