Breaking the Bottleneck is a weekly manufacturing technology newsletter with perspectives, interviews, news, funding announcements, manufacturing market maps, 2025 predictions, and more!
💥 If you want to chat, feel free to reach out at aditya@machinafactory.org.
🏭 If you were forwarded this and found it interesting, please sign up!
Spot Solutions Won’t Save You: Why Manufacturing’s Real Bottleneck Is Culture, Not Technology 🎙️💬
“Occasionally, I’ll hear that better AI will obviate these problems, but that’s wrong. The AI we want in manufacturing depends on strong foundations, standards and frameworks that provide shared structure, semantics, and context.”
You went from leading product strategy at MakeTime and Xometry to launching AMT’s SF Tech Lab as a bridge between Silicon Valley and traditional manufacturing. What was the biggest culture shock in that transition, and how did it shape your approach to accelerating the industry’s technology adoption?
The biggest culture shock between Silicon Valley and traditional manufacturing was the stark difference in the cost of failure.
In Silicon Valley, failure is expected and relatively cheap. Move fast, break things, iterate. In manufacturing, failure is expensive. Margins are thin, capital is tied up in equipment, and downtime has real consequences.
The SF Tech Lab was created to increase surface area between Silicon Valley and traditional manufacturing, to create more opportunities for exposure, translation, and mutual learning between two worlds that rarely intersect directly. One trend we saw over the past several years was genuinely impressive technologies stalling not because they lacked technical merit or market demand, but because the surrounding ecosystem wasn’t ready to absorb them.
That insight shapes how I think about technology adoption today. Accelerating progress in manufacturing isn’t about pushing companies to move faster or take bigger bets–many are willing to do exactly that when the foundations are solid. Our focus now is on de-risking adoption by building the tooling and standards that let that happen.
The SF Tech Lab exists to cultivate ties between Silicon Valley tech and traditional manufacturing. How do you actually sell a conservative Midwest machine shop on adopting or experimenting with a new manufacturing platform?
There are really two problems embedded in your question. The first is discovery: how does a manufacturer with a real problem even find a technology that can solve it? The second is trust: how does a solution provider position something genuinely new to customers who can’t afford to be early adopters?
One thing that’s interesting is that my roles at MakeTime, Xometry, and AMT all gave me a similar vantage point: they’re all marketplaces. MakeTime and Xometry connected buyers of manufactured goods with suppliers who could make them. AMT, through IMTS, connects buyers and sellers of manufacturing technology. In all three cases, the hard problems were discovery, trust, and onboarding.
AMT’s historical answer to the discovery problem has been IMTS. It’s a massive technology marketplace, over a million square feet, but what makes it work isn’t scale, it’s density. Thousands of one-on-one interactions where buyers and sellers learn from each other, and billions of dollars of commerce are ultimately driven by that trust.
One of the most interesting challenges we’re working on now is digitizing parts of that discovery process. Helping manufacturers find relevant solutions faster, and helping technology providers reach the right customers earlier. In many ways, it’s the same matchmaking problem I worked on at MakeTime and Xometry, just applied to manufacturing technology instead of parts. The trust problem is harder. Manufacturing technology companies often struggle because they’re selling something radically new into environments held together by brittle, aging tech stacks. Even when the value proposition is real, the risk feels existential.
When a conservative machine shop evaluates a new manufacturing platform, even if they believe the tech is sound, the real question is whether they trust the source, understand its relevance to their specific problem, and feel confident they can adopt it without destabilizing their operation.
Clear ROI gets vendors in the door, and shared technical underpinnings make it safe to keep going. Taken together, that opens up low-risk entry points where adoption adds capability instead of disruption. When vendors design for that from day one, their odds of long-term success rise dramatically.
MTConnect has been around for a while, but you’ve described a shift in what standards need to do as AI and autonomy advance. What’s new with MTConnect, and how is its role changing?
Historically, MTConnect focused on exposing reliable machine status and telemetry. That base is still there, but recent releases move MTConnect into modeling manufacturing context and execution.
With the latest release, we’ll have formal models for parts, processes, and tasks. That means MTConnect can describe what is being made, how it can be made across different routings and machines, and how work is coordinated between machines, robots, and other systems in a flexible, distributed way rather than through brittle, centralized control.
AI needs structure, semantics, and state, not just raw signals. MTConnect now provides machine-readable context about what’s supposed to happen, what is happening, and why. That makes it possible to build digital twins that stay in sync with reality, train AI on real operational behavior, and deploy flexible automation that adapts to change rather than breaking when something unexpected happens.
In short, MTConnect is evolving from a machine data pipe into the semantic backbone for AI-ready manufacturing systems. We’ve also increased our focus on developer relations and making it easier to get started with MTConnect.
You’ve called OpenUSD the “HTML of 3D” and positioned it as essential for the Industrial Metaverse. But most manufacturers still struggle with basic CAD interoperability. How do you help a mid-sized shop find a practical entry point into OpenUSD that delivers ROI today, and what’s the minimum viable use case?
The CAD interoperability comparison is a good one, and it’s important to say this isn’t a failure of manufacturers. It’s the result of decades of point-solution optimization, where tools evolved to be excellent at individual tasks but have not learned to interoperate at the system level.
From a practical standpoint, the entry point for a mid-sized shop isn’t adopting a new platform or chasing the Industrial Metaverse. It’s embedding standards quietly into the workflows they already have— CAD, automation, machines, and operations—so data, geometry, and state can finally line up.
OpenUSD plays a role by acting as a neutral composition layer, where geometry, kinematics, and metadata from existing CAD and automation tools can be assembled into a shared, referenceable model without forcing convergence on a single system.
A minimum viable use case can be as simple as having a shared digital model of one production cell, using the CAD you already have, and tying it to what the machines are actually doing on the floor. That delivers immediate ROI: faster commissioning, fewer integration errors, and a shared source of truth across engineering, operations, and vendors.
Occasionally, I’ll hear that better AI will obviate these problems, but that’s wrong. The AI we want in manufacturing depends on strong foundations, standards, and frameworks that provide shared structure, semantics, and context. As autonomy moves up the value chain, standards have to move up the stack as well, defining not just data formats but machine behavior and intent. MTConnect is a good example of that shift in practice.
If we get that right, most manufacturers won’t feel like they “adopted standards” at all.
You’ve highlighted AI’s role in quality inspection and predictive maintenance as part of the “Smart(er) Shop” vision. Beyond those use cases, what’s the next wave of AI applications in manufacturing emerging from Silicon Valley that most traditional manufacturers aren’t tracking yet?
The next wave of AI in manufacturing is realizing that AI at the edges can’t transform the system.
AI has an outsized impact only when it has context and can participate in feedback loops. Most factories weren’t built to support that. So the most radical progress is happening in greenfield environments where AI is assumed from day one.
So one of the most interesting trends I’m seeing is companies starting from a blank slate, building factories, workflows, and engineering organizations that assume AI is embedded from day one. The result is a very different kind of system.
That doesn’t mean legacy manufacturing is doomed, but the bar has moved. And once you’ve seen what AI-native manufacturing looks like, incremental improvement starts to feel very different.
You’ve said the challenge isn’t the existence of technology, but the industry’s ability to absorb it. What’s one specific cultural bottleneck in traditional manufacturing shops that consistently prevents technology adoption, and how do you actually fix culture rather than just throwing more tools at the problem?
The biggest cultural bottleneck I see is “tool-first” thinking.
Too often, new technology is evaluated as a standalone purchase: Does this tool work? Can I justify the ROI on this one problem? That leads to what I call spot solutions. Local optimizations that don’t compound because they’re not designed to work as part of a larger system.
I say this a lot to entrepreneurs selling into manufacturing: you’re not just competing with companies in your category. You’re competing with every other investment a manufacturer could make that promises a clearer or faster return. Most manufacturers already have a long list of technologies that could help them. The real question is, why should they adopt yours now, and what does it unlock?
Even when the desire is there, adoption breaks down if the foundations aren’t in place. Do they have systems that can absorb the technology? Do they have people who can define the problem, manage change, and integrate the solution into daily operations? Do they have the financial and organizational capacity to sustain ongoing investment?
Fixing that isn’t about throwing more tools at the shop floor. It’s about shifting the culture from buying technology to building capability. That means investing in interoperable systems and internal technical leadership so each new investment strengthens the whole instead of adding another isolated island.
The hard truth is that not every company will make that transition. Culture is difficult to change, and the time and capital required are real. But the companies that do succeed treat digital capability as a core, ongoing function of the business, not a series of one-off projects.
At a national level, this is a systemic issue, not an individual failure. That’s why I’ve argued for a more coherent techno-industrial strategy, because the ability to absorb technology at scale is just as important as the technology itself.
Thanks for the conversation. Manufacturing progress is driven by real interactions between people, problems, and technology. IMTS in September is where that happens at scale. Hope to see you there.
IMTS, the International Manufacturing Technology Show, is September 14-19 this year in Chicago.
To contact Ryan, reach out to him on LinkedIn here. He’s always open to chatting and sharing valuable insights.
