The Capture Layer

01Usually it only costs money. Sometimes it costs people.

That morning cost a crew half a day. Duplicated work, slow decisions, hours spent proving something that was already written down. Most days, that is the whole price of the gap — expensive, invisible, absorbed.

Some days the same gap costs more than money.

No crew can act on that many. The warning was in the system. The system had no way to make it findable.

At Texas City in 2005, fifteen people died on a start-up that should never have happened. Everything needed to stop it was known one shift earlier — a tower filling far past a safe level, watched by a crew who then went home. The handover was thin, the logbook was vague, and a decision to hold the unit never crossed the gap between two crews.

And in 1977 a relief valve stuck open at the Davis-Besse nuclear plant and left its control room confused. Engineers inside the manufacturer wrote it up; by February 1978 they had drafted new operator guidance for exactly that failure. Thirteen months later the same failure hit Three Mile Island, and the operators misread it the same way. The guidance was still inside the company.

The answer to Three Mile Island existed before the accident. In writing. Addressed to the wrong place.

Look at what those three have in common. In every one of them, somebody wrote it down. The information existed, in the plant, before it was needed — and it still failed to reach the person standing in front of the problem.

So the writing down was never the step that failed. Which raises the question this whole piece is about: what exactly is the step that did?

02Capture is not storage — and it is not analytics

Most plants are drowning in data and starving for information, and the reason is that three different things get called the same word.

Storage is where a number goes after a system records it. Plants are excellent at storage — the historian holds years of tags at sub-second resolution. Analytics is what you do to numbers you already have, and plants buy a lot of analytics.

Capture is the step before both: the moment a human observation becomes structured data. An operator notices a pump sounds wrong. A supervisor decides to keep feeding instead of holding. A technician finds a screen blocked ahead of a planned outage. Right there, in that second, the most valuable information in the plant exists — and in most plants it exists only inside a person.

That's the capture layer. It is the thinnest, least instrumented, most valuable layer in the whole operation.

03Every number remembered, every decision forgotten

We built plants that remember every number and forget every decision.

The historian can replay a trip to the millisecond. Nothing can replay the reasoning. Ask why the night shift held the unit instead of restarting it, six months later, and you will get a shrug and a guess — from a plant that can tell you that pump's discharge pressure at 3:47 a.m. to two decimal places.

Try the arithmetic on your own plant. Take one control room, one twelve-hour shift, and count the calls a supervisor actually makes before handover: restart or swap to standby, hold or keep feeding, escalate now or watch it one more hour. Put your own number on it — mine is around thirty, and I'd expect most supervisors to say that's low.

Then ask the question that matters. Of those, how many could the plant reconstruct a month later — not the outcome, the reasoning? On every plant I have worked, the honest answer is almost none.

04The three leaks: radio, notebook, memory

Frontline knowledge escapes through the same three holes in every plant I've seen.

The radio. The richest live stream in any operation is voice traffic — the state of the plant, transmitted continuously by the people standing in it. It has no storage. Every transmission evaporates the moment it ends. Nothing is written, nothing is searchable, and the only record is whatever the listener happened to retain.

The notebook. Not in the control room — in a field operator's back pocket, in the margin of a printed round sheet, on the inside cover of a technician's day book. The reading that runs hot on that exchanger but is nothing to worry about. The pump that primes on the second attempt and never the first. The starting sequence that only behaves on humid days.

People keep these because the official system is slower than the moment. By the time you've found a terminal, logged in and picked your way to the right field, the job has moved on — so the note goes in the pocket, where it takes three seconds.

That isn't a bad habit; it's an accurate read of the trade. But it means the plant's most practical knowledge sits in handwriting, in a drawer, in a locker — findable by one person, and only if you already knew to ask them.

Memory. The largest leak, and the slowest. A problem that beat the manuals for weeks gets solved in an afternoon because one operator remembered what the machine sounded like the last time it did this. That knowledge has one storage medium, and it retires.

05The problem isn't who's leaving. It's who isn't standing next to them.

Somebody is always retiring — on its own that proves nothing. What has changed is the way the knowledge used to move.

It moved by overlap. The new person worked beside the experienced one for years. Not in a classroom: on the plant, through enough upsets to see how they thought — what they checked first, what they refused to do at 3 a.m., which alarm they ignored and why. Nobody called it knowledge transfer. It was simply how long two people stood next to each other.

The overlap is what's gone. Crews are leaner, rosters pair people differently, more of the work sits with contractors who rotate through, and a replacement now arrives months before the leaver goes rather than years. The retirements aren't new. The overlap is.

What makes it bite is the other side of the ledger. US mining and mineral engineering programmes fell from 25 to 15, and graduates dropped 39% since 2016. Fewer people arriving, and less time beside the ones leaving.

For twenty years the answer has been to hope the next person picks it up by standing nearby. Nobody is standing there long enough anymore.

06Management of Change is real — and it stops at start-up

This is where anyone who runs a plant will object, correctly: we manage this. Every serious operation has Management of Change, and in the United States it is written into process-safety law — before a modification is made, the standard requires its technical basis, its impact on safety and health, any changes to operating procedures, an authorisation route, training before start-up, and the documents brought into line. It works. Knowledge still leaves, and not because anyone ignores it, but because of what it is scoped to:

• Replacement in kind is exemptBy the standard's own words, and it has to be — you cannot raise a change request to fit the same seal again. But "in kind" is a judgement, and a different seal material is defensible every single time.
• The drawing update is a carried itemRequired, unambiguously — but done later, by someone other than the people who made the change, in a queue behind more urgent work. Items that depend on being carried are the ones that go missing. A system that treats the document state as part of the change closes that by design instead of by diligence.
• It closes at start-upEverything the process-safety standard asks for orbits the modification itself. Then the plant runs on it for twenty years and teaches the people operating it what nobody knew the day it was signed. There is no form for that, because it isn't a change. It's operating experience.

The documentation tells you what was approved. That person tells you what the plant learned to do afterwards.

When they leave, the drawings stay and the operating experience walks out with them — which makes this a capture problem rather than a compliance one. And capture problems are fixable: the people who hold it are still on shift, and nobody has ever asked them for it in a form that takes seconds.

07The ladder underneath "AI-ready"

Here is the sequence I keep watching. A plant decides to get serious about prediction. Someone says, confidently, we have years of data. Then the project opens the data, and the data is tags — pressures, flows, temperatures. Every number the instruments produced, and almost nothing about what the humans saw, decided, or did.

You can build a model on that. You cannot build the model anyone actually wanted, because the interesting events — the near-miss, the workaround that became standard, the reason someone overrode the setpoint — were never captured in a form a machine can read.

Intelligence sits on top of a ladder, and the rungs are not optional:

• 1 · Structure the equipmentA register where every asset has one identity everyone uses. Without it there is nothing to attach history to.
• 2 · Structure the peopleWho was on shift, in which area, responsible for what. Attribution is what turns an event into evidence.
• 3 · Build the capture habitObservations logged at the moment they happen, by the person who saw it, in seconds — not reconstructed at the end of a twelve-hour shift.
• 4 · Accumulate real historyDepth on the same assets over time, including the reasoning, not just the readings.
• 5 · Then intelligenceOnly now does prediction have something worth learning from.

Almost every plant I've seen tries to buy rung five while standing on rung one. The vendor demo works because the demo data is clean; then it meets a real plant's history, quietly underperforms, and everyone concludes the technology isn't ready. The technology was ready. The data was never captured.

You cannot calibrate what you never measured. You cannot predict from history you never wrote down.

08The constraint expired — I went and tested it

For most of my career there was a good reason the capture layer stayed broken. Fixing it meant software. Software meant a vendor, a budget cycle, and a specification written by someone who had never stood a shift. By the time it arrived it solved the problem we'd had eighteen months earlier — so after a while you stop asking. People learn not to want things they can't have.

Mineral processing was never a slow industry — we put expert systems on grinding circuits in the 1990s and cameras on flotation froth two decades ago. When intelligence arrives packaged from a vendor, we buy it. But the everyday layer — logs, tickets, handovers — could only be bought, never bent. So it froze.

I know that constraint has moved, because I went and tested it. I'm a metallurgist, not a developer. I described what a shift actually needs — in the words I'd use to brief a relief operator — and built it, with an AI doing the coding. It runs. I use it myself.

I want to be careful here, because this is exactly where people oversell. It did not make me a software engineer. I still had to know precisely what was worth capturing, still had to test it against real shifts, and still found out the hard way where my own design was wrong. The machine writes the code; it does not know your plant, and it will confidently build the wrong thing if you ask it for the wrong thing.

But the gate moved. The question used to be can we get this built. Now it's do we know what's worth capturing — and that question was always ours to answer.

09What good looks like on Monday

None of this requires a platform to start. It requires a few design principles, and any operation can adopt them:

• File information to the equipment, not to a personA report addressed to an inbox dies with that inbox's attention. The same report attached to the asset answers whoever asks, for as long as the asset runs.
• Make important things survive by defaultAn open item should stay open until someone closes it — not until someone forgets to re-write it. "He'll remember to mention it" is not a safety system.
• Let items show their ageIf a watch-item has crossed two handovers it isn't open anymore. It's stuck, and it needs replanning, not another hand-off.
• Answer every report visiblyOperators don't stop reporting because they stopped caring. They stop because the last three reports went nowhere. The response side of the loop is what keeps capture alive.
• Judge any tool by the 3 a.m. testIf it only works when the A-team is awake and unhurried, it doesn't work.

None of these are exotic. They are the disciplines a good supervisor already applies by force of personality — written down and made structural, so they survive that supervisor moving on.

The plants that get real value out of AI in the next decade won't be the ones buying the biggest models. They'll be the ones that started capturing properly three years earlier — because when the tools finally arrive, they'll be the only ones with anything worth pointing them at.

So the question I'd take onto shift tomorrow isn't what could AI do for this plant. It's simpler, and harder: when your crew has a question at nine in the morning, is the answer addressed to them — or to someone's inbox?

Sources — BP-Husky Toledo alarm count and fatalities: U.S. Chemical Safety Board final investigation report, June 2024. Texas City 2005: U.S. Chemical Safety Board report 2005-04-I-TX. Davis-Besse 1977 and Three Mile Island 1979: NRC and Kemeny Commission record. US mining workforce retirement, programme and graduate figures: Society for Mining, Metallurgy & Exploration, Workforce Trends in the U.S. Mining Industry. Management of Change requirements: U.S. Occupational Safety and Health Administration, Process Safety Management of Highly Hazardous Chemicals, 29 CFR 1910.119(l).