back to portfolio
AI ProductActive development

SupremeOS AI — AI-Assisted ERP for Hardware Businesses

An ERP platform for hardware manufacturing and trading businesses, combining core operational modules with a natural-language assistant.

Category
AI Product
Role
Product Designer & AI-Assisted Builder
Focus
Hardware business operations
Status
Active development

/ the problem

Hardware manufacturing and trading businesses often manage purchases, stock, coating or job work, sales, supplier balances, employee attendance and payroll through disconnected registers or software that does not understand their actual workflow.

/ the solution

SupremeOS AI combines core ERP workflows with a natural-language assistant. The system is designed to let users retrieve business information and perform authorised operations through conversational commands while retaining structured validation and workflow controls.

/ business workflow

From raw material to finished sale.

  1. 01

    Purchase raw or semi-finished material

  2. 02

    Receive stock

  3. 03

    Send material for coating / job work

  4. 04

    Track quantities, shortages and rejections

  5. 05

    Receive finished goods

  6. 06

    Sell under the business's brand

  7. 07

    Track accounting and reports

/ core modules

Nine connected modules.

Products & inventory

Item master, stock levels and movement across the workflow.

Purchases & suppliers

Purchase entries, supplier records and outstanding balances.

Sales & customers

Sales entries, customer records and billing details.

Job work & coating

Material sent out, quantities returned, shortages and rejections.

Expenses

Business expense capture and categorisation.

Employee attendance

Daily attendance with weekly paid-leave handling.

Payroll

Salary calculation from attendance, advances and deductions.

Reports

Operational and financial reporting across modules.

AI assistant

Natural-language queries and authorised write operations.

Payroll

Payroll supports monthly salary, attendance, one paid weekly leave, double-rate pay when an employee works on the weekly paid-leave day, deductions, advances, overtime or extra-day pay and final net salary.

/ ai capabilities

Conversational, but controlled.

Natural-language business queries
Structured write operations
Confirmation before sensitive actions
Workflow planning
Multi-step execution
Validation and error handling
Auditability

/ architecture

How the system is put together.

Interface
Web dashboard with a conversational AI panel
AI layer
Local AI model plus Gemini API for language understanding
Operations
Structured, validated actions rather than free-form writes
Data
Relational business records across modules
Controls
Confirmation steps before sensitive or irreversible actions
Build approach
Designed and directed through prompting, testing and iterative refinement

/ challenges & solutions

The hard parts.

Hardware workflows do not fit generic ERP models

Modelled the real purchase → job work → finished stock → sale flow instead of forcing a standard trading structure.

Job work loses material to shortages and rejections

Tracked sent, returned, short and rejected quantities separately so stock stays reconcilable.

Free-form AI writes are risky in business data

Restricted the assistant to structured operations with validation and explicit confirmation before sensitive actions.

Payroll rules are unusual in this industry

Implemented weekly paid leave, double-rate pay for working that day, advances, deductions and overtime.

Language models misread ambiguous business requests

Added workflow planning and multi-step execution so requests resolve into reviewable steps.

Keeping the build coherent without hand-coding everything

Worked as designer and director: specified behaviour, tested output and refined through iteration.

/ current development progress

Where it stands today.

  • Core inventory, purchases and salesWorking
  • Job work and coating trackingWorking
  • Expenses and reportingWorking
  • Attendance and payrollIn progress
  • AI assistant read operationsWorking
  • AI assistant write operationsIn progress

/ what I learned

Lessons from building it.

  • How a real manufacturing and trading operation actually moves material and money
  • Translating business rules into structured, validated software behaviour
  • Designing AI interactions that are safe, confirmable and auditable
  • Specifying, testing and refining a product through AI-assisted development
  • Scoping a large product into modules that can ship independently

/ future roadmap

What comes next.

  • Complete payroll and attendance reporting
  • Expand authorised AI write operations
  • Add role-based access for staff and management
  • Add GST-ready invoicing and statements
  • Pilot the system with a live hardware business

/ request a demo

Want to see it working?

I'm happy to walk through the modules, the AI assistant and the business workflow behind SupremeOS AI.