AI applicant screening that screens like you would.
You cannot screen every applicant yourself. You teach notanATS what good looks like in one calibration, and it screens every applicant the way you would. It starts from their resume, asks about the work they have actually done, and checks it against your definition of a good fit. The people who match reach your calendar.
One round, the way it actually runs: a dispatch coordinator role at a growing delivery company, 86 applicants, one hiring manager with a day job.
It starts with the job description you already have. notanATS drafts your definition of a good fit from it, you correct the draft in calibration, and then it checks every applicant’s work experience against the locked definition of a good fit.
Built to match, not to judge. It matches what an applicant has done and can do to what the role and the team actually need.
Illustrative hiring round, not a customer story
Definition of a good fit: draft
Job description
Drop the job description here. Dispatch Coordinator.
Route board under pressure
Drivers directly
Handover the next shift trusts
City routes
notanATS draft lineHas worked in dispatch operations
Manager, correcting itHas dealt with drivers directly, not just through a system
This screening is run by AI. Your answers are used only for this application.
notanATSWalk me through a night the routes changed at the last minute. What did you do first?
MayaA driver called in sick at 6 a.m. I re-sequenced twelve drops and phoned the customers affected.
notanATSWere you speaking to the drivers yourself, or through a system?
MayaMyself, every shift. The drivers called me, I called them.
notanATSWhat did you put in the handover note that morning?
MayaThe new running order, the moved windows, and the pharmacy drop first.
Each applicant has a full chat about their own work, checked against the same locked definition.
Screening results, checked against the locked definition
Maya Chen
Re-sequenced twelve drops at 6 a.m. and left a handover the covering driver ran from.
Met the definition
Maya Chen · Interview booked
Maya ChenBooked · Wed 11:30
Jordan Ellis
Strongest replan answers in the pool.
Advance
Jordan Ellis · Interview booked
Jordan EllisBooked · Tue 10:00
Sam Ortiz
Finished the screening. One dimension needs the manager's own eyes.
Held for review
2 interviews booked1 held for the manager's review · 86 screened in total
Manager's calendar
Tue 10:00
Jordan Ellis · Interview booked
Wed 11:30
Maya Chen · Interview booked
Thu 14:00
Open
The strongest reach the hiring manager's calendar. One case is held for the manager's own review.
Position closed
Inbox · Sam Ortiz
Feedback on your applicationDispatch Coordinator · Position closed
One message for every applicant who finished, sent when the position closed.
Feedback on your application
Hi Sam,
The Dispatch Coordinator position has now closed. Thank you for the time you invested in the application process.
Here’s how you compared across the dimensions we evaluated:
18% of applicants in this pool had a stronger match on Has kept a route board moving when plans break
47% of applicants in this pool had a stronger match on Has dealt with drivers directly, not just through a system
52% of applicants in this pool had a stronger match on Writes a handover the next shift can trust
84% of applicants in this pool had a stronger match on Knows the city routes well enough to replan a run on the spot
Percentages reflect where you stood relative to all applicants in the pool.
We wish you the best in your search.
When the position closes, everyone is informed. Those who finished get where they stood, on each dimension.
The hiring manager adds the job description, corrects the draft, and locks the definition of a good fit. About 20 minutes.
0:00 / 1:10
Every screening ran against your definition of a good fit. The whole process is automated.
Your definition
Screened by your definition of a good fit
No applicant is measured against a stranger’s definition of a good fit, and there is no generic framework behind the screening. Every applicant is met against your definition of a good fit. That definition is what you said you are looking for, in your words, because that is the definition of a good fit this role actually has.
Control
You stay in charge
The dial is yours: it is you deciding how much of the advancing you keep in your own hands. Set it to auto-advance the top X of every N screenings, make the rule stricter or more open, or switch it off and approve every applicant yourself. You can change it at any time. You define a good fit, and you decide who moves.
AUTO-ADVANCE
2of every10
ManualOpen
Closure
Feedback for every applicant
When a round closes, every applicant who finished a screening is told where they stood, dimension by dimension, as the share of the pool that scored higher on each one. Never a score. Never an overall rank.
Because every applicant is matched against your definition of a good fit, each one’s standing in the pool can be shown honestly.
Plenty of job seekers have been ghosted by an employer. Silence after applying is the norm, not the exception. A clear answer is the minimum, and here it is part of the process, not an extra task.
The applicant’s side
A conversation about their experience
notanATS works to understand the applicant’s experience. It asks questions that go deeper into what they have done, and it gives them room to explain with examples from their own work.
The screening is a conversation about their experience, not a form, and not a one-shot answer. That is how an applicant is met against your definition of a good fit: on the work they have actually done, explained in their own words.
And it is always a chat. For this reason, notanATS will always remain in chat mode, and will never put an applicant through a fake audio or video act.
Who it’s for
For the person hiring lands on
If screening sits on the desk of whoever already has a full-time job, whether that is the founder, the CTO, or the one-person HR team, this is built for that desk.
The closing argument
The new form of ATS.
Hiring is supposed to be simple. The manager says: “I need someone to help with this task. This is the type of person I want.” The applicant answers: “I’ve done that. I’d like to apply.”
The old form of ATS manages applications. It captures data points, such as fields, stages, and tags, that are never used again, and it made that simple exchange complex and expensive. The new form holds something else: how you hire. notanATS holds how you hire. It holds your definition of a good fit, learned once in calibration, and puts it in every screening, so each applicant is met from their own resume, with their real work checked against what you are looking for.
The old form manages applications and captures data points. The new form holds how you hire and puts it in every screening.
*You might not understand this unless you have connected your agent to any of your tools.
Pricing
One month free Three months free. Then pay for what gets used.
A simple launch offer, followed by simple postpaid pricing. No prepaid screening packs.
Limited launch offer1 month free 3 months free
All completed screenings included
→
After the offer$49 / month
+ $1 per completed screening, billed after use
A completed screening is one applicant who finishes the screening conversation. Incomplete starts are not billed.
Your own agent can operate notanATS through the API: create jobs, read your applicants and positions, and advance or reject applicants. It uses your existing Settings API key once an admin enables agent access on it. Managers can also create their own personal keys. With your company's own key, those actions happen directly. Candidate decisions from the marketplace (MCP/OAuth) arrive as proposals a manager approves first. See the agent setup steps.
FAQ
Questions about applicant screening, answered plainly
The basics
notanATS is screening infrastructure. It holds the hiring manager's definition of the work, screens applicants against it, and moves the right people to the interview. You set that definition once, in a calibration of about twenty minutes. Every applicant is then screened against the same definition of a good fit.
Applicants who finish the screening receive, when the position closes, where they stood in the pool on each dimension the manager defined. There is no single overall rank.
Recruiters and hiring teams who see AI screening as a boon, not a threat. Basically, anybody who does hiring and needs help screening the applications coming in. You want the screening handled properly so you can get to the interview stage faster, and you want an AI that knows its place: it screens, a person decides. It works for technical roles, operational roles, and leadership roles. The structure does not care about the domain. It cares about specificity. If you want an AI that makes the hiring decision for you, this is not your tool, and it is not going to become one.
No, and it is not trying to. An ATS manages your pipeline: job postings, applicant tracking, offer letters, compliance paperwork. We handle one thing: screening. We sit between applied and interview, and we do that part properly. The screened applicants you advance go into your existing system as your shortlist.
Ironic, given the name. We are aware.
The hiring team defines what matters for the role. Applicants describe their experience against those same things. You see the evidence behind every result: the applicant's own words that supported it. It is not a score handed down from nowhere. It is a record you can read and judge for yourself.
We would rather you see exactly how it fails than believe vaguely that it works.
No. notanATS runs on its own. It holds your definition of a good fit and does the screening itself. When you need the shortlist somewhere else, export it to the tool or workflow you already use.
notanATS goes beyond the resume. The resume is a base understanding. It shows what an applicant says they have done. A conversation built around that history validates it, and matches the person to what the role actually requires.
For managers
About 20 minutes. You have a calibration conversation where the system asks what the role needs, what experience matters, and how you want the work done. It challenges vague answers until the criteria are specific enough to measure. If your answers are vague and generic, it will take longer. You are not filling out a form. You are describing the role the way you would explain it to a smart colleague.
Every role is calibrated on its own. You define the problem the role actually needs solved and the experience that matters, and the assessment then holds every applicant to that role-specific definition. Nothing is scored against a template for the job title. The criteria are yours. How answers are measured against them is not published.
Yes. The full conversation, every message from both sides, is available on the applicant detail page. Alongside it you get the evidence behind each result: which criteria were met and the applicant's own words that supported them.
Yes. Post a real role or a test role and run yourself through the applicant side. We do not police the difference. That is the fastest way to see how the calibration and the scoring work.
The best demo is the product. We do not trust slide decks.
Your criteria lock before the first applicant sees the link, and everyone already screened keeps the score they earned against the definition of a good fit they were told about. What you learn from the interviews sharpens the definition of a good fit you set next time. Re-scoring old answers against new questions is not fairness. It is retrofitting.
The locked definition is the point. Applicant number 12 is not judged by a rule written after applicant number 40.
The conversation is grounded in work the applicant says they have actually done, and the hiring team sees the full transcript alongside the evidence behind every result, so nothing rests on a score by itself. We do not publish how answers are assessed.
We cannot catch every lie. We can make lying harder than telling the truth.
Copy and paste is disabled in the conversation, and detection runs underneath it: pasted or machine-generated input leaves signals, and when we see anomalies we flag them to the hiring team alongside the transcript instead of silently rejecting anyone. How the detection works is not published. Then ask what gaming the screening would actually buy an applicant: an interview, in person, holding claims they cannot back. The screening only decides who gets talked to.
It is a strange thing to cheat your way into.
For applicants
After submitting their application, applicants choose: start the screening immediately, or schedule a slot for a later time. No account creation, no password, and no app download is required either way. Access uses a one-time code and is tied to the applicant's device, so they can pause and come back to finish. If they start now, they go straight into the conversation from the application page.
Yes. The screening is a chat conversation with an AI, and the applicant is told so before it starts. It is chat, and it will stay chat. We will never switch to voice or video screening and try to fake the process. Chat lets applicants explain their experience in their own words, in full, which is the point. It might not be a fit for hiring actors.
The AI does not decide who gets screened in. It measures alignment. The manager decides.
We are not pretending a person is evaluating you. It is an AI comparing your responses against what the hiring manager defined. Speaking introduces errors the AI might misinterpret. Typing lets you review your answers before submitting. There is no transcription to mishear you. Your answers stay your exact words, and your exact words are what gets screened.
Every applicant who completed their conversation but was not hired receives per-dimension feedback when the position closes. For each dimension, they see what percentage of the pool scored higher. That is a sentence worth receiving. It is not a generic rejection email. It is a comparison against the actual pool, delivered once the full picture is available.
If you ask someone for 30 minutes of their time, you owe them more than a form letter.
The conversation has a time budget, and the AI manages it: when time is nearly up, it closes the conversation naturally instead of cutting anyone off mid-sentence. Everything said up to that point is saved and scored. If an applicant booked a slot and did not join in time, the slot expires and they are offered one opportunity to reschedule.
Yes, it works on mobile. That said, you may find a keyboard more comfortable for thirty minutes of detailed answers. You can schedule the conversation for a time and device that suit you.
When a round closes, every applicant who finished a screening is told where they stood, dimension by dimension, as the share of the pool that scored higher on each one. Never a score. Never an overall rank.
AI & fairness
No hiring process can promise the absence of bias, and we will not claim otherwise. What we can tell you is where bias usually gets in and what happens to it here. Bias travels through proxies: names, schools, accents, gaps, and resume polish. The two-sided calibration takes them out of the comparison: the hiring manager sets what the role's outcome requires before seeing a single applicant, and every applicant is scored on work they have actually done against that same definition of a good fit. Selection rates are tracked per pool and built to surface bias if it appears. The AI is used to match, not to judge. It looks for experience that matches what the manager defined for that role, and screens against that definition alone. Organizations can enable voluntary, anonymous self-identification that is kept separate from scoring, never shown to the hiring manager, and has no effect on any result.
We would rather show you the numbers than ask you to trust us.
The system scores for evidence of capability, not writing quality. Spelling, grammar, and phrasing are not evaluated unless the hiring manager explicitly defined written communication as a dimension of the role. What matters is the work an applicant has actually done, not how neatly they describe it. The conversation is designed to surface work evidence, not to grade English.
Data & privacy
We do not claim a certification. The facts are these: names and emails are masked before any AI processing, access to applicant data is audit-logged, and erasure is available on request. The organization running the screening is the data controller, and applicant data is isolated per organization. We do not sell applicant data or use it to train models. For jurisdiction-specific obligations, we recommend confirming with your own counsel.
notanATS is built around the safeguards that matter most in hiring: the hiring manager defines the criteria before any applicant is assessed, every applicant is told an AI is being used before they take part. Advancement runs on the definition the manager locked and the rule the manager set for the role. The cases the system cannot settle are never decided by the system. A failed must-have, or an answer that looks automated, waits for the manager personally. Selection-rate data is available for review. These are practical protections, not a legal guarantee. Regulation varies by country, state, and city and is changing quickly, so we recommend reviewing your specific obligations with your own counsel.
We would rather be honest about what requires legal counsel than give you a blanket guarantee.
Yes. Applicant conversation data is exportable at any time. The calibration model is not exportable.
No. Google Analytics and Microsoft Clarity, described in our privacy policy, run only on this public website. They are never active inside the hiring platform and never see applicant names, resumes, screening responses, or any other applicant data.
Pricing & billing
A screening counts when the applicant completes the full conversation. Unfinished conversations do not count.
How to screen 300 applications without a recruiter
The math of a 300-application pile, and the process that survives it when there is no recruiting team.
screen-300-applications-no-recruiter5 min read
Small companies now get more applications per job posting than any other company size, and it lands on the teams least equipped to handle it: no recruiters, no HR department, just the people doing the work.
Reading all of it takes more time than it looks. At a brisk minute per resume, a few hundred applications is hours of reading. At a careful two minutes, it is a day and a half, and that is before anyone is interviewed, scheduled, or answered. The pile per hire keeps growing, while the team reading them has not.
Accept that you cannot read everything
Nobody reads 300 resumes carefully. What actually happens is skimming, gut feel, and whoever applied on Tuesday getting more attention than whoever applied on Friday. A process built on heroic reading fails quietly, so it is worth designing for the team that exists rather than the one you wish you had.
Define a good fit before you open anything
Write down the must-haves and the knockouts before the first application arrives. Three to five things the role truly requires, and the two or three things that rule someone out. Write them first, because every resume read before that point nudges the definition a little, and the definition of a good fit has to exist before the pile arrives.
One fast pass for knockouts
Go through the pile once, looking only for the knockout criteria. The pass moves quickly because each resume gets one question instead of a full evaluation, and most of the pile never needs a second look.
Screen what is left the same way
Everyone who survives the knockout pass gets the same treatment: the same definition of a good fit for each of them, measured against the definition of a good fit written down at the start. This is where small teams usually lose the thread, because applicant twelve gets a different evaluation than applicant two. The comparison only means something if everyone was measured the same way.
Tell everyone the outcome
Ghosting is common enough that most job seekers have a story about it. Silence feels like saved time and costs you later: those applicants are your future customers, your future applicants for better-fitting roles, and the people who tell their friends what applying to you was like. The minimum is an outcome and roughly where they stood on each dimension, and sending it takes less time than most teams expect.
None of this requires a recruiter. It requires a definition of a good fit written down before the pile arrives, one fast sorting pass, the same definition for everyone left standing, and an answer for every one of them.
Blog post
The resume screening checklist for small teams
A numbered checklist that gets a small team through a resume pile without a recruiter and without the guesswork.
resume-screening-checklist5 min read
Resume screening goes wrong in the same ways every time: no written definition of a good fit, a different definition per applicant, and the loudest resume winning instead of the best fit. A checklist addresses all three, because it makes the process the same every time. Here is the version we would hand a small team.
1. Write down the must-haves first
Three to five things the role truly requires. Not nice-to-haves, not the wishlist from the job board. The things you would reject an otherwise great applicant for lacking. Write them before opening a single application, because everything read first will quietly rewrite them.
2. Write down the knockouts next to them
The two or three things that rule someone out: missing legal requirements, a salary band that does not overlap, a location that will not work. Knockouts are what make the fast pass possible. Without them, every resume demands a full read.
3. One fast pass for knockouts only
Go through the pile once, looking only for the knockout criteria. Do not form opinions or compare applicants. The point is sorting, not judging, and the pass moves quickly because each resume gets exactly one question.
4. The same definition of a good fit for everyone left
Now assess what survived against the must-haves, every applicant against the same list. This is the step small teams skip, and it is why applicant twelve gets a different deal than applicant two. Write a line or two of evidence per must-have so the comparison rests on the notes rather than your memory of them.
5. Shortlist a fixed number, not a feeling
Decide in advance how many move forward: five, eight, ten. A fixed number forces the ranking to mean something. A feeling produces a shortlist that grows every time someone finds one more interesting applicant.
6. Tell every applicant the outcome
The ones who did not make it get an answer and roughly where they stood on each dimension. This is not charity. Ghosting is common enough that an answer alone already stands out.
Blog post
How to reject applicants without ghosting them
Why silence happens, what the minimum viable rejection looks like, and templates you can copy for every stage.
reject-without-ghosting5 min read
Ghosting is common. Most job seekers have been left waiting on silence from an employer. It is not a mystery why. Replying feels like extra work that belongs to nobody, so teams answer the people moving forward and the rest get the void. It is rational, and it is expensive: those applicants are future customers and future applicants, and they remember the silence.
The minimum viable rejection
You owe every applicant two things: an answer, and something they can use. Not a coaching session. Two sentences: the outcome, and roughly where they stood on each dimension. “You did not move forward. 18% of applicants in this pool had a stronger match on [the dimension that mattered most here].” Anything beyond that is generosity.
After the application screen
Keep it short and honest. “Thanks for applying for [role]. We have decided not to move forward. We needed [must-have] for this role, which is where the definition of a good fit landed. Happy to keep your application on file if you are open to it.” Copy it, fill in the two brackets, and send it.
After a screening conversation or interview
An applicant who gave you real time deserves a real reply. “Thanks for the time you put into the screening for [role]. We are not moving forward. You were strong on [specific], but [specific gap] is what this role needs most right now. Happy to keep you in mind for [other type of role] if that is of interest.”
The silver medalist
The one you liked but could not hire. “You were one of our strongest applicants for [role] and this was a close call. We went with someone whose [specific] matched this particular role more closely. I would like to reach out first if a similar role opens. Is that alright?” Only send it if you mean it. A keep-warm note with no follow-up behind it is worse than a clean no.
Three templates, two sentences each, sent the day the decision is made. The companies people reapply to do this. The ones people warn their friends about do not.
Blog post
What automated applicant screening actually does
A plain-English explainer: what it replaces, what it never replaces, and how to evaluate vendors without the demo hype.
what-is-automated-screening5 min read
Automated applicant screening has a naming problem. It sounds like a robot deciding who gets hired. What it actually does is narrower and more useful: it replaces the resume pile triage, the part where someone skims 300 applications and picks twelve on gut feel.
What it replaces
Manual review. Application review eats hours of a hiring manager’s week for every vacancy, inside a larger pile of hiring admin. Automated screening takes the review step. Everyone gets the same definition of a good fit, and the conversation starts from each applicant’s own resume, so no two conversations are alike. The evidence is attached to the result.
What it never replaces
The interview and the final call. Screening answers one question: who is worth talking to. It does not answer who to hire. A tool that claims to make the hiring decision is not one worth buying. The hiring manager decides, and the system exists to make sure the decision rests on a real comparison instead of a skimmed pile.
How to evaluate vendors
Ask four questions. First, who defines what good looks like: you, in your own words, or their fixed template. If it is their template, every customer hires the same person. Second, can you control how much the system acts on its own: does it advance applicants automatically, and can you turn that off. Third, does every applicant hear back, or only the winners. Fourth, can you see the evidence behind a ranking, or do you get a score with no explanation.
Those four answers tell you most of what you need to know. A vendor that hedges on any of them is selling a black box with a friendly demo.
Blog post
How to cut time-to-hire when you don’t have an HR team
Screening and scheduling are the two bottlenecks. Fix those and time-to-hire drops without hiring a recruiter.
cut-time-to-hire-no-hr5 min read
Time-to-hire has two bottlenecks and neither is the interview itself. The first is screening: hours of application review per vacancy, and that is just the reviewing. The second is scheduling: the week of calendar tag between deciding who to interview and the interview happening. Shorten those two and the whole timeline shortens with them.
Define a good fit once
Every hiring delay that looks like screening is actually re-deciding what the role needs, one resume at a time. Write the must-haves and knockouts down before applications open. One conversation, up front, replaces dozens of micro-decisions later. It is also the fastest part of the whole process, which makes it the step teams most often skip, and they pay for the skip for weeks.
Screen everyone the same way
The pile grows while you deliberate, and small companies get the largest piles of all. If each applicant gets a slightly different evaluation, you cannot compare them, so you re-read, you second-guess, and days pass. One definition of a good fit, applied the same way to everyone, produces comparable results, and decisions get faster when there is less to argue about.
Keep interview slots real
The most avoidable delay in hiring is advancing applicants you cannot interview. Only move people forward when there is actual interview capacity behind the decision, so offers go out the same day instead of after a week of finding a slot. Without that capacity behind it, a shortlist is a list of names, not a pipeline.
None of this requires an HR team. It requires a written definition of a good fit, the same definition applied the same way, and interview slots that exist before you need them. With those three in place, the bottleneck moves somewhere you can see it.
Blog post
Your agent should run your hiring, not just read it
Asking an assistant about your hiring data was the first step. The next one is letting your own agent act on it, through the API, inside boundaries you set.
talk-to-your-hiring-data5 min read
A quiet shift is happening in how people use their software. Instead of opening five dashboards and exporting spreadsheets, they connect an AI assistant to the system that holds the data and ask questions in plain language. Which stage loses the most applicants. How long screening takes on average. Whether this quarter’s pools are stronger than last quarter’s. The assistant reads the data and answers.
Why this beats dashboards
A dashboard answers the questions someone predicted you would ask, months ago, when it was built. Real questions do not arrive on schedule. They arrive mid-meeting, when a hiring manager asks why the pipeline stalled, and the honest answer is that the chart for that does not exist. An assistant connected to your data answers the question you actually have, in the shape you asked it, and the follow-up question too.
From asking questions to running the work
Asking questions is the smaller part of the shift. The larger part will be letting your own agent act: setting up a role and advancing the applicants that meet your definition of a good fit, through the API instead of clicking through our UI. Your own agent can operate notanATS through the API: create jobs, read your applicants and positions, and advance or reject applicants. It uses your existing Settings API key once an admin enables agent access on it. Managers can also create their own personal keys. With your company's own key, those actions happen directly. Candidate decisions from the marketplace (MCP/OAuth) arrive as proposals a manager approves first.
The boundary is enforced where the data lives
An agent connected to your hiring data needs clear rails, so the rails sit at the data layer, not in the agent’s manners. Keys are org-scoped: an agent can operate notanATS through the API — create jobs, read your applicants and positions, and advance or reject applicants — inside your own organisation, and nothing outside it. If a key leaks or an agent misbehaves, you revoke the key and the access is gone. That boundary is what makes a connected agent safe to use.
What teams hand over first
Nobody starts by giving an agent the keys to everything. The pattern we expect starts with the morning brief: where the pipeline stands and what moved overnight. Then screening review, with the agent summarising who met the definition of a good fit and why. Acting on the review (opening roles, advancing applicants) is write access, and it is live: your own agent can operate notanATS through the API once an admin enables agent access on your Settings API key.