It is a story that plays out thousands of times a day. Someone twenty-some years into their field, good at it, the kind who actually fixes things when they break. They lose a job, or just go looking for a better one, and they start applying. Forty applications in six months. Tailored CVs, real cover letters, the works.
And they hear nothing.
Not a no. Not “we went with someone else.” Nothing. The applications go into a box and the box never answers. After a while you stop feeling rejected and start feeling invisible, which is worse, because there is nobody to be angry at.
Here is the part that should bother all of us. When a thousand people apply for one opening, very little of that first cut is made by a person. Software scans, scores, and ranks every CV in seconds, and the ones it ranks low get buried so deep that no human ever scrolls far enough to reach them. You are not so much rejected as quietly filtered out, before anyone with the authority to hire you ever really looked.
And while this is happening, the same companies are on stage talking about the desperate shortage of talent.
The talent shortage and the talent filter are the same companies
Start with one number. Workday, one of the biggest makers of hiring software, disclosed in a lawsuit that roughly 1.1 billion job applications had been rejected through its platform over just a few years. In 2025 a US federal court let that case proceed as a nationwide collective action, on the claim that the system was screening people out by age. The lead plaintiff says he applied to more than a hundred jobs through it and was turned down by every single one.
And this is neither new nor a secret. Years earlier, a large Harvard Business School study bluntly titled “Hidden Workers” found that 88% of employers admitted their own hiring software filters out qualified candidates. Not unqualified people. Qualified ones, screened out because their CV did not fit the shape the system expected.
This is bigger than hiring. We have quietly handed a lot of decisions about people over to algorithms. Not recommendations. Actual decisions, with real consequences, made at a speed no human could match. And we did it without ever really deciding to.
The machine that says no
Most large companies run applications through an Applicant Tracking System, and a lot of it is keyword matching. The job says “five to eight years of experience,” and the system is tuned for that. So a candidate with twenty-five years gets flagged as overqualified and dropped. Not because they cannot do the work. Because a number fell outside a range.
A Hungarian survey of senior job seekers this spring put hard numbers on it: nearly 74% got automatic rejections with no explanation, around 80% were ghosted entirely, almost 59% were rated “overqualified.” A gap, a career switch, a title that does not match the template, any of it can be enough for the parser to lose the thread.
And here is the punchline that should end the argument. The Harvard study found that when companies actually hired these “hidden” workers, they performed as well or better, and they stayed longer. The machine is not protecting quality. It is throwing away good people and calling it efficiency.
It is not just jobs
Once you notice the pattern, hiring is just the loudest example.
A gig driver whose whole income depends on an app can be deactivated by that app, over a metric they never see, with no person to call. One day the login just stops working. A machine ended their livelihood, and there is no door to knock on.
A credit score, assembled from data you cannot inspect, quietly decides whether you get the apartment, the loan, the card. When it says no, you rarely get a why. You get a number, and the number is final.
And then the quietest one, the feed. The algorithm decides what you see, which over time shapes what you think about and what you fear. It never rejects you, so it never feels like a decision being made about you. But it is the same machinery, just pointed inward.
In every one of these, the structure is identical. A machine makes a judgment about a person. No human is in the loop. And there is no door to knock on.
Why we actually let it happen
The story we tell ourselves is that the machine is objective. No moods, no bias, just the data, fairly applied.
That is mostly a comforting fiction. An algorithm has no fairness of its own. It is a mirror of whatever it was trained on, our own past decisions reflected back at industrial scale. If the old data was biased, the machine is biased, just faster and with a clean technical alibi.
But objectivity is not even the real reason these systems spread. One reason is volume: a thousand applicants for one role, something has to do the first cut. The second reason is the one nobody says out loud. “The system flagged your application.” “Your score did not meet the threshold.” “The model deactivated your account.” The decision still gets made, someone still suffers it, but no human has to own it. The responsibility quietly evaporates into the software.
That, to me, is the actual function. Not speed, not fairness. Deniability.
One law is trying to put the human back
It would be easy to stop on that bleak note. But something is happening worth knowing about, and since Quiet Nodes is mostly about AI, this is where the two stories meet.
The EU’s new AI Act names exactly the two examples from this post: software that filters job applications, and software that scores your creditworthiness, are both classified as high-risk.
For those systems the law demands things that sound obvious yet barely exist today: meaningful human oversight so a machine is never the final word, a right to an explanation, and an outright ban on scoring people purely by algorithm — that last one already in force. A human who has to be in the loop. A door you are allowed to knock on. The exact things whose absence made all of this feel so cold.
I am not naive about it. Laws are slow, and this one just proved it: the high-risk rules were due to start applying this August, and only days ago the EU pushed most of them to the end of 2027. It is Europe only, and paper is not enforcement. But the instinct is right: not to repeat, with AI, the mistake we sleepwalked into with ordinary algorithms. It is the first serious attempt to keep an accountable human in the decision, on purpose, before it hardens.
The algorithm is the alibi
The algorithm did not take the human out of the decision. The decision is still being made, about real people, with real stakes. What the algorithm took out is the human who has to answer for it.
That is the trade we made without noticing. We did not automate the judgment so much as we automated the excuse. Somewhere right now, a perfectly qualified person is staring at a rejection that no person actually chose to send, and there is nobody on the other end to ask.
I do not think the answer is to smash the machines. Some of this automation is genuinely necessary at scale. But we have quietly accepted something we should not: decisions about people’s lives, made by systems that no one is accountable for.
So I will leave you with the question I keep turning over. Do you think we can still take this back, and put people back in charge of the decisions that fall on people? Or have we quietly accepted that some algorithms just shape our world now, with no one to answer for it?