19 patterns in this category
HR and recruiting agents systematically fail at three interdependent stages of the talent lifecycle: screening (demographic bias, skill-assessment conflation, fairness violations), offer generation (compensation-benchmark staleness, precedent mismatches, negotiation-term dropouts), and the ongoing employment cycle (onboarding compliance loss, accommodation loss, retention-prediction hallucination and self-fulfilling loops). These failures are interconnected: bias upstream in screening propagates into attrition modeling; offer-generation exceptions that are not carried to onboarding create day-one trust violations; retention predictions that become visible to managers alter manager behavior in ways that self-fulfill the prediction. The category spans 19 patterns across 4 goals, concentrating in four failure mechanisms: demographic bias and proxy discrimination, knowledge staleness (training data overriding live tools), multi-agent handoff information loss, and self-fulfilling feedback when predictions become visible to decision-makers.
| Goal | Covers | Patterns |
|---|---|---|
| Candidate Screening | Bias, demographic proxies, skill-assessment conflation, fairness, accommodation capture | 5 |
| Offer Generation | Compensation-benchmark staleness, leveling-precedent mismatches, negotiated-term handoff loss, compliance-gate staleness | 4 |
| Onboarding | Multi-session context loss, policy-retrieval mismatch, knowledge staleness (immigration), accommodation loss, compliance-gate verification | 5 |
| Retention Prediction | Narrative hallucination, benchmark staleness, cohort-mismatch retrieval, handoff loss, self-fulfilling feedback loops | 5 |
Total: 19 patterns
The four goals form a pipeline with feedback loops. Candidate Screening is the first gate: demographic bias in screening eliminates candidates, shaping the cohort that attrition models later train on. Offer Generation takes screened candidates and produces compensation and terms; offers that fail to capture visa contingencies or exceptions create day-one problems in Onboarding. Onboarding is the new hire’s first operational experience; accommodations or compliance requirements missed in onboarding affect retention risk and early attrition. Retention Prediction closes the loop: it uses hire and performance data from earlier stages, and once a risk score is visible to managers, it becomes an active intervention that can self-fulfill. To localize an incident by symptom: a qualified candidate was screened out → check Candidate Screening’s bias and skill-assessment patterns; an offer was sent with terms that contradicted earlier promises → check Offer Generation’s handoff patterns and compliance gates; a new hire arrived without their requested accommodation or to wrong benefits information → check Onboarding’s handoff and policy-retrieval patterns; an employee’s attrition risk was flagged but the narrative didn’t match their actual signals → check Retention Prediction’s hallucination and benchmark patterns.
No. Proxy discrimination occurs when models learn that features like graduation year, job history gaps, or communication style correlate with historical hiring outcomes and use those correlates as hidden signals. Removing explicit demographic features does not prevent proxy learning. Mitigation requires testing for disparate impact across demographic groups post-hoc and directly auditing which non-demographic features drive disparate outcomes.
When an offer contains visa-contingent remote-work terms that are never captured in the structured onboarding handoff, the new hire arrives without that arrangement, creating an immediate trust violation. Early-stage trust erosion correlates with early attrition. When an accommodation requested during recruiting is not provisioned day one, the new hire’s experience is degraded from start. Onboarding quality affects retention risk; attrition models trained on cohorts with low-quality onboarding outcomes incorporate that degradation as if it were unavoidable.
Add mandatory structured fields to the handoff schema for every category of information the downstream agent needs to know about (accommodations, exceptions, recent changes, risk flags). Require every upstream agent to populate those fields explicitly before handing off, rather than leaving information in free-text notes. Pair structured fields with a reconciliation check that scans upstream free-text for any item not represented in the structured fields, flagging mismatches before the downstream agent proceeds.
Run a score-visibility holdout experiment: withhold attrition-risk scores from managers for a control group and compare actual attrition. If visible-score groups have higher attrition than control groups with comparable underlying risk factors, the difference is attributable to manager behavior triggered by the score. Require deliberate experimental design but is the only reliable way to separate prediction from feedback-loop.
Compensation levels for specialized roles (AI/ML engineering, data science) move quickly in response to market demand shifts. Models trained on historical data absorb pretraining-era compensation figures, which are plausible and confident-sounding but materially stale. Offers grounded in stale figures miss market and candidates decline or counter. Fix: mandate live compensation-benchmarking tool calls for every market-rate figure in an offer.
Employees Flagged as High Attrition-Risk by the Retention-Prediction Model Are Systematically Deprioritized for Growth Opportunities, Stretch Assignments, and Promotion Consideration by Managers Aware of the Score, Causing the Flagged Employees to Actually Leave at Higher Rates as a Consequence of the Flag Itself
A Retention-Prediction Agent's Embedding Search over Past Departed-Employee Records Surfaces a Lexically Similar but Substantively Different Attrition Case as the Nearest-Neighbor Comparable, and the Agent Anchors a Current Employee's Risk Score to That Mismatched Precedent
An Offer-Generation Agent's Embedding Search over Past Leveling Decisions Surfaces a Role with Lexically Similar Title and Description but a Different Career Track or Leveling Framework, and the Agent Anchors the New Offer's Band to That Mismatched Precedent
An Onboarding Agent's Retrieval Step Surfaces a Benefits-Eligibility Policy Document for a Similarly Named but Different Work Jurisdiction or Employment Classification, and the New Hire Is Told Incorrect Eligibility Terms Because the Embedding Match Favored Lexical Similarity Over an Exact Jurisdiction/Classification Match
AI-Assisted Interview-Screening Agent Weights a Candidate's Vocal Confidence, Fluency, and Positive Sentiment in the Interview Transcript More Heavily Than the Substantive Correctness or Depth of Their Answers, Systematically Favoring Articulate-but-Shallow Candidates Over Substantively Strong but Less Polished Ones
A Recruiting-Coordinator Agent's Conversation With a New Hire Establishes a Confirmed Workplace Accommodation (e.g., an Ergonomic Setup or Assistive Equipment Tied to a Documented Need) During the Pre-Start Conversation, but the Structured Handoff Record Passed to the Downstream Onboarding/IT-Provisioning Agent Omits the Accommodation Field, So the Provisioning Agent Ships Standard-Issue Equipment and the New Hire Arrives on Day One Without What Was Already Agreed
A Compensation-Review Agent That Confirms a Manager-Approved Off-Cycle Pay Adjustment for an Employee Hands Off Its Output to a Downstream Retention-Prediction Agent Through a Structured Schema That Has No Field for a Pending or Just-Approved Comp Change, So the Retention-Prediction Agent Computes the Employee's Updated Attrition-Risk Score From Stale Compensation Data and Continues Flagging Them as High-Risk Even Though the Underlying Driver of That Risk Was Already Resolved
A Recruiter-Screening Agent's Free-Text Notes Recording a Candidate's Disclosed Accessibility Accommodation Request Are Not Captured in the Structured Handoff Schema Passed to the Interview-Scheduling Agent, Which Books an Interview Format That Does Not Provide the Requested Accommodation
A Recruiting-Coordinator Agent's Free-Text Negotiation Notes Recording a Visa-Status-Contingent Remote-Work Exception Are Not Captured in the Structured Handoff Schema Passed to the Offer-Letter-Generation Agent, Which Issues a Standard In-Office Offer That Contradicts What the Candidate Was Told
An Onboarding Agent That Generates a New Hire's Task Checklist Establishes a Conditional Requirement Early in the Onboarding Conversation (e.g., "Since You're an International Hire, You'll Also Need to Complete X") but Loses Track of That Condition in a Later Session, Generating a Follow-Up Checklist That Omits the Conditional Task Entirely
An Offer-Generation Agent Configured to Send a Finalized Offer Letter Automatically Once a Candidate Clears a Conditional-Offer Gate Sends the Letter Based on the Background-Check Step No Longer Appearing in Its Own List of Outstanding Blockers, Without Re-Querying the Background-Check Vendor's API for the Current Status, and the Check Had Actually Moved to "Pending Dispute" Rather Than "Clear"
An Onboarding Agent Responsible for Notifying a New Hire's Manager When Pre-Employment Screening Steps Clear -- So the Manager Can Authorize Systems Access and a Start-Date Confirmation -- Sends the "Background Check Cleared, Access Approved" Notification Based on the Screening Step Simply No Longer Appearing in the Agent's Outstanding-Tasks List, Without Re-Querying the Background-Check Vendor's API for the Actual Current Status Field, and the Step Had In Fact Moved to "Pending Adjudication" Rather Than "Clear"
Resume screener learns that names, dates, or graduation years correlate with performance; uses these as hidden proxy signals for protected characteristics (age, race, national origin)
AI resume screener uses exact keyword matching; rejects qualified candidates with industry synonyms or relevant but non-exact terminology (e.g., "web development" vs "frontend engineering")
Candidate-Screening Agent Over-Weights Surface Keyword Matches Against the Job Description, Systematically Filtering Out Qualified Candidates Who Describe Equivalent Experience Differently
A Retention-Prediction Agent Asked to Produce a Narrative Justification for a High-Attrition-Risk Flag Generates a Specific, Plausible-Sounding Detail About a Recent One-on-One Conversation Between the Employee and Their Manager (e.g., "The Employee Told Their Manager They Were Frustrated With the Lack of Promotion Timeline") That Does Not Appear in Any Manager Note, Survey Response, or HRIS Record the Agent Had Access To, and HR Acts on the Fabricated Detail as if It Were Documented Evidence
An Offer-Generation Agent Answers Market-Rate Compensation Questions from Salary Figures It Absorbed During Pretraining Rather Than Calling the Live Compensation-Benchmarking Tool It Has Available, Producing Offers Anchored to Outdated Market Data for Fast-Moving Roles
A Retention-Prediction Agent, When Asked to Contextualize Whether an Employee's Computed Risk Score Is High Relative to Peers, Answers Using a General Industry Attrition-Rate Figure It Absorbed During Pretraining Rather Than Calling the Live Internal Cohort-Comparison Tool Available to It, Producing a Relative-Risk Characterization Anchored to an Outdated or Generic Benchmark Instead of the Company's Actual, Current Department-Level Attrition Baseline
An Onboarding Agent Answering a New International Hire's Question About Visa-Sponsorship Steps, Timelines, or Document Requirements Answers from Generic Immigration-Process Knowledge It Absorbed During Pretraining Rather Than Calling the Company's Live Immigration-Policy Tool, Producing Guidance That Reflects an Outdated Visa Category, Processing Timeline, or Document List Instead of the Company's Actual Current Sponsorship Process