Skip to content
← All projects
In production · all plants

4M Change Points

Digitized the 4M change report with sign-off workflows, analytics and an AI assistant.

Node.jsExpressReactTailwind CSSSQL ServerSocket.IORAGConfidential code

Results

  • Report time cut from 55 to 15 minutes (−72%)
  • 1,000+ paper copies eliminated per month
  • Used by every production area across all plants

Screenshots

Puntos de Cambio 4M statistics dashboardChange point log by department and 4M factorForm to raise a new 4MS change pointGeneral reports in progress and rejectedManagement actions and departmentsSystem in Japanese: Spanish, English and Japanese supportPuntos de Cambio 4M sign-in screenNova 4MS AI assistant chat

The problem

The 4M change-point report (Man, Machine, Material, Method) was filled out on paper across every plant. It took 55 minutes, produced 1,000+ copies a month, and tracking actions and sign-offs was slow. Users are production areas and the Quality team.

Features

  • Digital 4M report capture
  • Action items with due dates and evidence
  • Sign-off chain via badge QR code, rejection with reason, and Quality close-out with full history
  • Dashboards with interactive charts and NG (non-conformance) reports
  • PDF and Excel export
  • Interface in Spanish, English and Japanese
  • Automated HTML email notifications
  • Nova 4MS: a RAG-based AI assistant with automatic failover across Grok, Gemini and OpenRouter

Stack & why

Node.js and Express, React, Tailwind CSS, SQL Server, Socket.IO and Highcharts.

Architecture

Browsers in each area hit an Express app running as a service on Windows Server over HTTPS. Express reads and writes to SQL Server, pushes real-time events through Socket.IO and sends email through Nodemailer. Nova 4MS retrieves context and queries an LLM, failing over to the next provider automatically.

My role

Solo project, end to end: analysis, design, development, testing, deployment and user training.

Challenges

  • Digitizing a multi-step approval workflow without slowing the floor down: sign-offs by badge QR code instead of username and password.
  • Three languages at a Japanese company, including PDF export.
  • Keeping the AI assistant available through failover across three LLM providers.

Next project

Andon Shift Reports