On October 6, 2026, Climate Investment held its Annual Investor Meeting, AIM 2026, in London. Climate Investment is the investment firm founded by the Oil and Gas Climate Initiative (OGCI), and the room held operators from OGCI member companies, their investors, and the founders those investors back. SensorUp CEO Terry Cunningham had a short slot on the main stage. He told one engineer’s story, showed what that engineer’s morning looks like when agents do the legwork, and let a four-legged robot finish the point.
Here is the full presentation, including the ending.
Watch it on YouTube to open it in a full-size player. The rest of this post explains what was going on and why the robot’s entrance was not the point.
The industry is not short on data. It is short on a system that acts on it.
Terry opened with an engineer he knows. When a potential high-stakes alert hits his desk, he gets maybe 10 minutes to work out what to do. He checks three systems to find the source, and sometimes none of them have it. To assemble one picture of what is happening on his site, he logs into seven or eight disconnected systems by hand. He calls it the easter egg hunt. By the time he has the picture, the 10 minutes are gone. He reports the problem and moves on. Nothing gets fixed for good.
Most industrial software described operators’ problems better than it solved them. Operators already have the data. What they lack is a system of action, a distinction we wrote up in system of action versus system of record.
Agents do the legwork. The engineer approves.
Terry walked the room through what the engineer asked for, and what it requires. Every signal from field monitors, control-system tags, and robots lands on the asset it describes, without a person dragging it there. Agents run continuously on that record, evaluating conditions against operational, production, and climate goals. When a detection arrives, the picture is already assembled, the risk is already ranked, and the action is already proposed. The engineer approves from his phone, and the problem gets fixed rather than filed.
SensorUp spent years building that inside operators. Today 5 OGCI members run SensorUp in production. The product is SensorUp Studio, and the one-line description Terry used is the agentic system of action. The architecture behind it is in the six pillars of an agentic system of action.
An OGCI operator already runs its compliance program this way
The most important part of the session happened before Terry reached the stage. A major upstream operator, an OGCI member that runs its emissions and compliance work on SensorUp, described how that work moved from manual, to automated, to agents doing the legwork with an engineer approving each action.
The operator shared two numbers from its methane program. The median time from detection to closed repair fell from 29.9 days to 5.7. About 855 field visits a year no longer need a truck.
Agent demos are common. What was different was an OGCI operator describing agents as the current state of a production compliance program, not the roadmap. The operator is now taking the same approach into the rest of its operations, which is how SensorUp intends Studio to be used: solve the workflow you came for, then compose the next one on the same data. How that loop gets written down and run is the subject of how to encode the loop and run it.
The robot was the next action on the same loop
Terry’s slides ended with a short video of a normal morning in the near future. An agent notices an equipment anomaly, assembles the evidence, ranks the risk, and proposes an inspection. An engineer approves. Studio dispatches an ANYbotics ANYmal robot to inspect the equipment. The video ends on the robot walking out of frame.
Then the actual robot walked onto the stage.
Terry’s line was that software can calculate the pain, but it cannot walk a rack to fix it. ANYmal, from SensorUp’s partners at ANYbotics, is the part that walks the rack. In Terry’s words, connecting it to Studio was a configuration, not an integration project. He then handed the stage to Péter Fankhauser, CEO and co-founder of ANYbotics, who presented next.
Three things operations leaders should take from AIM London
- The human stays in the loop by design. Every action in the demo, and in the operator’s production program, waited for an engineer’s approval.
- Agents already do the legwork on compliance work at an OGCI operator. Not a pilot. The operator described the current state, with repair-time and truck-roll figures from the same program, and the next step is extending it beyond emissions.
- The robot was the next action on the same loop. Terry described connecting ANYmal to Studio as a configuration, not an integration project.
If someone on your team owns emissions, compliance, or operations reliability and wants to see the same approach on their own data, the shortest path is the free two-day workshop: bring a real problem and a sample of data, we scope it on day one, and your team sees a working prototype on day two.