How it works
One definition of the work. Every applicant screened against it.
Here is how a position runs, in order — your side first, then the applicant's, and your agent's where you use one.
The manager's side
Hiring manager
Create the position
Add the job description. notanATS reads it and drafts a screening model: four or five dimensions of the role, each with the work it involves, what good looks like, and what capability looks like. Lines in the draft are grounded in the job description itself. The draft may include a line the description implies rather than states — and any such line is marked as inferred, so you can cut it in calibration.
Hiring manager
Calibrate, then lock
A short chat walks you through the draft, one dimension at a time. Confirm what is right, correct what is not, cut what does not matter. Mark the must-haves — the dimensions an applicant cannot fail. Then the model is presented back to you in full, and you lock it.
Locking is checked on the server, not taken on trust: a model with an empty dimension cannot lock. Later, your real interview notes can amend the model into a new version. Amendments apply to future applicants only — nobody is ever rescored against a definition that changed after they were screened.
Hiring manager
Open the position
The position goes live with a public application page. If you already run an ATS, notanATS can sit in front of it. The Workable integration is live: applicants flow in, and the ranked shortlist flows back with scores and transcripts attached. Greenhouse, Lever, Ashby and SmartRecruiters are coming soon.
Hiring manager
Review the pool
Scored applicants collect in one pool view: a score out of 100, the rating on every dimension, and flags where a must-have was missed. Advance people in batches or one at a time. Advancing is a person's decision — unless you turn on automatic advancement, which moves the top of the pool in batches. Flagged applications are skipped by automatic advancement and wait for a person either way.
Hiring manager
Interview
Interview times are offered and booked inside notanATS. Then the technology stops: the interview is a real conversation with your team. Record the outcome, and when the position closes, applicant feedback goes out automatically.
The applicant's side
Applicant
01Apply and book
The applicant applies on the position's page and adds a resume. If the resume cannot be read, the application still goes through. They pick a screening time — or start right away — and receive a private link. There is no account and no password; the link is checked against their inbox the first time it opens, and it only works inside the booked window.
Applicant
02The screening chat
A typed conversation, up to an hour, about the work in the role. The applicant has already seen what the work is and what capability looks like — those parts of the model are shown to them up front. The questions probe real experience, including the anti-patterns: the things that look right on paper but fail in practice.
Applicant
03An answer at the end
When the position closes, the applicant can see where they stood in the pool on each dimension, where the organisation's transparency settings allow it.
How scoring works
Evidence, matched against your rubric.
When a screening chat ends, the applicant's answers are taken as evidence and matched, dimension by dimension, against the private rubric from calibration. Each dimension is rated on the evidence in the conversation — not on the resume's keywords, and not on polish.
Dimension ratings combine by the weights set at calibration into a score out of 100. A must-have dimension rated as a miss does not end the application quietly: it sets a flag the manager sees, and automatic advancement skips flagged applications while leaving them in the pool for a manual decision.
The rubric stays private — by construction
The part of the model that says what good looks like is never given to the screening chat. The rubric is excluded from what the screening side can access at all, in the structure of the code.