In September, more than 91,000 people registered for IMTS 2026 in Chicago, and by day two I was fairly sure I'd talked to at least half of them, as well as causing permanent damage to my vocal cords. I came home from that show convinced that machine monitoring as a standalone category had run out of road. The data still matters, but the value has moved underneath the dashboard. The manufacturers growing fastest talked about their systems and the decisions those systems let them make. Very few of them believed it should stand alone as its own solution anymore.
Others on the floor saw a similar shift. ABI Research's analysts wrote that the software conversation has moved past whether manufacturers have good data to how they use it to execute. IoT Analytics described a new industrial AI tech stack forming on top of the traditional systems of record, with knowledge graphs, AI agents and AI-built applications, and noted CIOs taking over industrial software buying.
A trade show tells you what the present and future state of things is that vendors are selling. I wanted to see what manufacturers are actually buying, and why. So last week, my partner in crime, Rutherford Wilson, our Chief Product Officer, and I flew west for three days and three on-sites: Karman Space & Defense, Robinson Helicopter Company, and Conesys. These companies are doing remarkable things, using state-of-the-art technology to deliver best-in-class products, and they were generous with their time and candid about where they're headed. Here's what we took away.
Every CIO we met had deep manufacturing knowledge. They talked about running network cable to machines and deciding which system owns which record as comfortably as they talked about cloud strategy and identity management. Ten years ago, those two conversations happened in different buildings, between people who weren't always on speaking terms. That manufacturing knowledge is paying off. They're building enterprise IT architectures that are also pragmatic and create real value on the factory floor.
That matters because the two sides have long seen core systems differently. LNS Research's August MES findings describe CIOs who view MES as an IT application, while operations teams see it as an OT system that connects islands of automation. When the CIO understands the floor, that gap gets a lot smaller.
These leaders were also clear about where the lines sit. ERP, PLM and MES are the core. The ERP stays the system of record for orders, inventory and financials. Execution at the machine and the workstation needs its own layer. When ERP terminals double as the MES, operators end up working in two systems, the numbers stop matching, and someone spends their Friday afternoon reconciling them in a spreadsheet nobody admits to owning. The goal we heard over and over was one place for operators to do their work.
ERP, PLM, and MES may be the holy trinity, but there are still plenty of jobs that need to get done outside of them, and every company had a list. Are those jobs going to be handled by more vertical software, or by applications built on top of the core platforms? LNS counted four overlapping software categories for maintenance alone, each one adding an integration and a dashboard. We're betting on the latter.
All three companies had progressive AI strategies, and I was impressed by the visions they presented. Some of these roadmaps go out several years and well beyond the factory floor. That said, it was clear that to fund the longer-term investments, they need to create value in the short term too. Nobody gets a five-year AI budget on the strength of a good demo.
In the near term, the needs are pretty practical.
Demand in aerospace and defense is strong, and capacity is the bottleneck. Leaders want to know what every machine is doing without walking the floor, see where capacity is being lost, and prove a return on a first set of machines before scaling to the rest. That starts with data they can trust. Manually entered numbers drift, part counts don't match, and a scheduling engine is only as good as its inputs. Capturing data automatically from the machine fixes the inputs. Connecting it to the job, the operator, and the routing gives you a schedule people actually trust, and decisions you can make during the shift instead of after it.
All three organizations understood the value of machine monitoring and data, and all three knew it was only one piece of a bigger puzzle. The data is the great enabler of real-time process optimization, and that requires connecting it to how the work gets done.
Most manufacturers are at a similar stage. In LNS Research's September survey of about 200 manufacturing decision-makers, 53% were still piloting and scaling industrial AI, and 68% of those with wide implementation described it as a productivity tool for existing teams. IDC's FutureScape 2027 predicts that by 2028, 60% of enterprises will fail to convert their data into measurable business value. The best way to stay out of that 60% is to show value early.
For manufacturing technology to deliver what it promises, it has to make the job easier for the people doing the work. Paper travelers and whiteboards are still common, even in shops running world-class equipment. There is something humbling about a million-dollar five-axis machine with a clipboard taped to the side of it. Where digital tools do exist on the floor, they're often ERP terminals built for transactions, and operators don't like using them. The leaders we met were candid that the hardest part of going digital is cultural. People have to want to use the new tools.
Then there's the knowledge problem. Every shop has a Jimmy who knows exactly why that one machine runs hot on Tuesdays, and Jimmy is retiring in eighteen months. That know-how lives nowhere else, and most of it leaves with him. Leaders want tools that engage operators and capture that knowledge while it's still in the building. As I wrote after IMTS, that only works when capture happens as a byproduct of doing the job and never as an extra task.
LNS Research is seeing the same thing. It reports that over 55% of manufacturers have already scaled connected frontline workforce solutions for safety, quality, and security. Its August work on trust found that companies seeing significant results report three times higher full trust in advanced and autonomous technology in their plants, and that adoption suffers where workers see data as a monitoring tool.
You could put all three companies in the same vertical, but their needs and processes were remarkably different. No out-of-the-box software could support everything each of them does, and all three agreed that rebuilding from scratch was the wrong approach.
The ideal solution gives them the templates they need out of the box, plus the ability to quickly configure it for their own processes.
The differences show up inside companies too. Growth through acquisition means plants with different equipment, different habits, and different systems.
Leaders want to standardize across sites without losing what works at each plant. If one plant figures out a better way to run a cell, they want every other plant doing it too.
Both extremes have a cost. LNS Research warns that a copy-and-paste approach to front-line technology carries "great risk". At the other end, a pure application platform means "the execution logic is yours to build and manage," including design, validation and maintenance. For a long time, those were the only two options. AI has made configuration fast and cheap enough to change that, so you can start with something that works on day one and shape it around how each plant actually runs.
There's a difference between being impetuous and moving quickly. All three companies understood the power of AI, and all three were naturally (and rightly) hesitant to connect their internal data systems to outside AI tools. In aerospace and defense manufacturing, that caution is often a requirement. The requirements were set long before AI came up: deployment in compliant government cloud environments, single sign-on through the company's identity provider, and the controls needed to support defense contracts. Any new capability, AI included, has to fit inside that perimeter. Nobody in aerospace is pasting production data into a chatbot just to see what happens.
ABI Research found in July that the MES market favors AI that humans can validate over autonomous execution, with trust and compliance requirements shaping deployments in aerospace and defense. And the bar keeps going up. IDC predicts that by 2028, 70% of the world's largest enterprises will be able to continuously verify the decisions their AI agents make, up from 14% today.
This is why CarbideML (launching next month, more to come on this shortly) changes the game for our customers. Teams can build and deploy apps inside MachineMetrics with no custom code, with full governance, in a system their IT and security teams have already approved.
IMTS gave us a good look at where the technology is going. These visits gave us a much better sense of what manufacturers need from it. They're ambitious about AI, and they're also practical. They want to see value quickly; they want tools their people will actually use, and they won't compromise on security to get there. We're holding ourselves to the same standard.
Thank you to the teams at Karman Space & Defense, Robinson Helicopter Company, and Conesys for hosting us, and to Rutherford for the long drives, the late dinners, and three straight days of putting up with me.