Every conversation about AI and jobs starts in the same place: who gets replaced? It’s the wrong question, and it comes from a quiet assumption almost nobody says out loud — that the amount of work in the world is fixed. One pie, finite slices; if the machine takes a slice, someone goes hungry.
But the amount of work has never been fixed. It’s set by what’s affordable to attempt. And that is exactly what AI just changed.
Supply is about to explode
Software is the clearest case. For seventy years, building it has been expensive: specialist people, long timelines, high failure costs. So we only built the software that cleared a high bar — the obviously-worth-it systems. Everything else stayed a spreadsheet, a paper form, a “we’ve always done it this way”.
AI collapses the cost of the attempt. A working prototype now takes days, not quarters. Which means the bar drops — and an enormous backlog of previously-unaffordable software suddenly becomes worth building. The little tool for the five-person team. The automation for the process only one department cares about. The idea you’d never have taken to a development agency because the quote would have made your eyes water.
When making things gets cheaper, we don’t make the same amount with fewer people. We make radically more.
This has happened before, every time. Spreadsheets didn’t shrink the accounting profession — they multiplied what a finance team was expected to do. Word processors didn’t end writing jobs; the volume of written material exploded. Cheaper compute didn’t reduce the number of programmers; it created millions of them.
Demand chases supply
Here’s the part people miss: when supply explodes, expectations explode with it. Once your competitor ships custom software in a week, “we’ll scope it next quarter” stops being acceptable. Once one team automates the boring half of their job, every team is asked why they haven’t. The demand curve doesn’t sit still while the supply curve moves — it chases it.
That chase is why the replacement story falls apart. Meeting tomorrow’s demand doesn’t need fewer people. It needs people who can steer — who know what’s worth building, can judge whether the machine’s output is right, and can carry a prototype through to something reliable enough to trust.
Jobs change shape, not headcount
None of this means nothing changes. The shape of the work changes a lot. Less time producing the first draft — of code, of copy, of designs — and more time on the parts AI is genuinely bad at:
- Deciding what to build. The machine has no opinion about what your customers need. That judgement is the scarce thing now.
- Knowing when it’s wrong. AI produces confident nonsense at the same speed as confident brilliance. Telling them apart is a human job.
- Making it production-grade. A prototype that works on a Tuesday demo and a system your business depends on are different things. The gap between them is where experience lives.
- Owning the roadmap. When building gets fast, the bottleneck moves to ideas. What’s next matters more than how.
So what should you actually do?
Not panic, and not wait. The gap that’s opening isn’t between humans and AI — it’s between people who work with AI and people who don’t. The tool is sitting there. Start with a real problem you already have, something small and boring and yours. Use the machine as a colleague, check its work, and notice where it shines and where it falls over. That’s the whole skill, and it compounds.
This is the bet GoMacRae is built on: AI aiding, not replacing. We use it to make the attempt cheap — and people to make the result trustworthy. The work isn’t disappearing. It’s multiplying. Best get good at it.