Control Seat, a San Francisco startup backed by Y Combinator, launched FactoryOS in July 2026, positioning the AI-native platform as a unified alternative to the legacy industrial control systems sold by Siemens, Rockwell Automation, and other longtime incumbents. The company aims to offer a system that can replace or enhance current SCADA platforms like Ignition, WinCC, and FactoryTalk, layering atop the decades-old software that still runs most manufacturing plants.
The audacity of the pitch is matched only by the pedigree of its founders. CEO Warren Shepard and co-founder Jack Grodnick both worked on control systems at SpaceX and Tesla before launching Control Seat. Their bet: that the same AI techniques reshaping consumer software can finally modernize an industrial sector still running on pre-smartphone technology.
"The software that's used to do it right now is very old," Shepard said earlier this year. "We're building a much more modern platform that automates a lot of these workflows and makes the systems significantly easier to use."
That "very old" characterization isn't hyperbole. Walk into most factories today and you'll find SCADA systems designed before cloud infrastructure or machine learning existed at scale. These platforms control everything from assembly lines to power grids, but their age shows in ways that go beyond aesthetics. A McKinsey survey from May 2026 found that while three-quarters of manufacturing COOs claimed to have a global production system, fewer than a third said it was fully implemented. The gap points to persistent integration headaches that newcomers like Control Seat see as an opening.
Four modules, shared foundation
FactoryOS arrives as a suite of four modules, all drawing from a single data layer. Predictive Intelligence handles condition monitoring. SCADA & HMI provides visualization. Virtual PLC manages control logic. Operations Intelligence fields natural-language queries over plant data.
Customers can cherry-pick modules or deploy the full stack, according to the company. The platform connects to existing infrastructure through OPC UA, MQTT Sparkplug B, AVEVA PI, and Ignition. Connectors for Rockwell FactoryTalk and Siemens WinCC appear on the website as "coming soon."
A telling detail in Control Seat's terms of service: FactoryOS only reads plant data, never writing values back to control systems. It's a design choice meant to ease deployment anxieties in an industry where software glitches can halt production lines or worse. The built-in historian can handle sub-second streaming at over 10,000 tags per second, the company claims. Deployments run as managed cloud or self-hosted on-premises, including air-gapped installations for customers who won't or can't connect to the internet.
Sensors and predictions
Control Seat unveiled its Predictive Intelligence module in August 2026, combining sensor data with existing plant historians and SCADA logs to watch for trouble in rotating equipment, fluid systems, and electrical gear. "What's different is where the data comes from," the company wrote at the time.
The startup ships its own wireless sensors alongside the software. These aluminum pucks measure three-axis vibration from 200 Hz to 25.6 kHz, ultrasound up to 82 kHz, and temperature. They're rated IP67 for dust and water resistance, run on battery for up to five years, and communicate over a 900/868 MHz mesh network with roughly one-mile line-of-sight range. The system looks for bearing wear, misalignment, cavitation, clogged filters, leaking valves, phase imbalances.
Whether this approach will gain traction against established predictive maintenance vendors remains an open question.
AI as engineering assistant
FactoryOS embeds AI assistance throughout, letting users draft SCADA visualization screens in natural language while the system handles the data tag connections. The Virtual PLC module offers AI-assisted logic development and compiler diagnostics, deploying to supported Linux industrial PCs. Safety functions, Control Seat notes, remain on dedicated certified hardware.
The Operations Intelligence layer answers natural-language questions over SCADA and historian data. Think "Compare production efficiency between first and second shift last week" without issuing control commands. It's a feature set that echoes what enterprise software vendors have been racing to add everywhere else, now applied to factory floors.
Young founders, big ambitions
Shepard and Grodnick hold computer science degrees from Dartmouth. Before founding Control Seat, Shepard worked on Starlink networks, manufacturing, and flight software at SpaceX, along with a stint in security engineering at Microsoft. Grodnick built SCADA and controls at Tesla and SpaceX, then moved to Amazon where he worked on generative AI infrastructure for what the company describes as 1.5 million employees.
Y Combinator invested $500,000 in 2026 through its standard deal structure, according to an announcement from Dartmouth. The team remains small. Control Seat's Y Combinator profile lists two people, though LinkedIn showed three employees as of mid-September. The company exhibited at the AI Manufacturing & SCADA Technology conference in Atlanta in August, an early signal it's courting the industry's attention.
Incumbents aren't standing still
The giants Control Seat hopes to challenge aren't sitting idle. Honeywell launched its Experion Operations Assistant in March after pilot tests. AVEVA announced in June that it had woven AI across its HMI, SCADA, and data infrastructure products. Rockwell Automation shipped FactoryTalk Orchestration software the same month and positioned its Optix platform for what it calls modern workflows. Even Inductive Automation markets Ignition 8.3 as "ready to help you leverage the power of AI" through REST APIs and Git integration.
The market they're all chasing is substantial and accelerating. MarketsandMarkets projects the global SCADA market will climb from $13.94 billion in 2026 to $24.69 billion by 2032, a compound annual growth rate around 10 percent driven by infrastructure upgrades and digital transformation mandates.
The full-stack wager
"A system people can finally understand is a system AI can eventually run," Control Seat wrote on its mission page. It's a revealing statement of ambition. The company frames FactoryOS as the complete controls stack delivered AI-native, collapsing what traditionally required separate vendors for PLC programming, SCADA development, networking, historians, and analytics.
In the company's July launch announcement, Shepard positioned FactoryOS as a replacement for legacy platforms like Ignition, WinCC, and FactoryTalk. The modular architecture allows manufacturers to adopt pieces incrementally rather than ripping out entire systems overnight.
Whether factory operators will trust critical infrastructure to a startup remains the central question. Industrial software moves slowly for good reason. But if Control Seat can demonstrate reliability while delivering on its automation promises, it may find an audience among manufacturers tired of wrestling with software that predates modern computing paradigms. The founders, at least, seem convinced the moment has arrived.
