Multi-Agent System for $3.1 Billion 🤑

Schneider Electric acquired the Norwegian company Cognite for $3.1 billion in cash. The Cognite project grew out of an oil and gas conglomerate and does one thing: it collects data from fields and plants into a single workspace for engineers. Let's break down what they paid such money for, how it works, and what clients get.

🔡 What Happened
The parent company Aker, which invested in the startup in 2017, received approximately a 20x return — about $1.48 billion. A year before the deal, Cognite's revenue exceeded $170 million, compared to $100 million the previous year.

🔡 What Problem It Solves
On a typical oil platform, less than 1% of collected data is used. Everything else lies in disparate systems: sensors separately, drawings separately, technical passports separately, work orders somewhere in another world altogether.

Cognite collects all this through 90+ ready-made connectors and links it into a single "plant map" — something like an industrial Google Maps. Click on a pump — and you immediately see its readings, repair history, drawing, and location on a 3D diagram.

🔡 Why It Works Better Than Just an LLM
An ordinary language model knows nothing about your specific plant. Like a brilliant consultant who wasn't allowed into the facility: talks beautifully, but knows nothing about your pump and, if asked, will start making things up.
Cognite first "takes the consultant through the workshops" — gives it all documents and instrument readings — and only then places AI agents on top. The number of clients using these agents grew by almost 700% over the year.

🔡 How It Looks in Practice
Agents work like a repair crew, not a reference book. One finds data, another analyzes instrument readings, a third builds a fault cause map, a fourth prepares a work order — and they automatically pass everything to each other.
Result at the Aker BP field: finding the cause of a breakdown, which previously took weeks (in some cases up to 9 months), now takes hours — a reduction of 70–97%. Document digitization accelerated by 85%, manual tests at the Valhall field by 80%.

🔡 How Much They Earn from This
According to an independent Forrester study, companies recoup their investment in less than 6 months and get 465% ROI over 3 years. Average net benefit is $29.4 million. Of that, $14.5 million comes from reduced downtime, $10.7 million from a 1–2% production increase, and another $10.5 million from employees spending a quarter less time searching for information.
Chemical giant Celanese scaled the system to 30 of its 50 plants in less than 2 years and called it "industrial Google": searching for engineering documents went from hours to clicks.

🔡 Why This Matters
Artificial intelligence works more productively where data has been organized before it. First, all plant systems are brought together and linked, and only then do algorithms make sense.

Video: Cognite

🌐 Website | 🎰LinkedIn