Why we do not scale the front door

A convincing resume, a portfolio, a reference-shaped LinkedIn history, even a live video interview, none of it is hard to fabricate any more. Detecting that after the fact is a race, and it is one staffing is structurally set up to lose: whoever is faking it gets to try again with better tools every time a detector catches up. That is not a reason to build a better detector. It is a reason to ask why detection was ever the plan.

Why detection loses

Every fraud-detection tool runs downstream of a front door that was left open on purpose, usually in the name of volume: post the role widely, take everyone who applies, sort the pile afterwards.

Sorting a pile of thousands necessarily means trusting signals, a resume, a headshot, a calendar invite, that are exactly the signals now cheapest to fake. The fix is not a sharper filter at the bottom of that funnel. It is fewer people entering it, chosen deliberately.

Why this profession specifically cannot afford to get it wrong

Most professions with a real fraud problem have a backstop: a licensing board, an accredited exam, a register someone can actually check.

AI ethics does not have one yet, that makes the profession a softer target than most for exactly this kind of fabrication, and the irony is not lost on us: the one role whose entire value is a trustworthy human check on an AI system is precisely the role a bad actor would most want to fake their way into, and currently one of the most exposed, which is exactly why we built evidencing around it.

What a curated desk actually buys you

This is the real reason we run a small, curated desk instead of an open board, not only the ranking and leaderboard argument we make elsewhere.

Deliberate, low-volume sourcing means a person, not a script, has an actual relationship with a candidate before anything is put in front of a client, not a form submission processed at scale. Trust gets established early, through that relationship.

It also means our evidencing has somewhere to stop. If something about who a candidate says they are, their identity, their claimed history, does not hold up under ordinary professional scrutiny, the evidencing process stops there and we say so, the same honesty we already apply to competence gaps, extended to authenticity itself.

What this is not

This is not a claim that fraud is solved, or that any process, ours included, is immune to a sufficiently determined and resourced attempt. Nobody honest can promise that. What a smaller, relationship-first funnel actually does is remove the easiest version of the attack, the one built for volume, run against a process that was never built to handle volume in the first place.

Trust, at this point, is not something you get to assume from a CV and a calendar invite. It has to be engineered, on purpose, and kept up, deliberately, for as long as the role exists.