Industry and energy · Intelligent Core · 2024 to 2026
Intelligent Core · Industrial AI from data to action
Modular AI which meets industry where it already runs. SCADA, IoT, drones, and APIs, drawn into one operational layer.
£2.7M pre-seed and seed10 modules and solutions90 days first pilot
Industrial operators are measured on uptime and judged on emissions.
A midstream operator runs a plant on systems which were never designed to speak to each other. SCADA in the control room, IoT at the wellhead, inspection data in a folder, compliance in a spreadsheet. Visibility falls apart exactly where a decision has to be made.
The cost is not abstract. A leak found late is a fine and a headline. A compressor drifting out of tune is money leaving the site every hour it runs.
We sat in the control rooms the dashboards forgot.
Intelligent Core began inside a running operation, not on a whiteboard. We sat control room shifts at Purestream, rode the field routes, and logged where decisions stalled. The data always existed. The action came hours later. Two questions carried every shift: what breaks next, and who acts on it.
The gap is not data. It is the distance from data to action.
Every operator we sat with was instrumented. None were early. Signals arrived, meetings formed, and the window closed. The finding set the product: not another dashboard which explains the past. Shorten the distance between a signal and a decision until it fits inside the moment.
Close the distance, not another dashboard.
One layer, assembled from parts which stand alone or work as one.
The architecture is modular on purpose. Each engine is a product on its own, and the same engines assemble into one domain solution when the work calls for it, which is how a plant can adopt it without a two-year program. Insight, Foresight, Autonomy, Relay, Anomaly, and Move make up the core.
On top sit the domain solutions: Horizon, Flow, Seismic, Dose, and Detect. Interface work covers the control room, the field team, and everything distributed between them, with compliance and governance aligned to industrial requirements from the start.
Six engines. Each a standalone product, each addressable alone or in concert.
The lease streams in.
SCADAFlow metersPLC / RTUGISDrone thermal / OGISafety formsNormalized, time-aligned, geo-tagged to lease, section, and pad. Cross-validated against independent signals.
Self-reported in, verified out.Sense-making for the organization drowning in its own data. Volume becomes meaning.
Signal on any time series. The slip is seen days before it lands.
Agentic execution which acts on processed intelligence from Insight and Foresight.
Bidirectional conversational command: ask what the field is doing, then tell it what to do.
Irregularity across infrastructure, flow, production, and seismic.
Crews, routing, and dispatch to the emergency. In build.
Each engine runs alone. Together they become solutions. Between them sits the MCP layer, directing traffic: context in, decisions out.
Five domain solutions sit on the layer. What each one does, and what it runs on.
Every lease and location in one view, with the completion date forecast before it slips.
SoftwareFluid tracked across the network, so every molecule is accounted for from intake to reinjection.
SoftwareInjection volumes read against ground movement, so a disposal well is throttled before a tremor arrives.
SoftwareChemistry measured at the injection point and the pump set from the reading, not from a schedule.
HardwareSoftwareEmissions and leaks found from the air, the ground, and the water, then placed on the map the operators already read.
HardwareSoftwareControl-room telemetry read in place. No rip and replace, no downtime window.
Field devices streamed at the edge, from pressure heads to flow meters.
Aerial inspection folded into the same map the operators already read.
Clean contracts to whatever speaks HTTP. The layer is additive by design.
Years of process history put to work instead of left in the archive.
Work orders and cost lines joined to the physical event which caused them.
Ask the field a question, get a decision back.
The experience is defined as a conversation with the operation. A pressure drop announces itself with its price attached. A question in plain words returns an answer with an action inside it. No operator learns about their own plant from a phone call.
Each engine stands alone. Composed, they run the operation.
Anomalies are detected and routed automatically. Alerts arrive with a recommended response attached, not as a raw signal for somebody to interpret. Conversational control lets an operator ask the plant a question in the language they already use.
The platform was validated in real conditions with Purestream, against live operations rather than a sandbox. Intelligent Core has secured £2.7M across pre-seed and seed funding.
Six engines compose into five domain solutions. Two run in production today. All five are drawn here, each at the stage it has reached.
Upstream oversight. The engines over the basin: schedule to production.
Midstream control. The same engines over the network: intake to reinjection.
Induced seismicity. The same engines over the disposal well: rate against ground movement.
Chemical dosing. The same engines at the injection point: the pump set from the reading.
Emissions and leak detection. The same engines over air, land, and sea: the signature on the map.
The basin, its leases, every location. Delays are forecast before they land.
The basin
Leases and locations in one view. A lease inherits the worst state of its locations, so the eye lands on the risk.
The location
One lease is off plan. The map goes to it before anyone asks which one.
The problem
Pump 7 is slipping completion. Not a lease problem. One pad, one job, one date.
The forecast
Foresight prices the slip at five thousand dollars a day and marks the completion window before it closes.
The response
Autonomy moves the crews, Relay briefs the site lead. The schedule realigns before the slip compounds.
CORE Horizon at intelligentcore.io →
Fluid control across the network. Every molecule accounted for, intake to reinjection.
The network
Intake, three loops, reinjection. Every barrel is counted at the meter it passes.
The loss
Line PL-A2 comes up 980 barrels a day short. The balance does not close, and the network says so.
The location
The gap sits between meters 4 and 5. One segment, not one line, and not one field.
The close
The loop is isolated and the volume reroutes. The balance closes and the number holds.
CORE Flow at intelligentcore.io →
Injection volumes read against ground movement. The well is throttled before the tremor, not after the report.
The injection
Water goes into the disposal formation at SWD-14. Ground movement runs on the same clock as the pump.
The cluster
Small events gather on a fault below the well. Alone unremarkable, together a pattern.
The attribution
The cluster measures 1.4 kilometres from SWD-14 and is tied to it. A distance, not a suspicion.
The throttle
Injection drops 30 percent. The magnitudes fall with it, and the well keeps running.
Chemistry measured at the injection point. The pump follows the reading, not the calendar.
The schedule
The pump doses inhibitor at a fixed rate. It has no idea what the water is doing.
The drift
Residual downstream reaches 26 parts per million. Over band, and paid for twice.
The loop
The analyser reading goes back to the controller. The measurement becomes the instruction.
The setpoint
Dose falls to 2.3 gallons an hour. Residual returns to target and the chemical bill follows the water.
Three ways of looking at the same field. Air, ground, and water, resolved onto one map.
The sweep
A drone line, a ground traverse, and a marine track cover the same coastal field at once.
The signature
Methane reads on the aerial pass. Real, and somewhere inside a circle no crew can be sent to.
The bearing
The ground station cuts a line through the circle. Two hundred metres, from three kilometres.
The order
Pad 07, flange 3, 142 kilograms an hour. The work order lands on the map the operators already read.
Minutes to act, ninety days to prove.
Also recorded in the same pilot: a 40% reduction in flow imbalance events through automated anomaly detection, and a 30% reduction in methane and emission events through predictive monitoring. These figures come from one ninety-day pilot with a midstream operator, and we state the scope because the scope is the point. They are a pilot result, not a fleet average.
Venture. Intelligent Core is co-founded and operated by a Formist founder. Robert Girvin, Co-Founder and Chief Innovation Officer since April 2024.
Client. Intelligent Core, with Purestream as validation partner.
Discipline. Industrial AI · environmental intelligence · platform design.
Team. Robert Girvin and Tim Rowe. Dr. Ndidi Nzeako Anyakora, data science and machine learning. Tom Greenlees, Ismayil Rzali, David Jung, and Adam Gilles. Roman Revazov and Yanina Olenych.
Ownership. Venture-owned, studio-built.
One layer, in the control room, the truck, and the pocket.
Every surface above is the same layer wearing a different frame: the monitor for the room, the tablet for the truck, the phone for the pocket. The state is one; only the glass changes.
The future of industrial operations is a field which answers.
Bring the ambition. Start a conversation.