Building the AI Operating System for American Manufacturing: Inside Harmony AI's Origin Story

There’s something uniquely compelling about a young engineer who gets invited to join a rocket ship. For Brian McKeown, that moment came via a college mentor who saw potential in his hunger to build things, move fast, and solve real problems. What followed is a masterclass in how to take proven product-market fit and scale it into a repeatable, platform-based solution—the exact challenge facing Harmony AI.

From Old School to Manufacturing Innovation

Brian’s journey to Harmony AI is telling. After college at Texas A&M, he found himself working at a traditional company in Dallas—the kind of place that moves at a pace he found frustrating. He wanted to build things. He wanted to move quickly. He wanted to talk to people and see the impact of his work immediately.

When Zachary Hanchui, a mentor from his college days who works in the 3D streaming space, reached out about an opportunity in Chattanooga, Brian was ready. The founders—George, Vic, and G—had already proven their product worked. They’d validated that manufacturing plants desperately needed what they were building. What they needed now was engineering talent to turn that proof of concept into something scalable.

That’s the moment a founding engineer becomes crucial.

What Is Harmony AI, Anyway?

Let’s cut to the core of what Harmony AI does, because it’s genuinely fascinating: it’s an AI operating system for factories.

Here’s the play: Harmony sends someone into a manufacturing facility for 3 to 9 months. During that time, they digitize all the paperwork on the factory floor, connect machine data, and integrate every internal tool the plant uses—ERPs, MES systems, CMS platforms, all that alphabet soup of manufacturing software. Everything gets connected to what Brian calls an “ontological data layer,” which is essentially a unified data foundation for AI systems to build on top of.

Once you have that foundation, the real magic happens. Harmony can deliver AI-powered production scheduling, quality intelligence, inventory optimization, and even automate tedious back-office workflows with AI agents. The kind of work that kills your best people’s productivity every single day.

The philosophy is elegant: AI is only as good as the data it has access to. Get everything working in harmony, connect it all to one data layer, and suddenly your AI can make genuinely intelligent decisions.

Building the MVP: Speed Meets Quality

Here’s where Brian’s insights get really valuable for any founder thinking about how to build and deploy a semi-custom product at scale.

The MVP development process at Harmony is unique because they’re delivering something that’s both standardized and customized. They’re not building a one-size-fits-all product; they’re building a platform that adapts to how different factories actually operate.

When Brian and the team show up at a new facility—let’s say a cabinet manufacturer in Alabama—they start by doing what any good engineer should do: they listen. They talk to everyone. They get in the weeds. They ask questions until they have a complete mental model of how that factory actually behaves.

Here’s the critical insight: the problems manufacturers think they have often don’t align with what a fresh set of technical eyes identifies as the real solution. That’s why the listening phase is so important.

Once you understand the factory’s actual workflow, you can use modern AI tooling to accelerate your learning process and build a vision for how everything connects into the Harmony platform. Then you pitch a solution that fits within Harmony’s scope while actually solving their specific problems.

The Speed vs. Quality Tightrope

This brings us to a decision every early-stage company faces: how much technical debt are you willing to accept for speed?

For Brian and Harmony, the answer came down to their pilot program structure. They commit to going into a plant for 4-6 weeks and solving their hardest problem. It sounds audacious—and it is. But here’s the kicker: they’ve closed every single pilot they’ve run. That 100% conversion rate tells you something about the quality they’re delivering.

But here’s the tension: in those 4-6 weeks, you need to build something useful enough to wow them while also building a foundation solid enough that when the actual contract closes, you’re not asking them to do the pilot all over again. You can’t cut corners that will create a mess downstream.

It’s a delicate balance. Move too slowly and you don’t prove your value in the timeframe. Move too fast and you build technical debt that haunts you during implementation. Brian and the team have clearly found their sweet spot.

Real Stories from the Shop Floor

One of the best parts of this episode is Brian’s story about Well-Born Cabinets, one of their early customers in Alabama. These are true patriots, he says—gun enthusiasts who have a clay shooting range at their factory to entertain customers.

As Brian and the team were building their relationship with Well-Born, they realized something important: personal connection matters. So they did something creative. They reached out to Mossberg (one of their other customers) and asked if they could get two custom shotguns laser-engraved with “Well-Born Cabinets” and “Freedom 2024” on them as gifts.

Mossberg said yes. Brian navigated the entire gun-gifting process in the US. And after demoing the final product at the end of the pilot, they gave Well-Born the guns and then went out to their shooting range to test them out.

It’s a wild story, but it illustrates something crucial: building relationships with customers in manufacturing isn’t just about delivering good software. It’s about understanding who they are, what they care about, and showing them you actually listened.

Building a Roadmap by Living in the Weeds

How do you decide what to build next when you’re early and trying to scale? For Harmony, the answer is elegant: have your team—people who literally live in these factories during implementations—share what they’re seeing.

Weekly product roadmap meetings become a real-time feedback loop from the factory floor. The team shares problems they’re seeing, and the group discusses whether those problems are specific to one plant or if they’re universal manufacturing challenges.

This is how you find the patterns. Every factory owner thinks their operation is unique—and in some ways, they’re right. But Brian’s insight from the International Woodworking Fair is perfect: whether you’re making guns, cabinet parts, or cutting wood, you have machines, you have materials, you have workflows. The underlying patterns are similar enough that a unified data layer can work across industries.

What It Takes to Build This Team

Brian’s final insight on hiring is simple but powerful: “You got to be a dog.”

It’s one of the vetting questions everyone gets asked when joining the team. What does it mean to be a dog? You have to have the hunger. You have to want it. You have to be willing to show up at factories, get your hands dirty, learn fast, and execute faster.

That’s the kind of person who belongs on a team building the future of American manufacturing.

Want the full story? Listen to the complete episode with Brian McKeown on Code Story to dive deeper into Harmony AI’s journey, their technical architecture decisions, and what it really takes to build in the manufacturing space. You’ll hear more about their early customers, the challenges of scaling a platform-based solution, and why Brian thinks manufacturing is the next frontier for AI innovation.