We build software for the people who keep buildings cold.
This site is not a product launch. It is an open research direction into a problem we kept running into and could not find anyone solving properly — and an invitation to three operators who want to solve it with us.
This is field service software. That is the point.
The category already exists. What does not exist is a version of it built for the half where being wrong costs a facility instead of a return visit.
Your BMS, BAS, DCIM and controls observe the equipment, and they do it well. We read them. We never take control authority and we are not trying to replace any of it.
Dispatch, work orders, approved procedures, technicians, closure. This is a field operating system — that layer is what we are, not something we integrate with.
The two halves talking to each other, so the dispatch decision is made from what the plant is actually doing rather than from what somebody typed into a ticket.
Today, in almost every critical facility we have looked at, the join between those two halves is a person on a phone at three in the morning.
We ran out of software problems.
It did not start with cooling. It started with noticing that every problem we had left was on the other side of the screen.
FieldCamp builds the operations software these contractors run on — dispatch, routing, scheduling, work orders, invoices. Years of that work solved a long list of problems, and every one of them was the same kind of problem: software talking to software.
Routing is solved. Scheduling is solved. Any decent platform can find the closest available technician and put the job on a board. What none of it does is reach past the screen.
The next problem is not another workflow. It is getting software to talk to the physical economy — the plant, the loops, the equipment the work is actually about. Not integrate with it. Talk to it.
Integration means one system exports a number and another system files it. Talking means the software understands what the hardware is doing, decides what that means, and acts on it. That is the step the industry has not taken, and it is the step this initiative exists to take.
It is an AI-first operating system, so the decision belongs to the software: what failed, what it threatens, what happens next. The hardware is where that decision stops being theoretical. When something is physically wrong a person still has to go and put their hands on it — and that person is the one who finds out whether the decision was right.
FieldCamp keeps going on the software end. This project has one focus: making that software meet the physical economy, starting with cooling, because cooling is where being wrong is least forgiving.
Not integrate with the hardware. Talk to it.
Jeel Patel
Leading the initiative · FieldCamp
Jeel leads product at FieldCamp, where he has spent the last three years building AI-native field operations software for mechanical and HVAC contractors — dispatch, scheduling, work orders, and the agents that run them.
That work put him in front of a lot of service managers, and the same conversation kept happening: the software knew where every technician was, and still could not answer whether any of them was allowed to touch the equipment in front of them. On a residential call that is an inconvenience. In a data hall it is an outage.
He started this research direction to find out whether the gap is big enough, and painful enough, to be its own product. That question is genuinely still open, which is why this is a design-partner programme and not a launch.
Three positions we are prepared to be wrong about in public.
They shape what we build first, and what we refuse to build at all.
Read before you write.
Nothing we build gets control authority over a cooling plant. The first version is read-only, human-approved, and outside the control path. Software earns write access over years, not in a pilot.
Topology beats thresholds.
A threshold tells you a number moved. A topology tells you what it means, what is downstream, and how much time you have. Every useful decision in this domain needs the second one.
The boring half is the moat.
Certifications, clearances, shift rosters, warranty boundaries, approved procedures. It is unglamorous work that nobody wants to model, and it is exactly why the generic platforms cannot do this.
Three design partners for 2026.
One service contractor, one operator, one facilities management team. If the problem on this site is recognisable from your own operation, we would rather hear how yours actually goes than tell you how ours works.