Made for Swiss self-employed & SMEs
In October 2026, an open-source project called AI Job Search made the rounds in tech circles: it turns a command-line AI assistant into a whole application team. You describe yourself once; the software then searches job portals, scores how well each opening fits, and writes a CV and cover letter tailored to that specific posting — before a second AI reads the draft, criticises it and improves it. Its author, a geophysicist who lost his job in 2025, reports 69 tailored applications, 21 first interviews and one signed contract. If you hire for a small Swiss business — an apprentice, an administrator, a new specialist — this story is really about your inbox.
What the tool actually does
In plain terms, the project chains several AI steps into one pipeline. One step holds your profile, built from a CV, an export of your LinkedIn page or a short interview with the AI. Another searches job portals and ranks the openings that suit you. A third writes the application, and a fourth — a “reviewer” agent — criticises the draft and revises it before anything is sent. The result is typeset with professional software, so it looks immaculate.
The honest footnotes: it needs a paid subscription to Claude, Anthropic’s AI assistant (there is no free tier); it was built around Danish job portals, so other countries have to plug in their own boards. And the 69 applications, 21 interviews and one contract are the author’s own account, not an independent study. Your profile data stays in your own copy of the project — provided that copy is private, not a public one on GitHub.
Why this lands on your desk
Until recently, a tailored application cost an evening. That was the signal: the candidate had read your posting carefully and cared enough to answer it specifically. When a machine does the tailoring in seconds, the signal disappears. Every application arrives polished — your company name used naturally, your requirements mirrored back, the right tone everywhere.
Volume rises too. A pipeline that starts with a few clicks can run every week. The small business that normally receives eight applications for a vacancy may receive forty, all flawless on paper. That is not a reason to panic; it is a reason to change what you look at.
What still tells you something
The good news: the parts of hiring that matter were never really on paper. Four habits keep your process sound:
- Ask for one concrete piece of work. Not “describe your experience”, but “how exactly did you handle this, with which tools, and what came out of it”. AI writes fluent generalities; the details come from the person.
- Set a small, real task. A supervised hour close to the actual job — paid, if it is substantial. Swiss fairness, and the single best filter you have.
- Ask everyone the same core questions. A structured interview is easier to compare and easier to justify than a friendly chat that drifts.
- Phone the references. A Swiss reference letter must be truthful but benevolent; ten minutes on the phone usually say more than the paper does.
One thing not to do
Do not treat “used AI” as cheating. For most office jobs today, using AI well is part of the job. What you want to test is judgment: does the candidate check the machine’s work, notice when it is wrong, and take responsibility for what goes out under their name?
And no AI attends the interview for anyone. How someone answers an unexpected question, what they ask you, how they react when something goes wrong — that part you still have to see for yourself.
“The applications will get better. Your process should improve with them: hire for the work, not for the wrapper.”
