Warren Shepard has a theory about why factory floors still look like time capsules from the Clinton administration. It's not that industrial operators are technophobes—it's that getting their equipment to actually talk to each other requires the kind of custom integration work that would make even a seasoned systems engineer wince.
"One system monitors equipment, another stores historical data, a third handles control logic," he explains. Months of wiring these pieces together is standard. The result? Operators squinting at dashboards that wouldn't look out of place in a 1998 control room.
Shepard and his co-founder Jack Grodnick think they've found a crack in that old edifice. Their startup, Control Seat, emerged from Y Combinator's S26 batch this year pitching what they're calling "the operating system for the physical world." The premise is straightforward, if ambitious: unify industrial monitoring, control, and analytics under one interface while still connecting to the legacy SCADA systems, historians, and programmable logic controllers that run most manufacturing operations.
It's a market wedged between enormous incumbents and decades of technical debt. Whether a two-person team can scale fast enough to matter remains unclear. But they're moving quickly—perhaps more quickly than the plodding sales cycles of industrial software typically allow.
Software That Speaks Factory
The platform launched publicly on July 27, though an initial release shipped in April. By late July, the team had added soft PLC capability, a sprint that reflects both founders' pedigree. Shepard spent time on Starlink at SpaceX and in security at Microsoft. Grodnick worked on SCADA at Tesla and SpaceX before a stint in generative AI at Amazon. They met at Dartmouth, founded Control Seat in 2025, and got into YC's batch shortly after.
At its core, the product has four pieces that slot together. The SCADA layer handles monitoring and visualization—Control Seat claims sub-100 millisecond latency and the ability to process over 10,000 tags per second at 99.9% uptime. It connects to legacy systems via OPC UA, MQTT (Sparkplug B), AVEVA PI, and Ignition, pulling real-time and historical data into one view.
Then comes the Intelligence layer, which is where things get interesting. Engineers can ask questions in plain language—"find anomalies in pump pressure over the last week"—and the system surfaces correlations, generates dashboards, runs 24/7 anomaly detection. Support exists for VTScada and Ignition today. Siemens WinCC and Rockwell FactoryTalk are marked "coming soon," though no timeline is specified.
The Virtual PLC is a software-defined runtime that executes IEC programs on standard Linux industrial PCs or a preconfigured hardware platform the company calls the FR202. The pitch: abstract I/O mapping from tag addresses, share tags across the entire stack, manage deployments with versioning and rollback. Cloud DevOps patterns, essentially, applied to factory-floor logic.
Rounding out the suite is predictive maintenance, integrating wireless vibration, ultrasound, and temperature sensors that mesh up to two miles line-of-sight and claim five-year battery life. Sensor data flows into the same tagging and alarming system as process data—one source of truth, at least in theory.
The Integration Gambit
Software platforms face a particular kind of friction in critical infrastructure. Deployments are measured in months. Downtime is unacceptable. Asking a plant manager to rip out working systems and replace them with a startup's product is a hard sell.
Control Seat's answer, announced in June: become an "AI-native integrator." Offer to deploy the full stack—PLCs, SCADA, networking, historians, analytics—as a service. The claim is roughly 10x faster integration compared to traditional methods, with AI assisting configuration and commissioning.
It's a strategic hedge. By bundling software and integration, the company positions itself to own the entire implementation rather than license a product and hope customers figure it out. Whether that model scales with a two-person team is another question entirely.
The platform already handles 100,000+ tags, according to updates posted in June and July 2026. The company added MQTT write paths, external historian support, and a trends workbench with 21 analysis tools—forecasting via AutoARIMA, FFT, pattern matching, regression—over a span of weeks. That pace is unusual for industrial software, which typically moves with the urgency of a glacier.
Competing With Giants
The SCADA market sits somewhere between $13 billion and $29 billion, depending on which research firm you ask and how they define the category. Methodologies vary. But the direction is consistent: growth driven by aging infrastructure and mounting pressure to digitize.
Honeywell launched an AI-powered control room assistant in March, with early pilots reportedly predicting pressure issues 10 to 18 minutes earlier than human operators. ABB introduced "Unification X" for control room integration in April. The incumbents are moving, if not quickly.
Control Seat is entering a space where sales cycles are long, incumbents are entrenched, and the software often feels decades behind consumer tech. The founders are betting that fragmentation creates an opening—that the sheer number of systems operators are forced to stitch together creates enough pain that a genuinely unified platform might gain traction despite the odds.
Beyond YC backing, third-party aggregators like Dealroom show a $125,000 seed entry, though that figure appears incomplete and carries the flag "Made with AI," suggesting automated data collection rather than verified reporting.
For now, the team is shipping features, integrating deeply, and making a straightforward pitch: one platform, one source of truth, and an AI layer that speaks English instead of ladder logic. Whether that's enough to crack a market dominated by industrial giants will take more than a few months to answer.
The product is live. The integrations are happening. But in industrial control, where a single misconfiguration can mean catastrophic failure, trust doesn't come quickly. And trust, more than technology, may be what Control Seat needs most.
