AI won't tidy the house

Hi everyone.

A few weeks ago we were at a factory, working out where AI could help. Near the end of the meeting, the owner of the company told us there were areas still done by hand that they'd like to see evolve. Nothing new so far. Then came this: "there's no point in you getting there, because other work has to be done first."

We went there to talk about AI, and it was our host who told us where it wasn't worth using yet. It's one of the most sensible things I've heard all year.

It's usually the other way round. The expectation is that AI will show up and tidy whatever is untidy. It won't. AI runs the process that's already there, only faster. If the same file exists in three versions, it finds the wrong one in no time. If a rule only exists in two people's heads, it doesn't follow it. It makes up another one, with great confidence.

Augusto, Waldyn's COO, and I saw this first-hand long before there was a Waldyn. We met at Indie Campers, a company that already had nine-figure revenue and doubled it in 12 months. We saw up close the damage done by scaling a business without solid, well-defined operations. When revenue doubles, everything that was badly sorted out doubles too.

Much of what we do today comes from there. Before we talk about AI, we want to understand the state of the foundations: who does what, in what order, with what information, and whether that's written down somewhere or only in someone's head.

It has happened to us in-house too. In a recent issue I told you we put time and money into cold email outreach, with tools to send everything at exactly the right time. The tools did exactly what we asked. We became very efficient at doing something that didn't work.

That's why, in the proposal we made to that factory, there's a step before the technology and the AI, called "Improve the process": simplify and organise before any tool. Sometimes that's where the biggest gain is. And the rule applies to us too. The colleagues I introduced you to last time only work because we moved our rules out of our heads and into a place where they can be read.

The conclusion fits in one sentence: automating a badly designed process only makes it fail faster.

So if you're thinking of putting AI into some area of your company, the first question isn't which tool to use. It's whether anyone can explain, step by step, how that area works today. If nobody can, you already know where to start.

Talk soon,
Miguel