The Employment Crisis Nobody's Talking About

Everyone is debating whether AI will take our jobs. I think they’re asking the wrong question.

The question isn’t whether machines will replace human labor. They will, in many areas, and faster than most predictions suggest. The question is what happens to human agency when the systems that run our lives are owned by a handful of companies that have no obligation to the communities they serve.

I’ve spent over two decades building software. Enterprise systems, national platforms, and tools that millions of people depend on daily. And the pattern I keep seeing isn’t automation replacing workers. It’s something quieter and more dangerous: the slow transfer of community knowledge into corporate infrastructure that communities can’t access, modify, or leave.

Let me explain what I mean.

A small city has a housing authority. For decades, the people who worked there understood the local rules, the exceptions, the human judgment calls that made the system work. Who qualifies for assistance? How to handle edge cases. What happens when demand exceeds supply? That knowledge lived in people.

Then the city bought a software platform. The vendor encoded those rules into code. The employees who understood the old system retired. The new employees learned the platform, not the policy. And now the knowledge lives in a system that the city doesn’t own, can’t modify without paying the vendor, and can’t take with them if the vendor raises prices or shuts down.

This is happening everywhere. In healthcare. In education. In government services. In agriculture. In every sector, local expertise is being quietly absorbed into centralized platforms.

AI accelerates this. Not because AI is malicious, but because the economics of centralization are overwhelming. It is cheaper to build one AI model that serves ten thousand organizations than to build ten thousand systems that each reflect local reality. The result is software that works for the average case and fails for everyone else. Communities don’t get tools shaped by their knowledge. They get tools shaped by aggregate data, trained on someone else’s reality.

The employment crisis isn’t that people will lose their jobs to robots. It’s that people will lose their relevance. When the knowledge that made you valuable, your understanding of how things actually work in your community, your industry, your corner of the world, gets absorbed into a system you can’t influence, you become a user instead of a participant. You operate the platform. You don’t shape it.

When we talk about the future of work, we usually talk about retraining. Learn to code. Learn to prompt. Adapt to the new tools. But that framing accepts a premise I don’t think we should accept: that humans must reshape themselves to serve machines, rather than demanding machines reshape themselves to serve humans.

What if the question wasn’t “how do I become useful to the machine” but “how does the machine become useful to my community”?

There are paths forward. Open-source software is one, though open-source alone doesn’t solve the knowledge problem if the logic is still unreadable to non-developers. Cooperative ownership models are another, where the communities that depend on software collectively own and govern it. Data sovereignty movements are pushing in the right direction, insisting that communities retain control over the information that describes them.

What’s missing is something more fundamental: a shared language for expressing how things work. Not code, which only developers can read. Not natural language, which is too ambiguous for machines to execute reliably. Something in between. Structured descriptions of facts, rules, and events that both humans and machines can understand, verify, and modify.

Imagine if the rules that govern housing eligibility weren’t hidden inside a vendor’s codebase but expressed in a format that a policy expert could read, a community board could review, and a machine could execute. Imagine if switching software systems didn’t mean losing the institutional knowledge encoded in the old one, because that knowledge existed in a portable, readable format independent of any vendor.

This isn’t a fantasy. The building blocks exist. Formal specification languages, declarative rule systems, and open data standards have been around for decades. What’s new is that AI makes them accessible. An AI that helps a domain expert translate their knowledge into a structured specification is far more powerful and far more equitable than an AI that absorbs their knowledge into a black box.

The hard part isn’t the technology. The hard part is convincing people that they deserve to own their own tools. That local knowledge matters. That the person who understands how volunteer coordination actually works in their town is more valuable than the platform that claims to solve volunteer coordination for every town.

I don’t know exactly what the future of work looks like. Nobody does. But I know what it shouldn’t look like: billions of people operating systems they don’t understand, built by companies they can’t influence, encoding knowledge they used to own.

The real employment crisis isn’t about jobs. It’s about agency. And the fix isn’t retraining people to serve machines. It’s building machines that serve people.

Osamah Alghanmi is the co-founder of Almadar, a software studio based in Ljubljana, Slovenia, working on tools for community-owned software. He spent over two decades building enterprise systems, including national-scale platforms in Saudi Arabia, and writes about technology, agency, and the question of who AI should serve.