Our position
What autonomous operations actually takes
Maturity scales for industrial autonomy already exist. ARC Advisory Group and NAMUR both publish one, and both grade how much of the control a machine is allowed to hold. They are useful, and when NAMUR asked its own members in November 2025 where they sat, most answered level two.
We describe something different: what has to become true at a site, in the order it has to happen. A site cannot hold its own setpoints until it can explain itself, and it cannot explain itself until every signal and every hand-entered reading sit in one record. Each step is worth doing on its own, and each one is what makes the next possible.
- Step 1 Essentials It sees itself
- Step 2 Maintenance Management It runs its own work
- Step 3 AI Advisory It explains itself
- Step 4 KoPilot It holds its own setpoints
-
Step 5 Physical work and business decisions Not ours. Moving material is a robotics problem, and the commercial calls stay yours.
The steps, defined
- Step 1: Essentials
- Every signal the plant produces, in one place, live and historical. Displaces the control-room panel.
- Step 2: Maintenance Management
- The system decides what maintenance is due, dispatches it, and tracks what it cost. Displaces maintainx.
- Step 3: AI Advisory
- Reports that arrive without being asked for, saying what changed and what to look at. Displaces the analyst's monday morning.
- Step 4: KoPilot
- Closed-loop control of the process, minute by minute, with no human at the panel. Displaces setpoints tuned by hand, differently on every shift.
- Step 5: Physical work and business decisions
- A site that moves its own material and decides its own commercial questions. We do not build this, and the section below says why.
Who covers which step
Most tools in a site sit on exactly one step, which is why moving up usually means adding another system that cannot see the last one.
| Step | Usually covered by | What it cannot reach |
|---|---|---|
| Step 1 | Historians and SCADA reporting tools | They never see what a person records by hand. |
| Step 2 | Maintenance systems such as MaintainX, Fiix and UpKeep | They never see the trend that should have raised the job. |
| Step 3 | Analytics and business-intelligence tools | They report on one system at a time, so they cannot explain a run. |
| Step 4 | Almost nobody, in production | Closing the loop needs physics the site can trust, not a forecast. |
| Step 5 | Robotics vendors, and your own management | Neither is a control problem, and neither is ours. |
Koidra covers the first four on one record, which is the only reason steps three and four are reachable at all. An agent can act only on what it can see.
Step 5
The step we do not build
KoPilot automates a decision, and the decision is a setpoint. It reads the site, works out where the operating point should be, and writes it back to the controllers. That is the whole scope, and it is deliberate: setpoints are where the energy is decided, and they are the part a physics model can be trusted with.
Two things sit above it and neither is ours. Moving material with machines is a robotics problem, with a different supply chain and a different safety case. Deciding what to make, what to charge and when to run is your business, and a control system has no standing to answer it.
A vendor who says otherwise is selling the fifth step on the strength of the fourth.
Where most sites should start
Usually at step one or step two, and often at step two first, because maintenance is the budget that already exists and the whiteboard is the thing that hurts. Most customers run there for a while before going further, and that is the right order. Seeing the site and running the work well are worth doing whether or not you ever hand over the setpoints, and they are what makes the handover safe when you do.
Not sure which step you are on?
Tell us what you instrument today and how maintenance gets scheduled. We will tell you which step that puts you on and what the next one would take.
Talk to our team