Every hiring tool I have ever bought has been priced wrong. Per seat, per year, with a minimum spend. The contract said one number, the invoice the next year said another. The seat I paid for half the time sat empty, because the manager who needed it last quarter had moved on.
The first thing we did when we started Picked was throw the pricing model away. Not the price. The model. We do not sell seats. We sell one thing: a fully vetted candidate, ready for the hiring manager to read. The price is 99 cents per vetted candidate. The first 50 are free. There is no minimum, no annual contract, no per-seat tax.
Ninety-nine cents per vetted candidate is what we can profitably charge once Claude does the reasoning, LiveKit handles the voice infra, and Whisper does the transcription. Voice, reasoning, scoring, and the candidate-facing experience compose into a unit cost we can pin. The margin is thinner than a seat-licence SaaS would post, and we are fine with that. We would rather price what the customer values than what our finance team can defend in a board pack.
It also forces us to be good at the only thing that matters: the quality of the vetted candidate. If you pay per seat, the software vendor gets paid whether or not you hire anyone. If you pay per vetted candidate, we only get paid if the screen, the assessment, and the first-round interview actually run. That alignment is the whole point.
A typical software-engineer role on Picked gets 40 to 60 applicants in the first week. About 25 of those clear the inbound triage and run through the AI-vetted pipeline. At 99 cents per vetted candidate, that is $25 of spend for a shortlist of three. Pin that against the alternatives:
Twenty-five dollars vs five-figure cheques is the headline. The structural difference is bigger. When the variable cost is this low, the constraint is no longer "is it worth posting another role". The constraint becomes "do I have a real role to post". Which is the right constraint.
Three things, openly.
One, we will never beat an enterprise vendor on procurement-side feature checkboxes at V1. A buyer who needs SSO, SCIM, SAML, BYOK, and a master service agreement signed in legalese should wait for our V2 enterprise tier. We are not pretending otherwise.
Two, we will never be the cheapest at scale of one. A 5-role-per-year company will spend less than a hundred dollars with us per year. That is fine. We earn back the customer relationship when the next role hits and the manager remembers that posting was eight minutes and the shortlist arrived on Friday.
Three, we cannot bundle. A seat-licence vendor can throw in a feature module and call it value. We can only ship better vetted candidates. The pricing model holds us to it.
The 99-cent unit is a bet that the right pricing primitive for AI-native hiring is the candidate, not the seat. Seats made sense when software was a tool you opened to do work. Picked does the work. You read the output. The thing you are buying is the output.
If we are wrong, the model breaks because the unit cost rises faster than we can hold the price. If we are right, the model gets better every quarter as Claude, LiveKit, and Whisper improve under us. Right now, six months from V1, the unit is healthy and the bet is on.
Founder of Neuroworx. Eight years building psychometrics into hiring. Writes about the unit economics, the candidate side of the funnel, and what shipping with Claude looks like in practice.