Recruitment automation from application to pre-hire, with every AI score checked by a person and every decision explained
See one case, screen by screen ↓Built forRecruitersHiring managersAssessors and panel membersHeads of talent acquisitionHR operationsResponsible AI leadsHR systems administrators
Volume hiring runs on separate tools, so scores come back as bare numbers and seats are still empty on the start date.
Run across separate tools
Built-in controls
Shortlists in days, not weeks, and every score checked by a person, with the reason on record for each decision.
Recruitment Automation is an end-to-end recruitment solution: agents read every resume with its sources, apply each requisition's published eligibility rules, propose written and one-way video interview scores with their evidence, book panels and create the pre-hire in the HR system. Recruiters decide close calls, assessors verify every score, and a second person approves every selection.
How to choose AI solutions for back-office and enterprise work →
Updated
A volume campaign brings hundreds of applications to each requisition. Resumes are skimmed in seconds or filtered by keyword, tests and one-way video interviews come back from separate vendors as single numbers, and case studies are marked by hand. When a hiring manager asks why an applicant was turned down, the answer is in someone's head, if anywhere.
The slow stages are invisible until a seat is still empty on the start date: panels are booked by email, approvals travel as messages, and the accepted hire is typed again into the HR system. Pass rates by group surface at the yearly bias audit, long after the decisions they describe, while NYC Local Law 144, the Illinois AI Video Interview Act and the Colorado AI Act now expect notice, evidence and a person in the loop.
Estimated split for one requisition with a few hundred applicants, run across separate tools and by hand.
A Data Analyst requisition with 287 applicants and four seats, in a campaign of six requisitions and 1,512 applicants, at a company with 40,000 people in 30 countries. Applications closed on October 31; the start date is January 5.
A worked example on sample data. Names, IDs and figures on the screens and in this section come from the demo.
The recruiter opens six requisitions as six tracks, from applied to hired, each with a forecast of when its seats fill against the start date. Two are behind, and the agents say why: only 55% of the eligible Cybersecurity Analyst applicants have started the assessment. Data Analyst fills all four seats by December 4, 32 days early.
“Data Analyst fills all 4 seats by Dec 4, 32 days before its start date.”
She asks which requisitions will miss their start date. The answer names both, with the reason and its source, and offers the next step: a reminder to the 54 applicants who have not started the assessment.
The resume reader reads 14 new resumes and converts five grading systems to one 4.0 scale. The eligibility screener applies rules R1 to R5: 11 eligible, 2 not eligible on field of study, each with a respectful message. One resume holds hidden text telling the AI to rank the applicant first; the integrity guard sends it to the recruiter instead of obeying it.
“1 resume has hidden text: sent to you, not rejected.”
On an applicant's record the resume sits beside what was read from it: degree, course of study, school, academic result, experience and skills, each with a confidence. Clicking a confidence highlights its line; a GPA of 3.82 is also shown on the common 4.0 scale.
The assessor verifies a case study without seeing the applicant's name. For each rubric criterion the assessment marker shows its proposed score, its confidence and the passages behind it. It is unsure about writing for the audience, at 64%, so the assessor moves it from 3 to 4 and types the reason before the score counts.
“2 or more points from the AI: a second marker will be assigned.”
Each of four one-way answers is transcribed and scored from the transcript, with the quote the score rests on. No face, voice-tone, accent or background analysis; the assessor verifies all four.
Six finalists have every score. Ranked by composite (40% written, 30% video, 30% panel) with what the evidence shows for each, the hiring manager proposes the top three to fill the three open seats, with a reason.
The head of talent approves the selection with the reason beside it; the person who proposed it could not. Offer terms inside the approved band and a new requisition wait in the same place.
The applicant reads and accepts the offer on the careers site. The HR-system connector creates the pre-hire, links the position and starts the hire process, by API or SFTP flat file, with nothing typed twice.
Recruiters, assessors, hiring managers and the head of talent: how a campaign changes for each of them.
Ten agents read every resume, apply the published rules, propose every written and video score with its evidence and forecast each seat; recruiters decide close calls, assessors verify every score, and a second person approves every selection.
Reads every resume into degree, course of study, school, academic result, experience and skills, each with its source line and a confidence, and converts each grading system to one 4.0 scale.
Compares what a reader sees with what the file holds: hidden or off-page text, duplicates and repeat applications.
Applies each requisition's published eligibility rules the same way to everyone and names the rule on every exclusion.
Proposes a score for each rubric criterion of a written answer, with the passages that justify it and its confidence.
Transcribes one-way video answers and proposes competency scores from the words, with the quote behind each.
Matches panel calendars with applicant availability and books live interviews with video links.
Forecasts the date each requisition's seats are filled from the pipeline and stage conversion, against the start date.
Creates the pre-hire record in the HR system of record, links the position, attaches the signed offer and starts the hire process.
Decide; a second person approves.
Ask in plain words, or say what to prepare. Each answer cites the record it came from and offers one next step; the decision stays with you.
Which requisitions will miss their start date?
Two are behind: Cybersecurity Analyst fills 19 days after its start, because only 55% of eligible applicants have started the assessment, and Technology Counsel 11 days after, with 75 applications not yet screened.
Show the hidden-text resumes
Two Data Analyst resumes hold text a reader cannot see. One has 46 words in white 1-point type telling the AI to rank the applicant first; it was read as text and ignored. Both are waiting for a recruiter, not rejected.
How reliable is the scoring?
The AI and the verifying assessor agree within one point on 91% of criteria; weighted kappa is 0.78; 7% of scripts went to a second marker.
From the working solution, on its sample data. Pick one to see it large.
An administrator sets these in the settings, as new versions; nothing is coded per customer.
| Setting | Default | Choose from |
|---|---|---|
| Close-call decision clock | 2 working days | Any number of working days; escalates to the recruiting lead |
| Scoring clock | 3 working days | Any number of working days |
| AI confidence to flag | 0.70 | 0.50–0.95; below it a criterion is marked for judgment |
| Second marker | Gap of 2 points | 1–3 points |
| Sample check of exclusions | 10% | 5–50% a week |
| Stage weights | 40% written · 30% video · 30% panel | Set per requisition |
| Applications not hired | Kept 12 months after close | Per retention schedule; talent pool by opt-in |
Each figure carries its label: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = design goal, measured in the live solution · “typical” = published figure · “estimated” = our estimate · Sources: NEOGOV, public-sector time to hire · 29 CFR 1607.4(D) (four-fifths rule) · NYC Local Law 144 · Illinois AI Video Interview Act (820 ILCS 42) · Colorado AI Act (SB 24-205) · People named on this page are characters in the working solution.
Recruitment Automation is an end-to-end recruitment solution: agents read every resume with its sources, apply each requisition's published eligibility rules, propose written and one-way video interview scores with their evidence, book panels and create the pre-hire in the HR system. Recruiters decide close calls, assessors verify every score, and a second person approves every selection.
AI resume screening ranks one role's applicants against locked requirements. Recruitment Automation runs the whole campaign around it: published eligibility rules, written assessments, one-way video interviews, verified scores, panel booking, approvals and the hand-off to the HR system, with a fill forecast for every requisition.
No. The eligibility screener applies the rules you publish and names the rule on every exclusion; close matches, unrecognized grading systems, low-confidence fields and hidden text go to a recruiter. A recruiter re-checks a sample of exclusions each week and can overturn any with a reason.
From the transcript only, with the quote behind each score. There is no face, voice-tone, accent or background analysis; applicants are told before recording and can ask for written questions or a live interview instead. An assessor verifies every score.
Yes. The working solution runs six requisitions with 1,512 applicants, including a graduate program with 612, and reads the peak at a median of 11 seconds a resume. Bulk moves, blind review and booking panels from calendars are built in.
Every score shows its evidence and is verified by an assessor; changes need a reason and a wide gap brings a second marker. Selection rates are checked each week for every requisition and stage, and model or prompt changes need the Responsible AI lead's approval with a regression attached.
Your applicant tracking system, the HR system of record (by API or SFTP flat file), calendars, video meetings, email, the careers site and job boards. Open requisitions and their applicants can be imported from a flat file at go-live.
The Agentic Solution Engine builds and deploys it from your requirements, rules and documents, and it goes live once every quality gate has passed, usually starting with one campaign.
Every finding tied to quoted evidence and tested for gaps before anyone signs.
1-min demoForm I-9 Compliance86% less fine exposure on the sample audit, every correction signed and never back-dated.
1-min demoBackground Check AdjudicationEvery record weighed the way the law requires, and cleared candidates start on time.
1-min demoWe'll run Recruitment Automation on one of your own closed requisitions, with its applications, tests and video interviews, and show where the time went.