Why I Built a Manufacturing SaaS with GitHub Copilot

For nearly two decades, manufacturing plants across eastern Quebec relied on a Windows application I built in 2007. Andon-SSP did its job—tracking production status, alerting supervisors to issues, providing real-time visibility on shop floors. But 18 years later, the industrial landscape had changed. Cloud computing was standard. Real-time data expectations had evolved. Multi-tenant SaaS was the norm.

It was time for a complete rewrite.

But here's the problem: I'm a solo founder with a full-time job. Building enterprise-grade manufacturing software the traditional way would take years. I needed to move fast, maintain quality, and ship something production-ready in months, not years.

The decision: Embrace AI-assisted development from day one. Every line of code, every component, every test—built with GitHub Copilot as my pair programming partner.

This is Part 1 of a 4-part series on how I built Andon-SSP, a modern manufacturing SaaS platform, entirely with AI assistance.

The Manufacturing Problem: $500-2K/Hour in Hidden Downtime

Manufacturing environments are unforgiving. When a production line stops, costs accumulate fast:

• Direct labor costs: Operators standing idle
• Overhead allocation: Facility costs continue regardless of output
• Missed delivery commitments: Penalty clauses, customer dissatisfaction
• Downstream bottlenecks: Other departments waiting for parts

Industry estimates put downtime costs at $500 to $2,000 per hour for mid-sized manufacturing operations. For automotive or aerospace? Multiply that by 3-5x.

The worst part? Most downtime is invisible until it's too late.

Traditional approaches rely on:

• Manual reporting (slow, inconsistent)
• Spreadsheet tracking (no real-time visibility)
• Legacy MES systems (expensive, complex, inflexible)

What manufacturers need: Real-time production visibility with instant alerts when issues arise. Not a $500K MES implementation that takes 18 months. A modern SaaS solution they can deploy in weeks.

That's what Andon-SSP delivers.

The Solo Founder Challenge: Enterprise Software Complexity

Here's what building enterprise manufacturing software requires:

1. Real-time data synchronization (production status updates in < 2 seconds)
2. Multi-tenant architecture (multiple plants, strict data isolation)
3. Reliable connectivity (shop floors have spotty WiFi, need RF backup)
4. Offline-capable UI (operators can't wait for network recovery)
5. Enterprise authentication (SSO, role-based access, compliance)
6. Scalable infrastructure (Azure cloud, auto-scaling, monitoring)
7. Comprehensive testing (unit tests, integration tests, E2E tests)

Traditionally, you'd need:

• Backend team (3-4 developers)
• Frontend team (2-3 developers)
• DevOps engineer
• QA engineer
• 12-18 months minimum

I had: Nights and weekends. A full-time job at Nmédia. A vision for what needed to exist.

Traditional approach: Impossible.

AI-assisted approach: Let's find out.

The AI-Assisted Development Decision: Why Copilot Changed the Equation

I'd been using GitHub Copilot for 6+ months at Nmédia for Power Platform development. I knew its strengths:

• Boilerplate generation: Entities, DTOs, CRUD operations—Copilot excels here
• Pattern replication: Once it learns your coding style, it's consistent
• Test generation: From zero tests to 85%+ coverage in minutes
• Refactoring assistance: "Convert to async," "Add logging"—mechanical changes
• Documentation: XML comments, README sections, API docs

But could it handle enterprise architecture? Could it build multi-tenant SaaS with proper security? Could it maintain code quality at scale?

I decided to find out.

Key Principles I Established:

1. AI generates, human validates: Every security-critical piece of code gets manual review
2. Incremental iteration: Don't ask Copilot to write 500 lines at once. Small pieces → test → refine
3. Prompt with context: Detailed comments describing requirements, architecture, edge cases
4. Human owns architecture: AI implements my design decisions, doesn't make them
5. Test everything: If Copilot generates it, I write tests (or have Copilot generate tests I review)

With these guardrails in place, I started building.

What You'll Learn in This Series

This is a 4-part series documenting the entire journey:

Part 1 (this post): Why I chose AI-assisted development for manufacturing SaaS

Part 2 (next): Choosing the tech stack—Blazor, Azure, Dataverse, SignalR—and how Copilot helped evaluate trade-offs

Part 3: Real examples of Copilot in action—generating entities, building SignalR hubs, creating the Windows Gateway for RF devices, writing comprehensive tests

Part 4: Lessons learned—what works brilliantly, what requires human oversight, development velocity metrics, best practices

The Result: 4-6 Months to Production-Ready SaaS

Spoiler for the series finale: It worked.

I shipped Andon-SSP in 4-6 months instead of the 18-24 months a traditional approach would have required. Code quality is higher (85%+ test coverage vs typical 40%). Development velocity is 3-5x faster than manual coding.

The platform is live. Manufacturing plants are using it today. It handles real-time production monitoring, instant alerts, multi-plant coordination—everything the legacy Windows app did, plus modern SaaS capabilities like cloud-based dashboards, mobile access, and multi-tenancy.

Most importantly: AI assistance didn't compromise quality. It amplified it.

See the Result

Andon-SSP is live at outsiderssolutions.ca. If you're struggling with production visibility, WiFi connectivity on shop floors, or outdated legacy systems, let's talk.

Next in this series: Part 2 explores the technology stack decisions—why Blazor Server, why Dataverse, why SignalR—and how Copilot helped evaluate trade-offs and generate proof-of-concepts.

Questions about AI-assisted development for enterprise software? Contact me—I'm happy to share what I've learned on this journey.