Walk a midstream control room and you're surrounded by data. Years of it, streaming off compressors, pumps, meters, and valves into SCADA and the historian. Almost none of it gets used to answer a forward-looking question. It's recorded, retained, and occasionally pulled when something has already gone wrong. The asset is sitting right there. The gap is that the data isn't answerable.
Closing that gap, not buying more sensors, is where the value is.
The under-used asset
Most operators don't have a data problem. They have an access problem. The information needed to catch a developing compressor issue or spot an anomaly is usually already in the historian. What's missing is a way to ask it questions in time to act, instead of reconstructing what happened after a trip.
That reframes the build. The goal isn't a new data platform. It's intelligence that sits on top of the telemetry you already collect and turns it into a timely answer.
Connect, don't replace
The instinct of many AI pitches is to replace the systems of record. Operators rightly resist, because swapping out SCADA or a historian is a multi-year project nobody asked for. The better build reads from those systems and writes decisions back, leaving the systems of record exactly where they are.
The historian stays the historian. The control room gains a layer that can answer questions against it. Nothing gets ripped out, which is the only kind of project that actually ships in a regulated operation.
The distributed-infrastructure build problem
Midstream's hard part is scale and geography. Hundreds of sites, thin local staff, and conditions that vary from one station to the next. A model tuned to one compressor won't automatically work on another. Building for that means validating across sites, planning to retrain as conditions shift, and investing in the data plumbing that makes retraining cheap. The plumbing is the durable asset, more than any single model.
What a working example looks like
Kaysee, Cortland's reliability platform, is a concrete version of this pattern on rotating equipment. It reads compressor and pump telemetry from the systems already in place, surfaces developing problems to the engineer who owns the decision, and keeps the human in the loop. It doesn't replace the historian or the SCADA system. It makes them answerable.
That's the shape of turning telemetry into decisions: connect to what you have, keep the person accountable, and build so it travels across sites. The data has been there the whole time. The work is making it answer.