There has been a lot in the press recently about a jobs apocalypse as a result of AI. This is very unlikely to happen. Jobs will change and some jobs will cease to exist, but that is normal progress. The same happened with the invention of the PC, the internet, the steam engine, the spear. Each time, people worried that the new thing would eliminate work. Each time, it mostly eliminated certain kinds of work while creating demand for capabilities that simply hadn't existed before.
Even in software engineering it isn't as bleak as it may first appear. Software has, since its inception, been an artisan discipline. We craft code by hand, line by line, in the same way a cabinet maker shapes wood. The difference is that a cabinet maker can only produce a few pieces a month, and everyone accepts that. In software, the labour cost of artisan production means the backlog never clears. Only products with the biggest budgets get the features they actually need. Everything else sits on a list, reviewed quarterly, deprioritised, and eventually forgotten. The teams who asked for it stop asking.
AI changes that equation. The marginal cost of building something drops dramatically, which means things that were previously uneconomic to build become viable. That is not a threat to engineers — it is a release valve for the enormous backlog of work that has accumulated because skilled engineering time is expensive.
Here is what that looks like in practice. These are the kinds of tasks that used to die in the backlog and now get done:
- Config changes done through a UI — teams have been manually editing config files or emailing requests because the admin interface was never finished. A week of AI-assisted work builds it properly.
- Manual processes triggered by email — "email IT to kick off the month-end run" is a workaround that has outlasted three generations of tooling. Automating it properly is now a day's work, not a sprint.
- Interactive features on company websites — a configurator, a calculator, a smarter search. Things that would have needed a full project with a business case now get prototyped in an afternoon and shipped in a week.
- New product categories with complex domain logic — the features that were always "too risky to estimate" become tractable when an AI agent can draft the scaffolding and you can iterate on top of it.
- End to end testing — the test coverage that never got written because there was always something more urgent gets written now.
All of these improvements ship in days, which means the benefits are visible almost immediately rather than sitting in a queue for nine months. That removes the single biggest headache technology teams face: the perception of inaction. Users ask for things, nothing happens, trust erodes. When the gap between a request and a shipped feature collapses, that dynamic changes entirely.
The companies that will benefit most from AI are the ones that use it to expand what they can build, not to contract their teams. The cost savings from cutting engineers are real but finite. The value of finally being able to build the product you always wanted to build is not.