Behind the tools everyone uses and nobody really looks at..
A job title almost never describes the real work.
Two people with exactly the same title can spend their days doing jobs that have almost nothing in common. One manages internal political crises inside a transforming organization. The other coordinates weekly meetings inside a stable structure. Both are called “Director of Operations.” The system sees a match.
When someone tells their story, they describe what they lived through in their own corner, with what they saw, understood and navigated inside their environment. Even when told honestly, an experience remains partial, compressed by context. Résumés and LinkedIn organize these stories. They make them clean, readable, comparable. We get the impression that we understand skills better, when in reality we mostly learned how to organize stories more efficiently.
Modern recruiting systems rely almost entirely on this logic: placing compressed narratives of work side by side and hoping the correspondence between the words says something about the reality behind them.
Mostly, it says the narratives use the same words.
On the company side, the mechanism follows the same logic. When someone leaves, organizations rarely begin by asking what is truly missing or what should actually change. Someone gets replaced by someone else — same role, same title, same job description — as if the problem necessarily came from the person and never from the work itself. The need gets reformulated out of habit, compared against experiences described from very different worlds, and the result is a real matching rate somewhere around 1.6%. At that point, companies could almost sort candidates alphabetically and decide people whose names start with M are more promising.
Before a human recruiter even looks at a profile, the platform has already worked: standardized titles, expected keywords, career paths that “look serious.” The human recruiter arrives afterward and looks at what the system already filtered on their behalf — a filter placed on top of another filter, whose accumulation distorts the message even further.
We call this recruiting but what we mostly compare are self-presentations, because that remains simpler than understanding what someone actually knows how to do.
What the system optimizes is the readability of narratives — not the reality of the work. And the more it optimizes readability, the more the narratives begin resembling one another, the harder reality becomes to see.
There are still places where this logic breaks — spaces where what gets examined are the traces left during execution, the attempts, the mistakes, the corrections, the choices, instead of a story carefully reconstructed afterward. These spaces remain rare and mostly exist inside technical professions. And when people discover them, the reaction is almost always the same: “Wait — you can actually see how someone works?”
The system is nevertheless beginning to become capable of functioning differently. Connecting trajectories that do not carry the same titles. Reconstructing context instead of simply comparing compatible formulations. Detecting comparable professional realities even when the narratives themselves look completely different. Comparing narratives remains much simpler but nobody is really asking the system to do that yet.
The real always leaves traces.
