Hey friends,

Before Enterprise Monkey, I was a mechanical engineer. Not a good one. I ran production at Honda, and the thing I remember most clearly is how often the line stopped for stupid reasons.

The first thing I noticed walking in was that everything the place knew was on paper. Operating procedures, countermeasures, the fix somebody worked out three years ago for a problem we were currently having again. All of it existed. None of it was findable.

So I went to my head of department, fresh graduate, no standing whatsoever, and asked why we didn't have a knowledge management system. He had no idea what I was talking about. To his credit he let me build one anyway.

I now walk into manufacturers and ask the same question in a different form. Do you have ten or fifteen years of data on what you ordered, what you consumed, and when?

The answer is almost always no.

Then somebody opens a storeroom, and there it is. Manila folders, floor to ceiling. The data was never missing. It just wasn't in anything you could ask a question of.

For most of my career that room was worthless, and everybody knew it. Nobody was ever going to key fifteen years of paper into a system. It's not a hard job, it's an impossible one, and anyone who assigned it would deserve what came next.

Here's the part I want to be direct about, because it's costing people years.

If you go and ask for advice on this, most of the people you'd naturally ask will tell you that first you need a data warehouse. Pull the information out of every system, clean it, standardise it, land it somewhere central, then start asking questions.

I'm not having a go at them. That was the correct answer, and it's what they trained on. Predictive analytics genuinely did need clean, structured, consistent data before it could do anything useful.

That is no longer the prerequisite. Language models read the way a person reads. A scanned page, a handwritten note, an Excel file somebody built their own way in 2014, a PDF nobody has opened since. It can work out what it's looking at and pull the information out.

So the sequence inverts. You don't organise the data so you can ask questions. You ask the questions, and the organising happens along the way.

Less impressive than it sounds, which is why it works.

Put everything in one folder on a computer. Scans, spreadsheets, exports, whatever shape it's in. Then use Claude Code, which comes with the same subscription as Claude and needs no coding at all. You talk to it.

Then give it four jobs, in this order.

One, work out what is actually in here. Not analysis, just an inventory of what exists and what state it's in. Two, make a plan for cleaning and organising it, and show me the plan. Three, start extracting the information and putting it somewhere simple. Four, now analyse it.

Then leave it running. This is the part people find strange. It runs for a day. You come back and check. It runs for another two. You can set it to loop and keep going.

Ten days later you have fifteen years of your own history in a form you can question. Not a warehouse. A simple store, and a machine that knows how to read.

Every single time we have done this, the owner has been surprised. These are people who have run the business for decades and know it better than anyone alive.

Some of what comes back they nod at, because it confirms what their gut already told them. Then there's the rest.

This supplier is late every year in the same season, reliably, and you have never planned for it. That other supplier has never once taken more than two days, and you are holding weeks of their stock for no reason. This station is the constraint, not the one everybody complains about.

None of it is clever. It's arithmetic across more rows than a person can hold in their head, which is exactly the work to hand over.

Here's why I care about this more than any automation we've talked about.

One manufacturer we walked into had close to half their cash flow tied up in inventory. Not because anyone was careless. Because nobody could say with confidence what the production cycle would actually need, so the safe move was to hold more. Safe, and enormously expensive.

Then there's the other failure, which I have also watched. A line down for a fortnight because one supplier was late on one component.

Those look like opposite problems and people manage them as though they are. Hold more, or hold less. They are the same problem wearing two coats: nobody knows what is needed, when.

It's also why single points of failure are worth hunting deliberately. The story that went around during the vaccine rollout was that one of the real constraints wasn't the vaccine at all, it was the little caps for the vials. Plenty of businesses have a cap. Almost none of them have written down what it is.

Answering "do you have the data" with no.

You have it. You have more of it than almost anyone you compete with, because you have been trading for thirty years and they haven't. Saying no because it isn't in a system is how a genuine asset gets described as an absence.

The second version is agreeing that yes, the warehouse should be built first. Six months and a real budget later, nobody has asked a single question, and the appetite is gone.

Pick the worst of it. The folder nobody owns, or the actual filing cabinet.

Take one year of it. Photograph the pages with your phone if that's the fastest route, because it usually is. Put it in a folder and ask one question: what patterns do you see in here that would be hard for a person to notice?

One year. One question. You will know within an afternoon whether the other fourteen are worth doing.

If you are running a manufacturing business in Australia, you are already doing something right. Doing it here rather than somewhere with a tenth of the cost base means you have solved problems your competitors never had to.

But most of what you solved, you solved by being good at it rather than by writing it down. That worked while the person who knew was in the building and had a spare minute.

The storeroom isn't a filing problem. It's thirty years of hard-won judgement, sitting in a form nobody could use, until quite recently.

See you next week,

— Aamir

🎧 Listen to the Podcast Episode on: Spotify | Apple Podcasts | YouTube

📘 Book: The CEO Who Mocked AI (Until It Made Him Millions) by Aamir Qutub

🚀 Dumb Monkey AI Academy

📱 Dumb Monkey AI Academy App: Apple | Android