Automated Employment Decision Tool (AEDT)
An AEDT uses machine learning, statistics, or AI to produce a scored or ranked output that substantially assists or replaces a human employment decision. New York City Local Law 144 was the first law to mandate independent bias audits and candidate-notice requirements for these tools.
Editorial review notice
This page has not received legal review and is for informational purposes only. Employment law obligations vary by jurisdiction and change frequently. Confirm current requirements and applicability with qualified employment counsel before making compliance decisions.
What is an AEDT?
An automated employment decision tool (AEDT) is a computational process — typically using machine learning, statistical modelling, or artificial intelligence — whose output is used to substantially assist or replace a human in making an employment decision. Covered decisions commonly include hiring, promotion, compensation, termination, and performance evaluation.
The term originates in New York City Local Law 144. Similar concepts appear in EEOC technical guidance, proposed and enacted state legislation, and the EU AI Act's classification of recruitment AI as high-risk.
What tools qualify as AEDTs?
Under the NYC LL 144 definition, an AEDT includes tools that:
- Rank or score resumes or job applications algorithmically
- Automatically assess candidates from video or audio recordings
- Produce scored or ranked shortlists from applicant pools
- Score test results or work samples using ML models
Tools that only facilitate human organisation without producing scored recommendations — such as a plain ATS with keyword filtering — may fall outside the definition, depending on the jurisdiction. Standard grammar aids and document formatters are generally not considered AEDTs.
Why are AEDTs regulated?
Regulators have identified three primary risk areas:
- Disparate impact: Models trained on historical hiring data may replicate or amplify existing biases against protected groups.
- Opacity: Candidates and employers may not understand why a recommendation was produced, limiting meaningful review.
- Displacement of human judgment: When humans defer to AI output without scrutiny, individual errors become systemic.
Regulatory frameworks typically respond with mandatory bias audits, transparency obligations, and rights for candidates to request an alternative process or human review.
Common employer obligations
Obligations vary by jurisdiction, but commonly include:
- Commissioning or conducting an independent bias audit before deployment
- Publishing audit results on the company website
- Notifying candidates and employees before using an AEDT to evaluate them
- Providing an accommodation or alternative process on request
- Retaining audit and usage records for a defined period
Even in jurisdictions without explicit AEDT laws, employers may face liability under existing anti-discrimination statutes if an AI tool produces disparate impact against a protected group.
Vendor vs. employer responsibility
Both parties typically share responsibility. The employer remains liable for employment decisions regardless of whether a vendor supplied the AI tool. Vendors may be required to supply technical documentation, facilitate bias audits, and disclose training data sources.
When evaluating vendors, request independent bias audit reports, documentation of training data and selection criteria, explainability of scores, and contractual commitments covering audit cadence and notification of model updates.