Three hours of work for 37 cents
I recently delegated some work to an AI that would probably have taken me three hours and it just cost 37 cents. The task wasn’t particularly difficult, but it involved understanding a system, reading through some data, and working out how to structure everything properly to follow correct procedures. It was the kind of work I knew how to do or taught a junior employee to do, but I would’ve spent a good part of a morning. The hard part was recognizing what I needed and figuring out the simplest result that would solve the problem well enough.
Experiences like this are why I’m optimistic about AI. I think it’s going to be an amazing tool for almost everyone. Being able to hand over time-consuming work gives us more room to do other things, and I expect we’ll get better at using it to produce quality work faster and with fewer errors. There is still judgment involved in deciding what to ask for and whether the result is useful, but the amount of effort needed to get there is changing quickly.
It also makes me think about what companies will value in the people they hire. Knowing how to do something has always been a large part of what we get paid for. When AI can do more of that work, having the knowledge becomes less of an advantage on its own. You still need enough understanding to spot mistakes, but remembering the exact procedure or syntax seems likely to matter less.
Someone still has to decide what needs doing
I think initiative will become much more valuable. Someone who notices a problem, understands why it matters and takes responsibility for solving it can get a lot done with these tools.
Companies will need people who can decide what is worth doing and think through the consequences.
Being able to execute a clearly defined task will be a smaller part of the job.
The difficulty is that thinking takes effort, and I believe a lot of people would prefer to be told what to do and how to do it. A clear set of instructions is comfortable. Deciding what needs to happen means dealing with uncertainty and accepting that your decision might be wrong. AI can help you explore options, but someone still has to care enough to ask the right questions and make a decision.
Working with AI is also a communication skill
When you work with an AI, you’re essentially telling someone what to do. You have a goal in your head and need to explain it clearly enough for the work to move in the right direction. If you haven’t worked out what you want, it’s much harder to communicate it.
Getting useful results means being able to:
- Provide context, explain constraints and identify which details matter.
- Interpret the result and decide whether it solves the problem.
- Explain what was misunderstood, why the result isn’t useful and what needs to change.
The work doesn’t end with the first instruction. Saying “this is wrong” gives very little direction. Helpful feedback gives the next attempt a better chance.
I expect good managers and people who have practice working with others to have an advantage here. They already spend time breaking down work, explaining expectations and reviewing what comes back. Those skills transfer quite naturally to working with AI, even when the task itself requires technical knowledge they don’t personally have.
I also think people who struggle to be good teammates will find this harder. If you routinely leave out context, assume others can read your mind or give vague feedback, those habits will affect the results you get from AI too. Being polite to a chatbot won’t compensate for unclear instructions. The useful skill is understanding what information is missing and helping the work move forward.
Experience still matters, but seniority isn’t enough
I do worry about juniors, in the short term. My expectation is that getting hired will become harder while companies work out how much they can delegate to AI. Longer term, though, I think companies will be able to reconsider if all they need are seniors. Not because AI models are becoming smarter and able to do “senior work”, but because years of experience don’t automatically tell you how productive, creative or proactive someone will be. A person who can identify useful work and carry it through may contribute more than someone with a longer CV who waits for instructions.
Sometimes I fear for my own job and those of my colleagues. At the same time, I think a substantial part of our value is difficult to replace because it comes from understanding the limits of technology. We can judge what is possible, where the complexity sits and what a proposed solution is likely to cost in effort. That understanding also helps us recognise when something that looks simple is going to cause problems.
As building software becomes accessible to more people, that judgment will matter. Getting an application to work doesn’t tell you how difficult it will be to change, maintain or debug later. A decision that saves time today might create dependencies and assumptions that make every future change harder. AI makes it easier to build the software, but someone still needs to understand the technical debt being created and whether the shortcut is worth taking.
No comments yet