AI Insight Lab — The Deployment MemoJune 17, 2026

The Hiring Algorithm

HireVue is in 750 enterprises. NYC Local Law 144 took effect July 2023. ~1 in 4 employers had filed the bias audit. The EEOC guidance is in enforcement posture. What your CHRO hasn't done yet.

The Deployment Memo — AI Insight Lab | June 17, 2026

Background

How enterprise AI hiring tools became a compliance exposure — and why procurement missed it

01 / Background

The Enterprise AI Hiring Market

•HireVue: deployed at 750+ enterprises — Unilever, Goldman Sachs, Hilton, Delta, Intel. Analyzes facial expressions, tone, verbal content in async video interviews.
•Eightfold.ai: skills inference from résumés and profiles; ranks and screens applicants before recruiter review. Clients: Rolls-Royce, Vodafone, Zurich Insurance.
•Paradox (Olivia): conversational AI handles initial screening and scheduling at McDonald's, Nestlé, Unilever — processes more candidates daily than most HR teams review manually in a week.
•Platform-embedded AI: Workday Skills Cloud, SAP SuccessFactors Recruiting AI, Oracle HCM Candidate Experience — AI scoring accepted as standard ATS feature without separate AEDT evaluation.
•Documented efficiency gain: 50–75% reduction in recruiter screening time. The tool class is real. The regulatory gap arrived while adoption was accelerating.

01 / Background

The Regulatory Landscape

•NYC Local Law 144 (effective July 5, 2023): annual independent bias audit required for AEDTs; results must be publicly posted ≥10 business days before use; candidate notice required; alternative process on request.
•AEDT definition is broad: "any computational tool that substantially assists or replaces discretionary decision-making" — covers AI résumé scoring, video interview analysis, and skills-inference ranking.
•Illinois AI Video Interview Act (effective Jan 1, 2020): consent + disclosure + 30-day deletion right for video interview AI. Most enterprise workflows have no deletion request process.
•Chicago Algorithmic Hiring Act (effective May 2025): extends bias audit and notice requirements citywide. EEOC May 2023 guidance: AI adverse impact on protected classes can violate Title VII, ADA, ADEA without intent.
•California and New York state bills advancing in 2026. NYC enforcement guidance (2024): per-candidate, per-day fines; DCWP has initiated enforcement on repeat non-compliance.

01 / Background

The Compliance Gap

•Approximately 1 in 4 NYC-area employers using AI hiring tools had published a compliant bias audit one year after Law 144 took effect.
•Most common gap: AI scoring embedded in ATS platform (Workday, SAP, Oracle) was accepted as a product feature update — no one classified it as an AEDT requiring a separate compliance program.
•Vendor documentation gap: vendor EEO compliance representations describe internal testing, not an independent audit. DCWP clarified: vendor documentation does not constitute a compliant independent audit.
•Retroactive exposure: candidates processed after July 5, 2023 without required notice are candidates whose statutory benefit was denied. Exposure depends on volume, duration, and adverse impact findings.
•Model update staleness: most enterprise AI hiring contracts have no model change notification provisions — a model update invalidates the audit on file without employer knowledge.

Decision Required

Which of your screening tools are automated employment decision tools under NYC Local Law 144 and comparable laws — and have you conducted the required annual bias audit for each?

What does your organization owe the candidates your AI tools have already processed? The retroactive exposure assessment needs to be completed before you publish the prospective audit — because the audit results are public, and a bias finding will generate questions about the retrospective period.

When your vendor updates their model — which every production AI system does continuously — does that update require re-validation of your bias audit on file? Most enterprise contracts do not include this provision.

Four Options

Option A

Continue with vendor-provided documentation and existing procurement terms

Defers the compliance exposure. DCWP has clarified: vendor documentation is not a compliant independent audit. Exposure accumulates with each additional candidate processed.

Option BRecommended

Commission independent annual bias audit, build candidate notice workflow, add model-change provisions to vendor renewals

The compliant path. Audit cost: $15K–$50K per tool. Candidate notice is an ATS configuration change. Model-change provisions belong in the next vendor renewal.

Option C

Restrict AI scoring to post-application stages only — skills assessments, work samples — where AEDT scope is narrower

Reduces compliance surface. Does not address retroactive exposure. Sacrifices efficiency gain in initial screening stages. AEDT requirements still apply to remaining AI use.

Option D

Full AI hiring governance audit before next recruitment cycle — classify every tool, run disparate impact analysis, build compliance architecture

Governance-complete. Requires 60–90 days. May require pausing tool use pending audit results. Right for organizations that discover AI is more embedded in hiring than HR leadership understood.

Recommendation

✓Commission the independent bias audit before the next NYC-area hiring cycle. Legal compliance obligation since July 2023. Audit methodology: four-fifths rule adverse impact analysis. Cost: $15K–$50K per tool. Do this before the next open requisition.
✓Build the candidate notice workflow into your ATS. NYC Law 144 requires notice ≥10 business days before use, with evaluation criteria disclosed and alternative process available. One configuration covers all current U.S. jurisdictions with enacted requirements.
✓Add model change notification and re-audit provisions to every AI hiring vendor contract at the next renewal: 30-day notice of model updates; right to re-audit at vendor cost if scoring behavior changes; technical documentation sufficient for auditor characterization.
✓Conduct retroactive exposure assessment with employment counsel before publishing the audit. Candidates processed after July 5, 2023 without required notice are candidates whose statutory rights were denied. Assess before the audit is public.
✓Audit your ATS for AI features added in platform updates that were not separately evaluated as AEDTs — Workday, SAP SuccessFactors, Oracle HCM all added AI scoring to standard features without AEDT flagging to enterprise clients.

Five Risks

1.NYC Law 144 fines: up to $500/candidate/day accumulate quickly at enterprise hiring volumes

An enterprise processing 1,000 NYC-area candidates per year without a compliant bias audit for 18 months has theoretical maximum exposure of ~$27M. Enforcement has been graduated in early cases, but DCWP has indicated sector audits are coming.

2.EEOC adverse impact: employer cannot transfer liability to the AI vendor

The EEOC confirmed that adverse impact on protected classes from AI tools can violate Title VII, ADA, and ADEA regardless of intent — and the employer's selection decision remains the employer's legal responsibility regardless of how much the algorithm determined the outcome.

3.Model update staleness: vendor updates invalidate bias audit without employer notice

Production AI systems are updated continuously. An audit filed in 2024 may not reflect the model running in 2026. Most enterprise contracts have no model change notification provision — the employer has no way to know the audit is stale.

4.Illinois AI Video Interview Act: missing deletion rights for candidate video data

Illinois requires 30-day deletion of candidate video data upon request. Most enterprise workflows have no deletion request process for candidate video recordings. Candidates who request deletion and receive no response have a statutory violation claim.

5.ATS platform scope gap: AI scoring embedded in standard features is not evaluated as an AEDT

The most common compliance gap is an ATS platform that added AI scoring as a feature update — accepted by HR technology teams as a productivity improvement, not evaluated by legal as a new AEDT triggering independent regulatory obligations.

Six Questions Before the Next Hiring Cycle

1.Has your organization inventoried AI and algorithmic tools used in hiring — and for each tool meeting the AEDT definition, has an independent bias audit been conducted and published on your company website?
2.Does your candidate notice workflow for NYC-area candidates include disclosure that an AEDT is in use, evaluation criteria, and an alternative process option — provided ≥10 business days before the tool is applied?
3.Do your AI hiring vendor contracts include model change notification provisions — and does a model update by your vendor trigger a re-evaluation of the bias audit on file?
4.Which AI scoring features embedded in your ATS platform (Workday, SAP, Oracle) were evaluated separately as AEDTs — or was AI compliance review applied only to standalone tools?
5.Does your organization have a deletion request process for candidate video data under the Illinois AI Video Interview Act — and when were data retention settings for candidate video last reviewed?
6.What is the retroactive exposure assessment for candidates processed by unaudited AEDTs in New York City after July 5, 2023 — and has that assessment been completed before your next bias audit publication?

AI Insight Lab

The Hiring Algorithm

The efficiency case for AI hiring tools is real. The compliance obligation has been in effect since July 2023. The bias audit is not a governance best practice — it is a legal requirement with per-candidate fines that accumulate at enterprise scale. Commission it before the next hiring cycle.

Read the full memo at aiinsightlab.cloud/memos/hiring-ai-screening

The Enterprise AI Hiring Market

•HireVue: deployed at 750+ enterprises — Unilever, Goldman Sachs, Hilton, Delta, Intel. Analyzes facial expressions, tone, verbal content in async video interviews.

•Eightfold.ai: skills inference from résumés and profiles; ranks and screens applicants before recruiter review. Clients: Rolls-Royce, Vodafone, Zurich Insurance.

•Paradox (Olivia): conversational AI handles initial screening and scheduling at McDonald's, Nestlé, Unilever — processes more candidates daily than most HR teams review manually in a week.

•Platform-embedded AI: Workday Skills Cloud, SAP SuccessFactors Recruiting AI, Oracle HCM Candidate Experience — AI scoring accepted as standard ATS feature without separate AEDT evaluation.

•Documented efficiency gain: 50–75% reduction in recruiter screening time. The tool class is real. The regulatory gap arrived while adoption was accelerating.

The Regulatory Landscape

•NYC Local Law 144 (effective July 5, 2023): annual independent bias audit required for AEDTs; results must be publicly posted ≥10 business days before use; candidate notice required; alternative process on request.

•AEDT definition is broad: "any computational tool that substantially assists or replaces discretionary decision-making" — covers AI résumé scoring, video interview analysis, and skills-inference ranking.

•Illinois AI Video Interview Act (effective Jan 1, 2020): consent + disclosure + 30-day deletion right for video interview AI. Most enterprise workflows have no deletion request process.

•Chicago Algorithmic Hiring Act (effective May 2025): extends bias audit and notice requirements citywide. EEOC May 2023 guidance: AI adverse impact on protected classes can violate Title VII, ADA, ADEA without intent.

•California and New York state bills advancing in 2026. NYC enforcement guidance (2024): per-candidate, per-day fines; DCWP has initiated enforcement on repeat non-compliance.

The Compliance Gap

•Approximately 1 in 4 NYC-area employers using AI hiring tools had published a compliant bias audit one year after Law 144 took effect.

•Most common gap: AI scoring embedded in ATS platform (Workday, SAP, Oracle) was accepted as a product feature update — no one classified it as an AEDT requiring a separate compliance program.

•Vendor documentation gap: vendor EEO compliance representations describe internal testing, not an independent audit. DCWP clarified: vendor documentation does not constitute a compliant independent audit.

•Retroactive exposure: candidates processed after July 5, 2023 without required notice are candidates whose statutory benefit was denied. Exposure depends on volume, duration, and adverse impact findings.

•Model update staleness: most enterprise AI hiring contracts have no model change notification provisions — a model update invalidates the audit on file without employer knowledge.

Decision Required

Which of your screening tools are automated employment decision tools under NYC Local Law 144 and comparable laws — and have you conducted the required annual bias audit for each?

Four Options

Option A

Continue with vendor-provided documentation and existing procurement terms

Defers the compliance exposure. DCWP has clarified: vendor documentation is not a compliant independent audit. Exposure accumulates with each additional candidate processed.

Option BRecommended

Commission independent annual bias audit, build candidate notice workflow, add model-change provisions to vendor renewals

The compliant path. Audit cost: $15K–$50K per tool. Candidate notice is an ATS configuration change. Model-change provisions belong in the next vendor renewal.

Option C

Restrict AI scoring to post-application stages only — skills assessments, work samples — where AEDT scope is narrower

Reduces compliance surface. Does not address retroactive exposure. Sacrifices efficiency gain in initial screening stages. AEDT requirements still apply to remaining AI use.

Option D

Full AI hiring governance audit before next recruitment cycle — classify every tool, run disparate impact analysis, build compliance architecture

Governance-complete. Requires 60–90 days. May require pausing tool use pending audit results. Right for organizations that discover AI is more embedded in hiring than HR leadership understood.

Recommendation

✓Commission the independent bias audit before the next NYC-area hiring cycle. Legal compliance obligation since July 2023. Audit methodology: four-fifths rule adverse impact analysis. Cost: $15K–$50K per tool. Do this before the next open requisition.

✓Build the candidate notice workflow into your ATS. NYC Law 144 requires notice ≥10 business days before use, with evaluation criteria disclosed and alternative process available. One configuration covers all current U.S. jurisdictions with enacted requirements.

✓Add model change notification and re-audit provisions to every AI hiring vendor contract at the next renewal: 30-day notice of model updates; right to re-audit at vendor cost if scoring behavior changes; technical documentation sufficient for auditor characterization.

✓Conduct retroactive exposure assessment with employment counsel before publishing the audit. Candidates processed after July 5, 2023 without required notice are candidates whose statutory rights were denied. Assess before the audit is public.

✓Audit your ATS for AI features added in platform updates that were not separately evaluated as AEDTs — Workday, SAP SuccessFactors, Oracle HCM all added AI scoring to standard features without AEDT flagging to enterprise clients.

Five Risks

1.NYC Law 144 fines: up to $500/candidate/day accumulate quickly at enterprise hiring volumes

2.EEOC adverse impact: employer cannot transfer liability to the AI vendor

3.Model update staleness: vendor updates invalidate bias audit without employer notice

4.Illinois AI Video Interview Act: missing deletion rights for candidate video data

5.ATS platform scope gap: AI scoring embedded in standard features is not evaluated as an AEDT

Six Questions Before the Next Hiring Cycle

1.Has your organization inventoried AI and algorithmic tools used in hiring — and for each tool meeting the AEDT definition, has an independent bias audit been conducted and published on your company website?

2.Does your candidate notice workflow for NYC-area candidates include disclosure that an AEDT is in use, evaluation criteria, and an alternative process option — provided ≥10 business days before the tool is applied?

3.Do your AI hiring vendor contracts include model change notification provisions — and does a model update by your vendor trigger a re-evaluation of the bias audit on file?

4.Which AI scoring features embedded in your ATS platform (Workday, SAP, Oracle) were evaluated separately as AEDTs — or was AI compliance review applied only to standalone tools?

5.Does your organization have a deletion request process for candidate video data under the Illinois AI Video Interview Act — and when were data retention settings for candidate video last reviewed?

6.What is the retroactive exposure assessment for candidates processed by unaudited AEDTs in New York City after July 5, 2023 — and has that assessment been completed before your next bias audit publication?

The Hiring Algorithm

Read the full memo at aiinsightlab.cloud/memos/hiring-ai-screening