Someone in that organisation had been trying to fix this for years.
Hundreds of data centres across multiple continents. Every facility had the same problem. Thousands of engineering documents locked inside incompatible systems. Tag naming conventions that changed from site to site. Operations teams who could describe the equipment they needed to maintain but could not find the file that told them how.
The estimates to fix it came back in the millions. Timelines measured in years. The project was shelved. Then proposed again. Then shelved again. Not because anyone doubted it mattered. Because nobody believed the cost.
If you have read the first three editions of this newsletter, you already know why. Edition 1: the £79m ($100m, €95m) daily cost of commissioning delay. Edition 2: 30,000 files, three naming systems, same document. Edition 3: the shift supervisor at 2:47 AM trusting a drawing that was already wrong.
This is the story of the team that solved it.
The first batch
They did not try to impose a universal naming convention. They did not ask every subcontractor to re-export in a single format. They did not write a 200-page specification for how documents should be classified.
They built an extraction engine that reads whatever exists. PDF. BIM. CAD. Excel. Revit. Any format, any naming convention, any classification schema. The engine ingests it all and normalises it automatically.
The first batch of documents went through on a Monday morning. Files that operations teams had been searching for across three systems for months appeared in a structured, searchable index within hours. Not renamed. Not reformatted. Understood.
By the end of the first week, 30,000 documents had been processed. That is the same volume I described in Edition 2 as the entire output of a single data centre construction project. One week. One facility. Done.
£4m to £800,000
The operator had budgeted over £4m ($5m, €5m) per facility for manual remediation. The kind of manual remediation I described in Edition 2: engineers walking the site, photographing installations, redrawing diagrams that already existed somewhere in the system.
The actual cost per facility: £800,000 ($1m, €960,000).
80% reduction. Not on a pilot. Not on a proof of concept. On a live production portfolio, across continents.
Handover timeline: 18 months compressed to 4 weeks. Processing speed: 1,000 documents per hour, where the previous manual rate was 10. The three-year backlog across their entire global portfolio was cleared in one year. Greenfield and brownfield. Hundreds of facilities.
The engineers who had been sorting files were freed to do what they were hired for. Commission the facility. Get it operational. Stop sitting dark.
Run the numbers on your own facility
Here is the calculation I use when someone tells me their handover timeline is "normal."
30,000 documents. 10 per hour manually. 3,000 hours of a commissioning engineer at £80 to £120 ($100 to $150, €95 to €145) per hour. That is £300,000 ($375,000, €360,000) in direct labour before a single system gets commissioned. For one facility.
Now add the cost of every day dark. The Uptime Institute benchmarks downtime at tens of millions per day for a 100 MW data centre. For a 1 GW AI factory, the number I cited in Edition 1: £79m ($100m, €95m) per day.
At 1,000 documents per hour, 30,000 documents process in 30 hours. Not 3,000. That is the difference between commissioning in weeks and commissioning in quarters.
If you run these numbers on your portfolio and the answer does not concern you, you are either very efficient or very optimistic.
The insight nobody expected
The team that built this did not wait for the industry to agree on a universal schema. They did not lobby for a new standard. They did not convene a working group.
They built a system that reads the mess. Every naming convention, every proprietary format, every legacy classification. The system does not need the world to change first. It works with the world as it is.
That is the line between aspiration and capability. The standards exist. I have spent three editions explaining why they conflict with each other. The breakthrough was not a better standard. It was a system that harmonises across all of them simultaneously, without asking anyone's permission.
The operator who shelved this project for years now starts every new facility with it from day one. The project that nobody believed in is now the template.
What this means for what comes next
Edition 5 is about the reason this system was necessary in the first place. Twenty-five standards walk into a facility. They were each developed by some of the most experienced engineers in the world. And none of them agree on how to classify the same piece of equipment.
That is next Thursday.
If you build, operate, invest in, or regulate infrastructure anywhere in the world, this is written for you. Subscribe to Still Dark.
This newsletter lives in the gap between digital delivery complete and permit to operate. That gap is where value dies, and where it can be recovered.
I also co-author The Vistergy Brief at vistergy.com/archive. Satellite and geospatial monitoring, facility lifecycle intelligence, and standards architecture across LNG, nuclear, data centres, utilities and construction. Subscribe to both for the full picture.
Permit to Operate.