What is agentic recruiting? AI agents in hiring, how they work and the rules that apply
On this page
- How does an AI agent in recruiting work?
- How is an agent different from an AI assistant or automation?
- What tasks do recruiting agents take on?
- Which products are sold as recruiting agents?
- What do hiring laws require when an agent screens or ranks candidates?
- What should human review of an agent look like?
- How do you evaluate an AI recruiting agent?
- What are the pros and cons of agentic recruiting?
- Related terms
- Questions people ask
Agentic recruiting is the use of AI agents in hiring: software that takes a goal from a recruiter, such as "find ten qualified people for this role", and carries out several steps toward it on its own, including searching, ranking, drafting or sending outreach, and booking screening calls, with a person approving at set checkpoints. An agent differs from an AI assistant, which answers one prompt at a time, and from rules-based automation, which only follows triggers someone configured. As of October 2026, LinkedIn, Metaview, Gem and hireEZ all sell products they describe as recruiting agents, and existing hiring laws apply to what those agents decide, whatever they are called.
For the wider picture of AI in hiring, see what is AI in recruiting; for software that conducts the interview itself, see what is an AI interviewer.
How does an AI agent in recruiting work?
Most recruiting agents run the same loop:
- Goal and context. The recruiter gives a brief: the job description, intake notes, examples of good and bad fits, and rules such as "don't contact current clients' employees".
- Plan. The agent breaks the goal into steps, such as search, rank, draft and send.
- Act with tools. It uses the systems it has been connected to: a profile database, the ATS, the recruiter's mailbox, a calendar.
- Check results and adjust. Replies, rejections and the recruiter's feedback change what it does next.
- Hand off. At the checkpoints the team set, a person approves, edits or stops the work.
Metaview's own FAQ puts the distinction simply: "Most AI only responds when you prompt it. Agentic AI can take action on its own, based on instructions you've already given." Its example is a sourcing agent that starts searching after an intake call without anyone pressing "go" (metaview.ai, as of October 2026).
How is an agent different from an AI assistant or automation?
| Rules-based automation | AI assistant | AI agent | |
|---|---|---|---|
| Who starts each step | A trigger a person configured | A person, every time | The agent, within limits |
| What it does | One fixed action: send this email when the stage changes | Drafts, summarizes or answers one request | Several steps toward a goal, using tools |
| Example | Auto-rejection email when a knockout question fails | "Write an outreach message for this profile" | Searches, ranks, drafts and queues outreach for a role |
| Typical failure | The rule was wrong, every time | A bad draft that a person catches | Many actions taken before anyone notices a wrong assumption |
The last row is the point. An agent's value is that it acts without being asked each time; its risk is the same.
What tasks do recruiting agents take on?
| Task | What the agent does | Risk | Sensible human checkpoint |
|---|---|---|---|
| Sourcing | Searches profile databases and past applicants for matches | Moderate: who is never found is invisible | Review the search criteria and a sample of results |
| Outreach | Drafts and, in some products, sends personalized messages | Moderate: reputation, deliverability, inaccurate claims | Approve templates and the first messages; see recruiting email deliverability |
| Application review | Ranks or scores applicants against criteria | High: a selection decision | A person reviews every "do not advance" |
| Pre-screening | Asks candidates questions by chat, message or voice | High: candidates are assessed by software | Disclosure, human review of outcomes |
| Scheduling | Books, moves and confirms interviews | Low | Spot checks |
| Notes and summaries | Writes up calls and fills scorecards | Moderate: invented or distorted details | The interviewer checks before sharing |
Which products are sold as recruiting agents?
What each vendor says on its own site, as of October 2026. We have not tested these products; the descriptions are theirs, not ours.
| Product | What the vendor says the agent does | Human control the vendor describes |
|---|---|---|
| LinkedIn Hiring Assistant | An add-on to LinkedIn Recruiter that sources, evaluates applicants "from LinkedIn and your ATS" (integration availability varies by partner), drafts personalized messages and pre-screens interested candidates. LinkedIn lists voice prescreening and interview scheduling for "Hiring Assistant 2" as coming soon. | The FAQ says recruiters choose which projects it helps with and "remain in control throughout the process, from intake to interview" |
| Metaview | Calls itself "the Agentic Recruiting Platform", with agents across sourcing, application review, screening, scheduling and interview notes | "Define what the agent can and can't do"; each recommendation includes candidate-level rationale, and the team can review, edit or override |
| Gem | An AI Sourcing Agent, an AI App Review Agent that "ranks and reviews applicants" with reasoning for every score, and an AI Fraud Detection Agent, usable on top of an existing ATS | Reasoning shown for every score |
| hireEZ | "EZ Agent sources, reviews, screens and schedules on top of your ATS" | "Nothing moves forward without your approval"; "nothing gets rejected without you" |
| Paradox | A conversational assistant, Olivia, that automates applicant screening, interview scheduling and other tasks by text and chat, including for Workday customers | Not described on the home page |
Not every AI recruiting tool is an agent. An AI note taker such as Interview Signal, which we make, assists the interviewer during a call, suggesting follow-up questions and drafting a scorecard for the interviewer to check, but it does not contact candidates or move anyone forward.
What do hiring laws require when an agent screens or ranks candidates?
No law regulates "agents" as such. The rules attach to what the software does to a hiring decision, and they put the duties on the employer or agency using it. As of October 2026:
| Where | What applies | Duty that affects agents |
|---|---|---|
| United States, federal | Title VII, the ADEA and the ADA apply to any selection procedure | The employer answers for discriminatory outcomes, whoever or whatever made the selection; see EEOC guidance on AI |
| New York City | Local Law 144 on automated employment decision tools | Independent bias audit within the past year, published summary, notices to candidates; see the bias audit guide |
| Illinois | Public Act 103-0804, in effect since January 1, 2026 | No AI use with discriminatory effect; notice that AI is used; see Illinois HB 3773 |
| Colorado | SB 26-189, from January 1, 2027 | Notice before use, disclosure after an adverse outcome, correction of data and human reconsideration; see Colorado's rules for employers |
| California | Civil Rights Council automated-decision system regulations (since October 1, 2025) and the CCPA ADMT rules (from January 1, 2027) | Discrimination rules and four-year record retention; for ADMT, pre-use notice, opt-out unless an exception applies, and explanations; see CCPA ADMT rules |
| European Union | AI Act, Annex III point 4(a): recruitment and selection systems are high-risk | Deployer duties including human oversight, monitoring, logs and informing people, applying from 2 December 2027; see the EU AI Act for recruiting |
One detail matters for agents in particular: California's ADMT rules define automated decisionmaking technology as technology that replaces or substantially replaces human decisionmaking, so a genuine human review step changes which duties apply. A checkpoint where someone clicks "approve all" is not genuine review. This is general information, not legal advice; the state-by-state summary is in AI hiring laws by state.
What should human review of an agent look like?
- Review every negative decision. An agent may rank and recommend; a person decides who is not advanced, and records why.
- Sample the positives. Each week, compare a sample of the agent's top picks with a manual review of the same applicants.
- Check outcomes by group. Selection rates by sex and race or ethnicity, using the four-fifths rule as a first screen.
- Approve what goes out under your name. Templates first, then the first batch of real messages.
- Tell candidates. Say where AI is used and how to reach a person; the AI disclosure templates have wording.
- Keep the log. What the agent did, what a person approved, and when.
How do you evaluate an AI recruiting agent?
- Which actions can it take without approval, and which can be undone? A sent email cannot.
- Which systems does it read and write: the ATS, your mailbox, your calendar? With whose permissions?
- Does it explain each ranking or score in terms of the job's criteria?
- Has it been tested for different outcomes by group, when, and can you see the results?
- Is candidate data used to train models, and can that be switched off?
- Can you set limits per role, such as off-limits companies or a cap on daily outreach?
- Is there an audit log you can export?
- How is it priced: per seat, per role, per action?
Pilot it on one or two roles with a recruiter doing the same work in parallel, and compare the shortlists, not just the hours saved. The broader evaluation questions are in how to use AI in recruiting responsibly.
What are the pros and cons of agentic recruiting?
| Pros | Cons |
|---|---|
| Work continues between recruiter sessions: searches, first-pass review, scheduling | Mistakes repeat at scale before anyone notices |
| Recruiters spend more time talking to people | Ranking and screening are regulated selection decisions |
| Consistent first-pass criteria across many applicants | Consistently wrong if the criteria or training data are wrong |
| Faster responses to candidates | Candidates may not realize they are dealing with software unless told |
Related terms
- AI assistant: responds to one request at a time; the person drives every step.
- Human in the loop: a design where a person must approve defined actions before they take effect.
- AEDT and ADMT: the legal terms New York City and California use for automated tools that substantially assist or replace hiring decisions.
- AI interviewer: software that conducts the interview itself; see what is an AI interviewer.
- Interview intelligence: tools that record, transcribe and analyze interviews; see what is interview intelligence.
More definitions are in the AI recruiting glossary.
Questions people ask
What is agentic AI in recruiting?
Agentic AI in recruiting is software that takes a goal from a recruiter, such as finding people for a role, and carries out several steps toward it on its own: searching, ranking, drafting or sending messages, and booking calls, with a person approving at set points. It differs from an AI assistant, which answers one prompt at a time, and from rules-based automation, which only follows fixed triggers someone configured.
What are examples of AI recruiting agents?
As of October 2026, LinkedIn sells Hiring Assistant as an add-on to LinkedIn Recruiter for sourcing, applicant evaluation, outreach drafts and pre-screening; Metaview describes agents for sourcing, application review, screening and notes; Gem offers sourcing, application review and fraud detection agents; and hireEZ describes an agent that sources, reviews, screens and schedules on top of an existing ATS, with recruiters approving decisions.
Are AI recruiting agents legal?
Using them is legal in most places, but the employer or agency stays responsible for what they do. Anti-discrimination law applies to any selection decision, New York City requires bias audits and notices for automated employment decision tools, Illinois requires notice of AI use in employment decisions, and Colorado, California and the EU have rules on automated decisions with notice, review and record duties. This is general information, not legal advice.
Will AI agents replace recruiters?
The tasks agents take on today are the repetitive ones: searching, first-pass review, drafting outreach and scheduling. Vendors themselves describe recruiters approving decisions and setting the rules, and the laws that regulate automated hiring decisions place duties on the employer, such as human review and explanations, that need a person who understands the role.