Is twenty years' experience still worth anything?

If you’ve been doing your job for twenty years, you’ve probably had this thought: the new tools make a beginner nearly as good as me. So what exactly am I for?

It’s a fair worry, and there’s real evidence behind it. But the same evidence tells a second story, and it’s the one worth hanging on to.

The bad news first

The best study we have of AI at work comes from a customer support centre. Erik Brynjolfsson, Danielle Li and Lindsey Raymond followed 5,179 support agents as a company rolled out an AI assistant that suggested replies in real time.

On average, agents with the tool resolved 14% more issues per hour. But the average hides the interesting part. The newest and least skilled agents improved by 34%. The most experienced and skilled agents barely improved at all.

Why? Because the tool had learned from the best agents’ conversations. It was, in effect, handing their know-how to everyone else. A new starter with the assistant behaved a lot like someone with far more months on the phones. The researchers found the tool helped new agents move down the experience curve faster.

So yes: on the routine part of a job, AI compresses the gap between a veteran and a newcomer. If what made you valuable was knowing the standard answer faster than anyone else, the machine now knows it too.

The good news, which is also evidence

Look at who is actually losing ground in the job market so far. Stanford’s Digital Economy Lab, tracking US payroll data, finds that young workers in the most AI-exposed jobs are being hired less, with employment 19% below their peers in less exposed work. Experienced workers in the same jobs show no comparable gap.

That’s the second story. Firms are, so far, keeping the people with experience and hiring fewer beginners. (That’s its own problem, and we wrote about it in where do apprentices come from now?)

Why would they? Because the part of experience that a tool can’t copy turns out to matter more once the tool is doing the routine part.

What experience actually buys you now

Think about what the support-centre AI couldn’t do: it could suggest the standard reply, but it needed someone to know when the standard reply was wrong. That’s the bit experience is made of.

• Knowing when the machine is wrong. A tool that’s right 95% of the time is dangerous in the hands of someone who can’t spot the other 5%. Spotting it takes years.
• The weird cases. Routine problems are exactly what AI learns from. The one-in-a-thousand fault, the customer who’s upset about something else entirely, the drawing that’s technically correct but will never get built: those live in people’s heads.
• Trust. Clients, patients and colleagues trust the person who’s been right before. That’s slow to earn and impossible to download.
• Knowing who to call. Half of getting things done in any organisation is knowing who actually decides. No tool has that map.

You can see it in the task lists here. Across the site, tasks like advising, deciding, negotiating and supervising are consistently the ones AI reaches least. Those are the tasks people are usually given after years in a job, not before.

What to do with twenty years

1. Stop competing with the tool on speed. You’ll lose, and it’s the least valuable part of what you know.
2. Become the person who checks it. Every team using AI needs someone who can look at the output and say “no, that’s wrong, and here’s why”. That’s you.
3. Write down what you know. Seriously. The support-centre tool was built from the best agents’ work. Experience that’s captured, taught and passed on is worth far more than experience that stays in one head.
4. Look at your own tasks. Slide the weekly audit on your job page towards the tasks you actually spend your time on. If your week is heavy on the routine, it’s time to move towards the judgement end of your job.

Twenty years is still worth a lot. It’s just worth it for different reasons than it was five years ago: less for what you can produce, more for what you can tell is wrong.