Faster at this job, rustier for the next one

Picture a couple who move to France for work.

One of them is fluent within a year. The other does a course, gets to a decent level, then lets the fluent one handle the “life admin”. The bakery. The landlord. The GP. The utility companies.

It works. Life runs smoothly. Faster, if anything.

3 years later the fluent one is away for a week and a pipe bursts. Water comes through the ceiling of the flat below. The other has to knock on the neighbour's door, then phone the insurer.

They can follow every word. They can't put a sentence together in reply.

Anyone who's lived abroad with a partner who speaks the language will recognise some version of that.

A short video on AI and skill loss sent me to 2 pieces of research this week that describe the same thing happening at work. Including to very experienced people.

Let's dive in.

Bernard Marr, who writes about technology and business, gave it a name this week: AI deskilling. You get faster with the tool, and less able without it. I went and read the research behind it.

The first study comes from medicine. 19 experienced endoscopists across 4 centres in Poland, each with more than 2,000 colonoscopies behind them. Their centres brought in an AI system that flags possible polyps during the procedure. The researchers then compared the doctors' work without the AI, in the 3 months before it arrived and the 3 months after.

Working without the tool, their detection rate for adenomas (the growths that can turn into bowel cancer) fell from 28.4% to 22.4%.

(Budzyń et al., The Lancet Gastroenterology & Hepatology, August 2025; endoscopist experience via The ASCO Post, 13 August 2025)

It's one observational study, and the authors are careful about what it can prove. But 3 months isn't long. And these weren't trainees.

The second is a randomised trial published in January by researchers at an AI lab. 52 software developers learned a coding library none of them had used before. Half had an AI assistant. Half worked it out by hand. Then everyone sat a quiz on what they'd just built.

The AI group averaged 50%. The hand-coding group averaged 67%. And the AI group wasn't meaningfully faster: the time difference wasn't statistically significant.

The detail worth your attention sat inside the AI group. The people who used the assistant to ask how the thing worked kept most of their understanding. The people who handed it the whole task and pasted back the answer scored lowest of anyone. Those sub-groups were small, so treat that part as a signal rather than a finding.

(AI lab research, "How AI assistance impacts the formation of coding skills", 29 January 2026)

Put the 2 side by side. The tool made the work feel easier. The skill and understanding of it thinned out. And in both cases nobody could formally realise it until someone was tested without the tool.

Most senior people I speak to would file this under early-career risk. Graduates who never learn the basics because the tool does them on their behalf. A real problem, and someone else's.

I'd be careful with that, for 2 reasons.

The first is the endoscopists. 2,000 procedures each. Years of pattern recognition. 3 months was enough to move the number. Experience slows the fade. It doesn't stop it.

The second is about which skills go first.

Look at what senior people are handing to AI right now. The first draft of the board paper. The forecast model. The market analysis before the strategy plan submission. The commercial proposal. The note to the CEO that frames the decision.

Those are the things you'd be hired to do anywhere else. A board wants the judgement behind the paper. A fractional client wants the forecast and the proposal. A consultancy wants the market read.

Inside your current company, the gap doesn't show. You have the tools, the analysts, the data and the brand on your email signature. Everything around you props the output up, and the output is fine.

A second option is where you get tested without any of that. A board interview with a blank sheet of paper. Week 1 at a smaller company with no data team and no licences. A client who wants to hear you think, live, in the room.

So the efficiency is real, and it accrues to the job you already have. The cost falls on the jobs you might want next...

Back to the couple in France.

The partner who let their French go didn't do anything wrong. Handing it over was the sensible choice every single day. The problem only showed up on the one day they had to speak for themselves, with water coming through the ceiling.

Career options work the same way. An option you can only exercise with your employer's tools, data and team behind you is weaker than it looks on your CV. And you won't find out which ones have faded until the day you need them.

Keep using the tools. I use them every day. What the research suggests is a change in the order: form your own view first, then use the tool to argue with it. That's the pattern that kept the skill.

So run a quick check. This week:

1) Name your 3 portable skills. The 3 things someone outside your company would pay you for. Be concrete. "Building a 3-year commercial forecast", rather than "strategy".

2) Date the last blank page. For each one, when did you last do it from scratch, without a tool or a team producing the first draft? If the honest answer is more than a year ago, that skill is on the fade list.

3) Do one the slow way this month. Pick the skill with the oldest date. Do the next piece of that work yourself first, start to finish, then use the tool to challenge it. Note where you hesitated. That's what you need to keep exercising.

None of this means leaving. It means making sure the options you think you hold are ones you could actually take, while you're still employed and the timing is yours.

This is the work Zelova is built for. The guide looks at your actual role, works out which of your skills other organisations would pay for, and turns that into a plan with more than one outcome in it. Built while you're employed, so the choice of terms stays with you.

It's a career strategy service, still being built. If any of this feels close to home, the waitlist is open at zelova.ai.

Which part of your job could you still do tomorrow on a blank page, with no tools and no team? And which part would you rather not be tested on?

Reply and tell me. I'd like to hear your version.

David