An always-on recruiting agent that finds candidates, scores each application against your job description, and conducts structured video or voice screening interviews around the clock. Resume ranking is $0.10 per resume and AI screening interviews are $5 each, on top of an applicant tracking system that stays free.
An AI recruiter is software that performs the repeatable parts of recruiting: it searches for candidates, scores every application against the requirements of a specific role, runs a structured first-round interview, and returns a ranked shortlist with written reasoning. HireCade's AI recruiter ranks resumes at $0.10 each and runs screening interviews at $5 each inside a free applicant tracking system.
The difference from an applicant tracking system is that an AI recruiter forms an opinion. An applicant tracking system stores candidates and moves them through stages; it leaves the reading, comparing, and scheduling to you. An AI recruiter reads every application against a rubric written for that role, interviews the strongest applicants, and tells you what it found and why.
The difference from older keyword screening is that it evaluates answers rather than documents. String matching taught candidates to paste the job posting into their resume and taught employers nothing about ability. Here the candidate is asked questions, the follow-ups adapt to what they say, and the score cites the evidence, so an unusual background described in unusual words still scores on its merits.
It is deliberately not the whole process. Cade replaces the reading and the scheduling, which is where most recruiting hours vanish without producing judgement. It does not decide who gets hired, it does not auto-reject anyone, and it does not close a candidate who has a competing offer. Those stay with your team, and the ranked list with its transcripts and flags is what you use to do them well.
A single open engineering role can attract several hundred applications in a week. Nobody reads several hundred applications. In practice a recruiter reads the first fifty carefully, skims the next hundred for familiar company names, and the rest are rejected by a timer rather than by a decision. The candidate you needed may well have been in the part nobody reached.
Time studies of manual recruiting put roughly 13 hours into sourcing one role, 20.85 hours into screening 250 resumes, 4.35 hours into scheduling, and 18.5 hours into reviewing first-round interviews. That is about 56.7 hours per role, and almost none of it is judgement work. It is reading, comparing, and calendar wrangling.
Cade takes that layer. It searches for candidates, scores every application against the requirements you actually wrote down, invites the strongest to a structured screening interview that runs whenever they are free, and hands your team a ranked shortlist with transcripts and reasoning attached. The review work that remains is closer to 3.75 hours, because it is only the top of the list.
Cade builds a search from your job description and works continuously rather than in the gaps between meetings, so pipeline keeps arriving overnight and at weekends.
Every resume is scored against the requirements of that specific role and returned with a written explanation of the score, so you can audit a decision instead of trusting a number.
A structured first-round interview conducted by the AI, with questions adapted to the candidate's background, available in video or voice depending on the role.
Screening sessions are proctored, with signals for pasted answers, model-generated phrasing, and off-screen assistance flagged on the report rather than hidden.
Candidates take the screen when it suits them, in their own timezone, which removes the scheduling loop that usually adds days between application and first conversation.
Results land in your pipeline as a ranked list with scores per competency, a full transcript, and the recording, so a hiring manager can check the evidence in minutes.
Screening is not a chatbot asking generic questions. Before a role goes live, the job description is turned into a scorecard: the competencies that matter, what a weak, adequate, and strong answer looks like for each, and which of them are disqualifying if missing. Every candidate for that role is then asked against the same scorecard, which is what makes the scores comparable at all.
During the interview, follow-up questions adapt to what the candidate says. Someone who claims ownership of a migration gets asked what broke and how they found it. The evaluation weighs communication, problem-solving approach, behavioural signals, and role-specific skill rather than resume formatting or keyword density, so a well-formatted resume with nothing behind it does not clear the bar.
Integrity matters more now that candidates have the same models you do. Sessions are proctored and the transcript is checked for answers pasted mid-sentence, phrasing that reads as generated rather than spoken, long unexplained pauses before fluent responses, and inconsistency between claimed and demonstrated depth. Flags are reported as flags, not verdicts. A human decides what to do with them, because the honest position is that detection is probabilistic and treating it as proof would fail good candidates.
What comes back on each candidate
Remote and entry-level postings where volume makes fair manual review impossible.
A founder who cannot spend two working weeks per role reading applications.
Support, sales, operations, and annual graduate intakes with identical criteria.
Score the applicants you never got to, rather than reposting the job.
Candidates screen themselves overnight instead of waiting on a calendar invite.
When you need a record of why each candidate was advanced or rejected.
Parallel pipelines without the review quality sliding on whichever role is third in line.
Push reading and scheduling to software so recruiters spend their day closing candidates.
Rank a client's inbound applications consistently before a consultant spends time on them.
These three cover the same stage of hiring with very different economics and failure modes. Most teams end up using more than one, so the useful question is which part of the funnel each should own.
| Consideration | AI Recruiter | In-house recruiter | Recruiting agency |
|---|---|---|---|
| Cost shape | Usage-based: $0.10 per resume ranked, $5 per screening interview, no subscription | Fixed salary and benefits regardless of how many roles are open that month | Contingent fee on placement, typically a large percentage of first-year salary |
| Throughput | Hundreds of resumes and interviews in parallel, with no queue forming | Realistically a few roles at once before quality of review drops | Limited by how many candidates their consultants can personally work |
| Consistency across candidates | Identical scorecard and question set for every applicant to a role | Varies with fatigue, time of day, and how many other roles are live | Varies by consultant, and their incentive is placement rather than calibration |
| Time to shortlist | Hours, because sourcing, ranking, and screening run continuously | One to three weeks once scheduling and review are included | Days to weeks, depending on whether they already know the market |
| Auditability of a decision | Every score cites quoted evidence, with transcript and recording on the record | Notes if someone wrote them, recollection if they did not | A written summary shaped by the incentive to place the candidate |
| Candidate experience | Everyone gets interviewed and gets feedback, on their own schedule, but it is not a human conversation | Warm and persuasive when there is time, silent when there is not | Strong for the shortlist they are pitching, thin for everyone else |
| Selling the role to a candidate | Weak. Software does not talk anyone out of a competing offer | Strong. This is the part worth protecting a recruiter's time for | Strong, and often their main value on senior searches |
| What happens when you stop paying | Nothing is owed, and the candidate records, scores, and transcripts stay in your free pipeline | The knowledge and relationships leave with the person | The pipeline usually leaves with the agency when the contract ends |
| Where it is weakest | Judgement calls, nuance, negotiation, and any signal that is not in the answer given | Volume. Fair review of 400 applications is not a human-scale task | Cost per hire, and a pipeline that leaves when the contract ends |
| Best used for | Sourcing, ranking, and first-round screening on every role | Closing, senior searches, stakeholder management, and hiring strategy | Confidential or executive searches, and markets you have no network in |
Cade is deliberately not the whole process. It replaces the reading and scheduling, not the final human loop or the conversation that gets someone to sign.
You pay for the AI work that costs us money to run, per resume and per interview. The pipeline it runs inside is free with no seat count and no contract.
Applicant tracking system
The pipeline Cade works in, with no cap that turns into a sales call.
AI resume ranking
Each application scored against the role, with the reasoning attached.
AI interview screening
A structured video or voice first round with a full report per candidate.
Full-service recruiting
When you would rather hand the search over than run it yourself.
There is no subscription, no seat licence, and no minimum spend. A role with 300 applicants and 20 screening interviews costs $130 in total.
Write or paste the job description. Cade turns the requirements into a scorecard for that role.
Candidates are found and every application is scored against the scorecard, at $0.10 per resume.
Strong applicants self-schedule a structured AI interview, at $5 each, in their own timezone.
A ranked list with scores, transcripts, and flags arrives in your pipeline for the human loop.
An AI recruiter is software that performs the repeatable parts of a recruiter's job: searching for candidates who match a role, evaluating applications against defined criteria, conducting a structured first-round interview, and producing a scored, ranked shortlist. It differs from an applicant tracking system, which stores and organises candidates but does not form an opinion about any of them.
The evaluation is what separates it from keyword filtering. Older screening tools matched strings, which taught candidates to stuff resumes with the words in the posting and taught nobody anything about ability. An AI recruiter asks questions, reads the answers, and scores against a rubric written for that role, which means a candidate with unusual background wording can still demonstrate the skill.
It is also worth separating an AI recruiter from an AI sourcing tool and from an assessment platform, because the three are often sold under the same label. A sourcing tool finds people and stops there, leaving the evaluation to you. An assessment platform tests a narrow skill, usually with a timed exercise, and tells you nothing about the rest of the candidate. An AI recruiter spans the whole top of the funnel, from search through to a scored interview, and hands over one ranked list instead of three exports you have to reconcile.
It does not replace your final human loop, and we would not sell it as though it does. Screening establishes that someone is worth your team's time. It cannot judge whether they will work well with the two people they will sit next to, it cannot read the room in a negotiation, and it cannot talk a hesitant candidate into accepting. Those are the parts of hiring where a person is genuinely better, which is why we deliberately hand the shortlist back rather than making the decision.
It also cannot see what is not in the interview. A candidate having a bad morning scores worse than they should, and a well-prepared candidate with shallow experience can present better than the transcript suggests. The mitigation is that every candidate is screened rather than only the first fifty, the reasoning is written down so you can disagree with it, and a person still makes the call. Used that way it widens the funnel and shortens the process. Used as an automatic reject button it will lose you good people.
One practical note for anyone comparing vendors: ask what the score cites. A product that returns a match percentage with no evidence is asking you to take a hiring decision on faith, and you will have nothing to show a candidate, a hiring manager, or a regulator who asks why someone was rejected. Evidence you can read is the difference between a tool you can defend and one you can only hope about.
An AI recruiter is only useful if you can see what it did and in what order. The pipeline below is the same for every role, which is what makes two candidates for the same opening comparable. Nothing in it is hidden behind a single opaque match percentage.
It starts with the job description. The requirements you wrote are turned into a scorecard for that role: the competencies that genuinely matter, the level expected at each one, and which requirements are disqualifying if they are missing. A vague job description produces a vague scorecard, so this is the stage worth ten minutes of your attention. If the posting says "strong communication skills" and nothing else, the scorecard has nothing concrete to test.
Sourcing then runs continuously from that scorecard rather than in the gaps between your meetings, so candidates keep arriving overnight and at weekends. Every application, whether it came from sourcing or from your careers page, is scored against the same requirements at $0.10 per resume, and each score arrives with a written explanation you can read rather than a number you have to trust.
Candidates who clear the bar are invited to a structured screening interview at $5 each. They book it themselves, in their own timezone, whenever they are free, which removes the scheduling round trip that normally adds days between application and first conversation. The interview runs in video or voice depending on the role, is proctored, and adapts its follow-up questions to what the candidate actually says.
What lands back in your pipeline is a ranked list. Each candidate carries a score per competency, the evidence behind each score, the full transcript, the recording, and any integrity flags raised during the session. Your team reads the top of the list, disagrees with anything that looks wrong, and runs the human rounds from there.
Scoring is per competency, not one blended number. A candidate can be strong on problem-solving approach, adequate on the role-specific skill, and weak on communication, and the report will say exactly that. A single overall figure would hide the trade-off you actually need to make, and it would be impossible to argue with.
Each competency is scored against the rubric written for that role, which describes what a weak, adequate, and strong answer looks like. Because the rubric exists before any candidate is interviewed, the fortieth applicant is measured against the same standard as the first. That is the mechanical reason the scores are comparable at all: consistency comes from the rubric, not from the model being clever.
The evaluation reads what the candidate said, not how their resume was formatted. Older screening tools matched keywords, which taught candidates to paste the job posting into their resume in white text and taught employers nothing about ability. Here, someone who describes the same skill in unusual words still gets credit for it, because the follow-up questions probe the substance. Claiming ownership of a migration invites a question about what broke and how they found it, and a shallow answer scores as a shallow answer.
Every score comes with the evidence that produced it, quoted from the answer. This matters more than the score itself. A hiring manager who thinks a candidate was marked down unfairly can open the transcript, read the exchange, and overrule the judgement in a minute. That is the intended use. A score you cannot interrogate is a score you should not act on, which is why we return the reasoning rather than a confidence percentage.
Integrity signals sit alongside the scores rather than inside them. Answers pasted mid-sentence, phrasing that reads as generated rather than spoken, long unexplained pauses before unusually fluent responses, and claimed depth that collapses under a follow-up are all reported as flags. They are never converted into an automatic rejection, because detection is probabilistic and a system that auto-rejected on suspicion would fail nervous candidates and non-native speakers first.
The division is deliberate and we would rather state it plainly than let you discover it later. Cade decides what order candidates appear in and what evidence to surface. It does not decide who gets hired, and it does not send rejections on your behalf as a silent automatic filter. Every advance or reject decision is made by a person looking at the report.
The reason is not caution for its own sake. Screening establishes that someone is worth your team's time, which is a narrower claim than it sounds. It cannot judge whether a candidate will work well with the two people they will sit next to, it cannot read the room in a negotiation, and it cannot talk a hesitant candidate out of a competing offer. Those are the parts of hiring where a person is genuinely better, and they are the parts worth protecting a recruiter's hours for.
There is also a limit on what any interview can see. A candidate having a bad morning scores worse than they should. A well-prepared candidate with shallow experience can present better than the transcript deserves. The mitigation is not a smarter model, it is process: every candidate is screened rather than only the first fifty, the reasoning is written down so you can disagree with it, and a human makes the call.
In practice the teams who get the most out of this treat the ranked list as a reading order rather than a verdict. They start at the top, they check two or three candidates from further down as a sanity check on the scorecard, and when they disagree with a score they fix the scorecard rather than the individual result. That last habit is what makes the second role easier than the first.
The honest framing is that structured screening removes some sources of bias and cannot remove all of them. Anyone who tells you their model is neutral is describing a marketing position rather than a technical property. What we can describe are the controls that are actually in the product and the ones you have to supply yourself.
The structural advantage is consistency. Every applicant to a role is asked against the same rubric, at the same difficulty, with the same competencies weighted the same way. Manual screening cannot do this. A human reviewer is stricter at five in the evening than at nine in the morning, is more forgiving after a good interview than after a bad one, and reads the first fifty resumes far more carefully than the next two hundred. Removing that drift is a real fairness gain, not a marginal one.
The second control is what is being measured. The evaluation weighs demonstrated skill, problem-solving approach, and communication rather than pedigree, company names, resume formatting, or keyword density. A candidate with an unusual background, a career break, or a non-standard way of describing their work is not penalised for failing to write like the job posting. This is also why we score interview answers rather than only resumes: an answer is a much better sample of ability than a document someone optimised.
The third control is auditability, and it is the one you should lean on hardest. Because every score cites the evidence behind it, a pattern of unfair outcomes is visible rather than buried. You can read why a candidate was ranked low, compare two candidates who were scored differently, and see whether the rubric or the model produced the gap. A system that returned only a number would give you nothing to review.
What you have to supply is the scorecard and the review. A scorecard that encodes a requirement the job does not really have will filter people out consistently and unfairly, and the system will do that faithfully. Review your hiring outcomes, check who is being screened out and why, and correct the rubric when it is wrong. Fairness here is a process you run, not a feature you buy.
The AI recruiter runs natively inside the HireCade applicant tracking system rather than as a separate tool you have to reconcile. Scores, written reasoning, transcripts, recordings, and integrity flags land on the candidate record next to your team's own notes, so a hiring manager reads one page instead of cross-referencing two products. That applicant tracking system is free forever, with unlimited job postings, unlimited candidates, unlimited team seats, and a branded careers page.
Unlimited seats is the part that changes behaviour. When interviewers and hiring managers cost nothing to add, everyone who needs to read a transcript can read it, and the debrief stops being a game of telephone through whoever holds the licence. Most tools price by seat precisely because that is where the leverage is, which is why we do not.
If you already run another applicant tracking system, you do not have to migrate to try this. The common pattern is to export your open roles and active candidates, run HireCade in parallel for a few weeks on one or two roles, and compare. There is no subscription, no seat count to negotiate, and no minimum spend, so running both while you decide costs only the per-resume and per-interview usage. A role with 300 applicants and 20 screening interviews comes to $130 in total.
On candidate data, the useful thing to know is where it lives and who can see it. Interview recordings, transcripts, scores, and flags are stored against the candidate record in your pipeline and are visible to the team members you have added. Candidates know they are being interviewed by software and recorded, because that is part of how the round is set up. Our handling of personal data is described in our privacy policy, and if your legal team needs specifics for a particular jurisdiction, ask us before you run the first role rather than after.
The questions below are worth asking any AI screening vendor, including us. A vendor who cannot answer them in writing is a vendor whose process you cannot defend if a candidate or a regulator asks about it.
Selling this into situations where it does not help is a short-term win and a long-term problem, so here is where it earns nothing. If a role has thirty applicants, you do not need it. You can read thirty applications properly in an afternoon, and the ranking adds a step without removing work. The value of automated screening scales with volume, and below roughly a hundred applications the curve is flat.
If the role is senior enough that the hiring decision turns on judgement, scope, and stakeholder history rather than demonstrable skill, a screening interview is the wrong instrument. A staff engineer or a VP is assessed on architectural trade-offs, influence, and what they chose not to build, and that assessment needs an experienced human in the room. Use the applicant tracking system to organise the search, and put real interviewers on the loop.
If the position is filled mostly through referrals and warm introductions, the top of your funnel is not the bottleneck and automating it will not speed anything up. The same is true for confidential or executive searches, where the constraint is discretion and network rather than throughput.
If your requirements are genuinely not written down, fix that before automating anything. A scorecard generated from a vague job description will screen consistently against the wrong thing, which is worse than screening inconsistently against the right thing, because it is systematic. Ten minutes spent making the job description specific is the highest-return work in this whole process.
And if what you actually need is for someone to persuade a reluctant candidate to take your offer, no screening product solves that. That is a human conversation, and it is the reason we hand the shortlist back rather than pretending the software closes anybody.
An AI recruiter is software that performs the repeatable parts of a recruiter's job: searching for candidates, scoring every application against defined criteria, running a structured first-round interview, and producing a ranked shortlist with the reasoning attached. It forms an opinion about candidates, which is what separates it from an applicant tracking system.
HireCade's AI recruiter ranks resumes at $0.10 each and runs screening interviews at $5 each, inside an applicant tracking system that is free forever. It does not make hiring decisions. It produces the shortlist and the evidence, and your team decides who advances.
AI resume ranking is $0.10 per resume and AI interview screening is $5 per interview. There is no subscription, no per-seat licence, and no minimum commitment, so a quiet month costs nothing.
The applicant tracking system underneath is free forever with unlimited jobs, candidates, and team seats. If you want the whole search handled instead, full-service recruiting is a 10% retainer plus a 20% placement fee against first-year base salary, and contract placements are 30% of engagement fees.
An applicant tracking system stores and organises candidates; an AI recruiter evaluates them. The first is a filing cabinet with a workflow attached, which is genuinely useful but leaves the reading, comparing, and scheduling to you. The second reads every application against the requirements of the role, runs the first interview, and hands back a ranked list.
You need both, which is why HireCade includes the applicant tracking system for free and charges only for the AI work. The scores, transcripts, recordings, and flags sit on the candidate record inside the pipeline, so there is no second tool to reconcile.
Accuracy depends on how well the scorecard describes the role, which is why we build one per role rather than using a generic template. Every score comes with written reasoning quoted from the candidate's answer, so you can check the judgement instead of trusting a number, and we would rather you overrule it than defer to it.
On bias, the structural advantage is consistency: every applicant is asked against the same rubric, and the evaluation weighs demonstrated skill and communication rather than pedigree, resume formatting, or keyword density. That removes several sources of drift in manual screening. It does not make the system neutral by decree, so scores are auditable, hiring outcomes are reviewable, and a human makes every advance or reject decision.
No. It removes sourcing, resume review, and scheduling, which is where most recruiting hours disappear without producing much judgement. Recruiters who use it spend their time on closing candidates, senior searches, and working with hiring managers instead of reading the four hundredth application of the week.
The final loop stays human. Cade produces the shortlist and the evidence behind it; your team decides who to hire. If you want experienced engineers to run the deep technical rounds on your behalf, that is a separate service and you can read about it on our interview as a service page.
Yes. Candidates know they are speaking to software, and the session is proctored and recorded as part of how the round is set up. We do not think a screening interview should pretend to be something it is not, and candidates behave differently when the format is clear, which makes the signal better as well as fairer.
What they get in exchange is a fair hearing and a fast answer instead of an application that vanishes. Everyone who reaches the screening stage is interviewed, on their own schedule and in their own timezone, and receives feedback regardless of the outcome.
Every applicant gets an interview rather than silence, which is the opposite of what happens when a human team is overwhelmed by volume. Candidates take the screen when it suits them and in their timezone, questions adapt to their actual background, and they receive feedback regardless of the outcome.
It is not a substitute for talking to a person, and we do not pretend otherwise. Candidates know they are speaking to software. What they get in exchange is a fair hearing and a fast answer, instead of an application that vanishes.
Sessions are proctored, and the transcript is analysed for answers pasted mid-sentence, phrasing that reads as generated rather than spoken, long silences followed by unusually fluent responses, and claimed experience that does not survive a follow-up question.
Those signals are reported as flags for a human to weigh, never as automatic rejections. Detection is probabilistic, and a system that auto-rejected on suspicion would fail nervous candidates and non-native speakers. The follow-up questioning is the real defence, because it is hard to fake depth about a system you did not build.
The scorecard is generated from the job description you write, so the most direct lever is the job description itself. Specific, written-down requirements produce a specific scorecard and sharper questions; a posting that only asks for "strong communication skills" gives the rubric nothing concrete to test.
Questions are not pulled from a fixed bank. They adapt to the candidate's background during the interview, which is why two candidates for the same role face equivalent competencies rather than identical wording. If a score looks wrong to your hiring manager, the fix is usually to sharpen the requirement in the scorecard rather than to override individual results one at a time.
It runs natively inside the HireCade applicant tracking system, so scores, transcripts, recordings, and flags sit on the candidate record next to your team's notes rather than in a separate tool. That pipeline is free, including unlimited seats for interviewers and hiring managers.
If you are on another ATS, most teams export their open roles and active candidates and run HireCade in parallel for a few weeks before switching. There is nothing to sign and no seat count to negotiate, so running both while you decide costs you only the per-resume and per-interview usage.
Transcripts, recordings, scores, and integrity flags are stored against the candidate record in your pipeline, visible to the team members you have added to it. Because seats are unlimited, every interviewer and hiring manager who needs to read the evidence can read it without a licence conversation.
How we handle personal data more broadly is set out in our privacy policy. If your legal or security team needs specifics for a particular jurisdiction, ask us before you run the first role rather than after, and we will answer in writing.
The scorecard is generated from your job description, so it is not limited to engineering. Teams use it across support, sales, operations, marketing, healthcare, and technical roles, and the question set changes with the role rather than being pulled from a fixed bank.
It works best where volume is high and the criteria are written down. For a role with 30 applicants you probably do not need it. At 400 it is the difference between reviewing everyone and reviewing whoever applied first.
They solve different problems. An agency brings a network, works a specific brief, and is paid a contingent fee on placement, which makes it strong for confidential searches, executive roles, and markets where you have no reach. An AI recruiter is usage-priced, runs on every role at once, and applies the same scorecard to every applicant, which makes it strong wherever volume is the constraint.
The economics differ sharply. Agency fees are a large percentage of first-year salary per hire; ranking and screening cost cents and dollars per candidate. Many teams use both: automated screening on the roles with real applicant volume, and an agency or our own full-service recruiting for the searches that need a person working the market.
For self-serve usage there is no placement fee to refund, because you paid cents per resume and $5 per screen rather than a percentage of salary. What we do instead is more useful: the transcript and scores from the original screen are still on the record, so you can look back at what the process actually said about that person and adjust the scorecard for the next round of hiring.
For full-service recruiting, where a placement fee applies, replacement terms are agreed in writing before the search begins. We would rather set that expectation up front than argue about it after a hire has left.
Screening is one stage of a hiring process. These pages cover the stages either side of it: the free pipeline the AI recruiter runs inside, the human interview rounds that follow a shortlist, and what it takes to actually employ the person once you have chosen them.
The pipeline the AI recruiter works inside: unlimited jobs, candidates, and team seats, at no cost.
When the shortlist needs a deep technical round, senior engineers run the loop and send back scored reports.
A free curated shortlist every Monday, for the roles where the problem is too few applicants rather than too many.
Every published rate in one place, including resume ranking, screening interviews, and full-service recruiting.
The full product line, from sourcing and screening through to employment, immigration, and device management.
How to run an engineering search end to end, and where automated screening fits into it.
Recruiting participants and specialists for research studies, where volume screening is the whole job.
Employ the person you just screened in a country where you have no entity, at $499 per employee per month.
Engage genuinely independent contractors compliantly, with classification liability handled for you.
Visa and relocation support when the strongest candidate needs the right to work where you are hiring.
Check whether the offer you are about to make is competitive before you start screening for it.
Level definitions that make a scorecard specific, so the rubric tests the seniority you actually need.
The questions candidates practise against, useful when you are writing your own competency rubric.
What the other side of the screen looks like, and what candidates prepare for before they meet Cade.
Post a role, let Cade rank the applications and run the first round, and spend your time on the shortlist instead of the inbox.
