The washing machine didn’t negotiate with the housemaid

Everyone’s arguing about which job AI takes first but what if that’s the wrong question entirely.

I was doing some copywriting for our HR team on the side at work as they wanted their job descriptions to be more on brand and edgier. ChatGPT arrived in 2023 and the first message I got back was we don’t need you anymore, we can do it ourselves now.

And they could…That’s the annoying part. They produced words quickly and felt empowered. Like anyone in that situation, I felt a bit threatened and impostor syndrome quickly kicks in. Am I going to be exposed? Will the HR team tell the founder that we don’t need marketing anymore? But this fear quickly disappeared once I’d seen the new job descriptions they had produced with AI. It was total crap, totally cheesy, cringy, and sounded the same as every other major corporate brand entity.

AI had given them the ability to make copy but it hadn’t given them the ability to know if the copy was any good.

Earlier in my career I made a TV ad in Slovenia for a digital bank in Ireland that featured a homeless guy. It’s a long story, but it kind of was a lesson in what can go wrong when you don’t hire a specialist. And this guy ended up homeless. I thought this ad was really clever and hilarious. It tested well even.

The problem wasn’t that I couldn’t make an ad. Clearly I could as the other two ads I created did really well. The problem was that I didn’t yet know enough to see what when something had crossed a line.

But as soon as it went live, the phones in the office went mental. Every homeless charity in Ireland rang in, totally disgusted and looking back I don’t blame them. I learned a valuable lesson the expensive way, in public, in about two hours. I don’t want to think about how much that ad cost to create.

My guilty pleasure is Downton Abbey. If you’ve ever watched the show, there are two different worlds happening in parallel. There’s the world upstairs, living in luxury, dealing with life’s trivialities. Then there’s the world downstairs, the domestic service, working, trying to scratch out a living. Domestic service around 1900 was one of the biggest employers in the UK. This included butlers, cooks, laundresses, housemaids and whole economies were built on the fact some jobs just needed a person with specific skills doing them.

None of them thought a machine was going to come along and take part of their job. If you’d asked a housemaid what threatened her livelihood, she’d probably have imagined another person who could do it cheaper or faster.

Then the washing machine arrived.

But the washing machine didn’t convince everyone else in the house that they suddenly knew how to be a housemaid.

AI is different.

A washing machine did one thing better than a person. What’s arriving now can have a go at almost any knowledge job. That creates a different problem. It doesn’t only threaten the person doing the job, it makes everyone around them think they might be able to do it too.

Copywriters and accountants keep getting named as the exposed ones. Maybe they are. But I think this is bigger and stranger than a list of jobs disappearing.

Even I catch myself thinking I could code a website or mark up a set of accounts.

And that’s the trap. With AI I can get surprisingly far. Far enough that I can start confusing access to the output with possession of the skill. But I’m nowhere near a skilled developer or accountant. I don’t know what I don’t know. And because the output can look convincing, I will not even realise when something is wrong.

If I can fall for that, plenty of people with even less experience will too.

Here’s the split I think actually opens up. People with experience get better because they know enough to catch AI when it’s wrong. Inexperienced people get faster. But speed can look a lot like competence when the output is polished. They haven’t built the eye yet. They haven’t made enough mistakes to know where the bodies are buried. They haven’t had the career defining failed TV ad.

And AI lets them skip straight past the years when that judgement would normally have been built.

I don’t know how this plays out. Nobody does. All you can do is place a bet, and mine is this. When producing the work gets cheap, knowing what good looks like becomes more valuable. So I wouldn’t go narrower. AI is frighteningly good at narrow. Go broad instead. Learn some philosophy, design, economics, code, copy, whatever you can stand to learn. Stay genuinely deep in one or two things, but know enough about the rest to ask good questions and spot BS. Because AI can make you feel competent at almost anything. The useful skill might be knowing when you’re not.