There's a take that's been gaining a lot of traction in manufacturing right now: that there's no such thing as an out-of-the-box MES. That every factory is so unique, you have to stitch together a dozen tools and rebuild everything from scratch every time.
I've spent over a decade on shop floors. I understand why people believe that. But I think they're wrong, and I wanted to prove it.
That's why we showed up at ProveIt! 2026.
For those unfamiliar, ProveIt! is one of the most challenging competitive environments in our industry. Fifty-one vendors. Reportedly, 125 manufacturers are in attendance. Three live virtual factories, intentionally messy and incomplete, because that's what real implementations look like. You connect your product to the factory, show what you built, and answer four questions on stage: What problem did you solve? How did you solve it? How long did it take? What did it cost?
No polished slide deck saves you. You either prove it or you don't.
We weren't there to claim that a perfect out-of-the-box MES exists. We were there to prove something more specific: that you shouldn't have to start from zero. A manufacturer shouldn't need 18 months, a team of consultants, and a blank page to get a system that actually works for their operation.
MachineMetrics was built for discrete manufacturing. CNC machines, stamping, and injection molding. That's our DNA. We're deeply connected to machining-centric manufacturers in aerospace, defense, medical devices, and precision job shops.
The ProveIt! virtual factory we were assigned? A multi-site beverage bottling operation. Three sites. Vats, fillers, labelers, palletizers. Continuous batch processing. Not a single CNC machine in sight.
I'll be honest: when I saw the assignment, my first thought was "what did I get myself into?" This wasn't our typical environment. At all.
So we decided to treat it as a real test. How far could our platform stretch before we'd have to start rebuilding?
The answer surprised even me.
I need to be upfront about something: I didn't start this until two weeks before the conference. My entire engineering team was buried implementing new customers after a strong quarter close. So I took it on myself with some help from Vicente, one of our applications engineers. I'm not a developer, or at least not much of one.
Machine connectivity in hours. We spun up our edge platform, connected to the MQTT broker on the UNS, and used our AI-assisted tooling to automatically map data items across more than fifty machines at the site. I'd never connected a tank before. We figured out mid-setup that one of the sensors was reporting weight and flow rate. The system handled it. That kind of flexibility is something I honestly didn't fully appreciate until I was in the middle of it.
Taught the AI to understand bottling. This was the part I was most uncertain about. We have a new feature called KnowledgeHub that lets you load process documentation and SOPs to train Max AI - our native agentic AI - on your specific environment. I fed it the plant description, the functional spec, and some general context on beverage manufacturing, and it started generating useful answers almost immediately. Changeover checklists based on actual SOPs. Shift handover summaries that understood the difference between a filler stoppage and a downstream labeler issue.
We didn't build a beverage MES. We taught our AI how beverage manufacturing works. That distinction matters. It's the difference between customization and configurability.
A custom operator panel. I built a custom operator interface using our developer MCP, APIs, and our front-end design framework. I want to be clear about what that means: yes, there is custom development involved. But it's front-end work on top of a platform that already understands manufacturing data, production scheduling, downtime classification, and shift events. It's not 12 months of back-end engineering starting from a blank page. Vicente and I built something that looks and behaves like a real product in days. That front-end framework has a name now: Carbide. It's how we're making this kind of development accessible to any customer, not just teams with dedicated engineering resources.
An intelligent agent. Using our MCP server and N8N, I built a workflow that runs on a schedule, pulls production data, runs it through our AI, detects anomalies and bottlenecks, and delivers an intelligent brief. The one I showed on stage identified a tank blocking issue. VAT 3 was holding the wrong product and creating a cascading downstream problem. The system recommended a specific action to prevent it from recurring, and the operator panel surfaced the alert in real time. Two prompts to Max AI. One imported N8N workflow. Done.
When I got on stage, the room had already seen a lot of vendors. Here's what I think landed:
The speed story hit differently than I expected. "Two weeks, with two or three people?" I heard that from attendees over and over. Manufacturers have been conditioned to expect 6 to 18-month implementation cycles. The idea that you could connect a multi-site operation, integrate an ERP, configure AI for a new manufacturing environment, and build a custom operator interface in two weeks is genuinely hard to believe. Until you see it live.
The shift handover demo was the moment that slowed people down. Operators recording shift notes. AI generates a handover summary that combines those notes with production events, downtime classifications, changeover times, and output rates. The incoming shift gets that brief before they touch a machine. Tribal knowledge is captured automatically, every shift. I've seen the pain this problem causes on real shop floors. The audience felt it too.
The configurability story resonated with the integrators in the room. Taking a platform built for discrete manufacturing, teaching it beverage manufacturing through KnowledgeHub, and having it generate useful, contextually accurate insights about a process I'd never worked with before. That's the proof that you don't need to rebuild everything from scratch.
The most valuable thing about a conference like ProveIt! isn't the stage time. It's what you hear when you're standing at a booth for four days talking to builders, integrators, and manufacturers who aren't filtering their feedback.
Here's what we heard, and here's what we're doing about it.
We're building faster UNS connectivity. At ProveIt!, the UNS was already set up, so machine connectivity was straightforward. In real implementations, that connectivity process is one of the most time-consuming parts of getting a customer live. During the conference, we built a new UNS Connector that handles ISA95 machine discovery, data item mapping, and adapter configuration in a single click. What took hours will now take minutes. That's the kind of improvement that compounds across every deployment we do.
We're sharpening how we talk about where we fit in the stack. The most common question at our booth wasn't "Does it work?" It was "Where does MachineMetrics fit in my architecture?" For a community that thinks in layers, that's a fair question and one we need to answer more clearly. We're a full-stack platform, which means we own more of the problem than most vendors in this space. But full-stack doesn't mean closed. We publish data, we share via MCP, and we integrate with the systems our customers are already running. Both things are true, and we're getting better at saying both out loud.
We're doubling down on the platform, and we're giving it a name. The version of MachineMetrics that showed up at ProveIt! could connect to a beverage bottling operation it had never seen, learn the process through KnowledgeHub, and generate contextually accurate operator workflows in two weeks. The operator panel, the intelligent agent, the AI-assisted app scaffolding: that's all being formalized into a single capability we're calling Carbide, our custom application builder. Carbide is how customers and our own team will rapidly build, deploy, and iterate applications that extend MachineMetrics into the specific workflows each manufacturer needs, without traditional development timelines and without starting from scratch. What you saw built at ProveIt! is the early proof. More is coming.
Here's what I actually believe after spending a week at ProveIt! with 51 vendors and reportedly 125 manufacturers:
AI is changing what it means to build manufacturing software. The apps that used to take months to develop can now be scaffolded in hours. The knowledge that used to live in SOPs and tribal memory can now be structured, stored, and surfaced to every operator on every shift. The barriers that made out-of-the-box MES feel impossible, the configurability gaps, the vertical-specific knowledge requirements, and the custom integration work are shrinking fast.
That creates an opportunity. Manufacturers who are waiting for a perfect custom solution are going to find that the gap between "out-of-the-box" and "built for us" is closing faster than they expected. The middle ground, a configurable platform that gets you to value fast and lets you extend from there, is becoming the rational choice.
We went to ProveIt! to make that argument. I think we made it.
Watch our full ProveIt! presentation below.
Have questions or want to know more? Join the conversation on LinkedIn here!
Bill Bither is the Co-Founder and CEO of MachineMetrics.