Use case

Recruiting screening automation that never decides who gets hired

Automating recruiting screening means removing the paperwork and logistics that surround hiring, not the decision itself. Calfy does not build systems that reject candidates automatically, score people against a model of who got hired before — which only encodes whoever a company happened to hire in the past — or infer anything about a candidate's protected characteristics. Who gets an interview and who gets an offer stays a human call, full stop; what a system can responsibly take off a recruiter's plate is everything around that call.

· Reviewed by Artur Horimoto, Founder & CEO

The line this work doesn't cross

Several jurisdictions now regulate automated employment decision tools specifically, and that isn't a footnote added after a system is built — it's the starting constraint the architecture is designed around. Anything that ranks, scores, or filters candidates against a target profile risks reproducing whatever bias already exists in who a company has hired before, quietly and at scale, which is exactly the failure mode those rules exist to catch. So a screening build here is scoped from the first conversation to leave ranking and rejection out of it entirely, and to make every automated step visible enough that it could be explained to a candidate, a regulator, or your own legal team if it's ever questioned.

That's a narrower job than "AI that hires for you." It's also a real one — most of what slows a recruiting team down has nothing to do with judgment and everything to do with logistics.

The job to be done

Ask a recruiter where their week actually goes and very little of it is deciding whether someone is a strong candidate — that judgment takes minutes once a recruiter is looking at a well-organised application. Most of the week disappears into getting to that point: opening résumés in a dozen different formats, cross-checking a certification or a right-to-work document by hand, sending the same acknowledgement email one at a time, chasing a reference who hasn't replied, and finding a shared interview slot across three calendars. None of that requires judgment. All of it requires time, and time is the one thing a stretched recruiting team doesn't have enough of.

The cost of skipping it isn't abstract. An applicant who hears nothing back for weeks — which is the norm at more organisations than would like to admit it — forms an opinion about the company before a recruiter has even opened their application, and that opinion travels: to review sites, to their network, to the next role they apply for somewhere else. Acknowledging every applicant promptly is one of the cheapest reputational wins available to a hiring team, and also one of the most commonly skipped, simply because doing it by hand at volume doesn't scale. An AI agent built for this job can pick up nearly all of the logistics above without ever touching who gets hired.

The manual way vs. the automated way

Handled manually, screening looks like this: applications land in an inbox or an applicant tracking system in whatever format each candidate happened to submit, a recruiter opens them one at a time in whatever order they arrived, checks the basics against the job requirements from memory, and replies — if there's time — with a generic acknowledgement days or weeks later. Interview scheduling happens by email, back and forth, until a slot sticks. References get chased manually, if at all, and easily fall through the cracks when the recruiter is juggling six other requisitions.

The automated version handles the same steps, immediately, without touching who gets hired:

Manual Automated
Application format Every résumé different; recruiter re-reads structure each time Parsed into one consistent structure so like is compared with like
Requirement checks Recruiter checks certifications, right to work, availability from memory Checked against stated, objective requirements and surfaced, not decided
Acknowledgement Days or weeks, if it happens at all Sent the moment an application is received
Interview scheduling Email back-and-forth until a slot sticks Offered directly against real calendar availability
References and documents Chased manually, easy to lose track of Requested and tracked automatically, with reminders
The hiring decision A person's call Still, always, a person's call

That last row is the one that matters most. Nothing above it changes who gets hired — it changes how much of a recruiter's week gets spent on logistics instead of on the candidates in front of them.

How a build actually works

Parsing applications into one consistent structure

A résumé arrives as a PDF, a LinkedIn export, a plain-text paste into a form — every candidate formats their own experience differently, and a recruiter reading several in a row is doing translation work before they ever get to judgment. The system's first job is normalising that: pulling the same fields out of every application regardless of the format it arrived in, so a recruiter is comparing like with like instead of re-orienting for every résumé.

Checking stated, objective requirements — and surfacing, not deciding

Some job requirements are simply factual: does the candidate hold the certification the role requires, do they have the right to work in the location the role is based, are they available for the shift pattern the role needs. A system can check facts like these against what the candidate submitted and flag a gap — a missing certification, unclear availability — for a recruiter to look at. It does not translate that flag into a rejection. The output is always "here's what's missing, handed to a person" — never "this candidate is out."

Acknowledging every applicant, immediately

The moment an application is received, the candidate hears something back — not a decision, just confirmation that a human being will actually look at what they sent, and roughly when. This alone fixes one of the most common failure points in hiring: the silence that makes a candidate assume they've been ignored, whether or not that's true.

Scheduling interviews and chasing what's outstanding

Once a recruiter decides to move a candidate forward, the logistics of getting them into a room, or onto a call, can run against real calendar availability instead of an email thread — the same underlying pattern as automating appointment scheduling anywhere else in the business. The same applies to references and documents: a request goes out, a reminder follows if nothing comes back, and a recruiter sees the status without having to chase it themselves.

Keeping candidates informed through the process

A candidate who applied weeks ago and hasn't heard anything assumes the worst, reasonably. The system keeps every applicant updated on where they stand — received, under review, moving to interview, not moving forward this time — without a recruiter having to write that update by hand for every person in the pipeline.

Auditability, because you may have to justify it

Every step the system takes is logged: what was checked, what was flagged, what was sent, and when. That log exists because a recruiting process sometimes has to be explained after the fact — to a candidate who asks why they didn't hear back, to legal counsel, to a regulator. A system that can't show its work isn't one we build, and that standard applies to the screening logistics just as much as it would to anything closer to a hiring decision.

The honest upside: pipelines can widen, not just narrow

It's worth saying plainly: automation in hiring has a poor reputation for a reason, mostly earned by keyword-matching résumé filters that silently discard strong candidates who phrased their experience differently than the filter expected. Built the way described above, the effect can run the other way. A system that reads every application with the same consistent structure — rather than however tired a recruiter is by the fortieth résumé of the day — can surface a candidate a manual pass would have missed entirely, simply because their résumé didn't use the expected wording. Whether a system narrows a pipeline or widens it is a design choice, not something inevitable about automation, and it's worth asking about directly before any build starts.

What it connects to

A recruiting screening build is only useful if it reaches into the systems a hiring team already runs:

  • Your applicant tracking system or careers site, wherever applications actually arrive.
  • Your calendar, so interview scheduling offers real availability rather than a guess.
  • Email or your messaging tool, for acknowledgements, status updates, and reference requests.
  • Document storage, for certifications, right-to-work documents, and references as they come in.
  • Wherever your recruiters actually work, so a flagged gap or a scheduled interview shows up in their existing workflow rather than a separate dashboard nobody opens.

None of this requires replacing what your team already runs. This sits squarely inside workflow automation — reducing repetitive administrative load, not building a decision-making model — and once a candidate is hired, the same approach carries into employee onboarding for the next stretch of paperwork.

Frequently asked questions

Does the system ever reject a candidate on its own?

No. It checks stated, objective requirements — a certification, right to work, availability — and flags what's missing for a recruiter to review. It never converts a check into a rejection, and it never scores or ranks candidates against each other. Every decision about who moves forward is made by a person.

Is automated recruiting screening even legal to use?

Several jurisdictions now regulate automated employment decision tools directly, which is exactly why this is scoped narrowly: parsing, checking objective facts, scheduling, and candidate communication, with no ranking or scoring of people. Every step is logged and explainable. Bring your specific location and hiring volume to the first call so we can talk through what applies to you.

Will this replace our recruiters?

No. It removes the logistics — parsing, acknowledgement, scheduling, chasing documents — so recruiters spend their time reading candidates and making calls, not reformatting résumés or writing the same email dozens of times over. The human in the loop isn't a fallback here; it's the design.

Does it work with our existing applicant tracking system?

In most cases, yes. We connect directly to applicant tracking systems and careers platforms that expose an API, and build a workable route in for the ones that don't. Bring the list of what your team runs to the first call.

What does it cost?

It depends on how many systems it touches and how many steps of the process it covers — a build that just handles parsing and acknowledgement sits at the lower end; one spanning scheduling, document chasing, and candidate updates sits higher. Every engagement gets clear pricing agreed before any build work starts.

Bring us the volume of applications that's currently sitting unopened or unanswered, and in a free 30-minute strategy call we'll tell you honestly what a screening build could take off your recruiters' plate — and what has to stay a human decision regardless.

Automate this job

Walk us through how it works today. We will map the build and give you a clear price before anything starts.

Free 30 minutes. No pitch deck. You leave with a plan either way.