Recruiting Agent
AI-powered candidate screening and resume analysis.
Meet the agent
This app ships with a production AI agent — permission-aware, tenant-scoped, audit-logged, and governed by the Agent Development Lifecycle.
AI Recruiting
Pipeline, screening and interviews through offer drafts.
Overview
Recruiting Agent is the AI layer on top of the candidates and jobs your team already manages in Job Board and Talent Hub — it keeps no pipeline of its own. Each new external application is parsed and scored within minutes while Auto-Screen is on (internal applicants when someone clicks Screen), against the job on four weighted dimensions — technical skills, experience, education and, when you write Company Culture & Values text, cultural fit — with the strengths, gaps and deal-breakers written out so a recruiter can see why. The AI is told to ignore protected characteristics and stand-ins for them. Rankings put a job's screened applicants in fit-score order, and the Approvals queue holds the autonomous agent's proposals until someone with approval rights decides, on desktop or from a phone: phone-screen invites and rejections always wait there, and acknowledgments and recruiter notices do too until the agent earns more trust.
It drafts personalized outreach in the brand voice you choose, books interviews and notifies the interviewers and the candidate, drafts offer letter text you can use when creating the offer in Offer Manager, and answers pipeline questions inside Ask AI through a purpose-built recruiting agent that asks before it moves a candidate, books an interview or sends anything. Autonomous mode stays off until an admin turns it on; once on, a scan every 15 minutes can propose acknowledgments for new applications, phone-screen invites for high scorers, rejections for low scorers and notices to a recruiter about candidates waiting on a decision — each behind its own switch. The Dashboard shows the agent's recent activity, AI Operations → Activity lists what it proposed and who decided, and a Playground lets admins run each tool against real candidates before trusting it. Licensed per recruiter seat and opt-in per tenant.
See how each new applicant is screened and scored within minutes of applying — with ranked shortlists, drafted outreach and offer letter text for review, job-related scoring weights, identifier masking, and a record of what the agent proposed and who decided.
Highlights
Capabilities
AI Candidate Screening & Scoring
-
AI resume parsing from PDF, DOCX, and legacy .doc files (with OCR for scanned PDFs)
-
Single-pass parse, score, and reasoning in one AI call
-
Fit scores across technical, experience, education, and cultural fit (scored only when you write Company Culture & Values text)
-
Configurable scoring weights that must total 100
-
Company culture and custom screening instructions applied to every candidate
-
Strengths, gaps, and deal-breaker call-outs per candidate
-
Force re-screen to clear the cache and re-parse a resume
-
Rankings of a job's screened candidates by fit score
Conversational Recruiting Assistant (Ask AI)
-
Natural-language recruiting chat in the Ask AI assistant
-
Automatic tool selection across the recruiting tools each person's role allows
-
Instant pipeline-status answers without a full AI round-trip
-
Filter candidates by job, stage, or application date in chat
-
Pipeline summary with stage counts across active jobs
-
Market compensation benchmarks for any role, level, and location
-
Write actions for recruiting staff (advance stage, fix contact info) with a diff preview + confirm step
Candidate Outreach & Email
-
AI-drafted acknowledgment, rejection, phone-screen, and interview-invite emails
-
Passive / cold-sourcing outreach with strength highlights
-
Outreach saved as a draft for review by default
-
Admin switch that keeps Ask AI outreach as drafts for review (on by default)
Interview Coordination
-
Schedule phone, video, onsite, and panel interviews
-
Edit the panel, length, meeting link and place of a booked interview
-
Reschedule and cancel with an audit trail; correct a recorded outcome
-
Candidate and interviewer notifications when an interview is booked, moved, edited or cancelled
-
Booked interviews on interviewers' Google and Outlook calendars
-
Candidate self-scheduling: leave the time blank and the candidate is sent a link to pick from the times the interviewers' calendars show free
Autonomous Recruiting (opt-in)
-
Scheduled pipeline scan every 15 minutes that surfaces work automatically
-
Application acknowledgments that send on their own once the agent reaches the Standard trust level
-
Phone-screen invite proposals for candidates the AI recommends advancing, always approved by a person
-
Rejection email proposals for candidates the AI recommends declining, always approved by a person
-
Recruiter notices when screened candidates are waiting for their next decision
-
Progressive trust levels that earn autonomy over time
Approvals & Audit
-
Screening scored on four factors you weight (technical fit, experience, education, and fit with your Company Culture & Values text when you write one)
-
Screening AI told to ignore protected characteristics and stand-ins for them, including any part of your own instructions that asks otherwise
-
Resumes and cover letters treated as untrusted data (prompt-injection defense)
-
SSNs and card, bank-account, license, and passport numbers a recruiter types are masked before the agent's model call
-
Approval queue for the autonomous agent's acknowledgments, phone-screen invites, rejections, and recruiter notices
-
Candidate PII and scores limited to people with candidate-data access (managers and above plus recruiting staff by default; configurable)
-
Recent agent activity on the Dashboard, and every proposal with who decided in AI Operations → Activity
Limits & Specs
-
Recruiting tools available to the assistant: 22 (read, action, and write; each person is offered the ones their role allows)
-
Resume file size limit: 10 MB
-
Supported resume formats: PDF, DOCX, DOC (+ OCR for scanned PDFs)
-
Autonomous pipeline scan cadence: Every 15 minutes
-
High-score recruiter alert: When a screening recommends Advance (at or above your Auto-Advance Threshold, 70 by default)
-
Screening job retries: 5 attempts, each waiting longer than the last
-
Default daily autonomous action limit: 50 (admin-configurable)
-
Pipeline stages: Applied → Screening → Qualified → Potentially Qualified → Interviews In Process → Interviewed → Ready for Offer → Offered → Hired
-
Pricing: License required (opt-in per tenant)
Screenshots
Use cases
Resources
FAQ
It parses the resume and scores the candidate against the job in a single AI call, rating technical skills, experience, education, and cultural fit (cultural fit only against Company Culture & Values text you have written), then combining them with weights you configure (they must total 100). Each screening returns strengths, gaps, deal-breakers, and a confidence level worked out from the score. The AI is told to ignore protected characteristics and stand-ins for them. You can add company culture notes and custom screening instructions that apply to every candidate.
Not from its own proposals: everything the autonomous agent proposes waits in Approvals at the starting Probation trust level, and phone-screen invites and rejections always do. In Ask AI, the agent asks the person to confirm before it moves a candidate, updates contact info (with a diff preview), books an interview, or sends an email — and by default outreach is only saved as a draft for review. Offer letters are drafted as text only; the offer itself is created and approved in Offer Manager.
Autonomous mode is off until an admin turns it on. When enabled, a scan every 15 minutes can propose acknowledgments for new applications, phone-screen invites for candidates the AI recommends advancing, rejections for candidates it recommends declining, and notices to recruiters about candidates waiting on a decision — each behind its own switch. Invites and rejections always wait for a person's approval; acknowledgments and recruiter notices wait too until the agent earns the Standard trust level. A daily action limit (default 50) caps activity, every action is logged, and you can pause the agent at any time.
PDF, DOCX, and legacy .doc files up to 10 MB. Text is read directly from PDF and DOCX files (a .doc file is read by the AI provider), and scanned PDFs fall back to OCR so they can still be parsed and scored.
Each screening is scored on four factors you weight — technical fit, experience, education, and cultural fit against your Company Culture & Values text (left out when you have not written any). SSNs and card, bank-account, license, and passport numbers a recruiter types are masked before the agent's model call, resumes and cover letters are treated as untrusted data (so hidden instructions in them can't hijack the model), and candidate names, emails, and scores are shown only to people with candidate-data access — by default managers and above plus recruiting staff (admins, HR admins, and recruiting app admins), configurable in Talent Hub settings.
The agent works on your existing recruiting pipeline — the jobs and applicants managed in Job Board and Talent Hub. It reads and acts on that same candidate data rather than maintaining a separate database of its own.
Job Board and Talent Hub are where you post jobs, collect applications, and manage the pipeline. The Recruiting Agent is the AI layer on top of them: it screens and scores candidates, ranks them, drafts outreach and offer letter text, coordinates interviews, and can propose follow-ups autonomously — all against the same candidate records. Use the apps to manage the pipeline; add the agent to automate the work inside it.
Each person is offered only the tools their recruiting role allows. Viewing candidates, scores, pipeline counts, and upcoming interviews is for recruiting admins, a job posting's recruiters and hiring managers, and the people your candidate-data setting admits (managers and above plus recruiting staff by default, configurable in Talent Hub settings). Screening, ranking, scheduling, and drafting outreach or offer letters are for recruiting admins, a job posting's recruiters, and the people that setting admits. The candidate's own Job Board permissions still apply to stage moves and contact-info changes.
It requires a license and is opt-in per tenant — an administrator enables it for your business from the Apps Marketplace, and it's disabled by default until then. Contact your OneMango administrator about enabling it.
