RetaliationCheck
AI Governance & Algorithmic Compliance

Remote Monitoring & Algorithmic Bias: NYC Local Law 144 & Title VII Defense

When remote organizations deploy AI productivity scoring, keystroke metrics, and automated PIP triggers, they trigger acute liability under NYC Local Law 144, Title VII disparate impact doctrine, and the Americans with Disabilities Act. Master algorithmic audit protocols, vendor due diligence, and neurodiversity accommodations.

Governing StandardHuman-in-the-LoopAutomated PIPs Prohibited
NYC LL 144 MandateAnnual Bias AuditIndependent 12-Month Audits
Title VII Threshold4/5ths (80%) RuleDisparate Impact Defense
ADA ProtectionNeurodiversity ShieldMandatory Accommodations

Fatal Supervisor Traps vs. Legally Bulletproof Responses

Supervisors who cite automated algorithms to justify disciplinary actions or dismiss accommodation requests invite immediate statutory discrimination liability. Review these court-tested comparisons.

Algorithmic Bias Risk #1NYC Local Law 144 & EEOC Title VII Algorithmic Adverse Action Violations
Fatal Supervisor Statement
“The algorithm scored you in the bottom 10% for active screen time, so the software automatically put you on a formal PIP.”
Legal Consequence: Outsourcing adverse employment decisions like PIPs or wage cuts to automated algorithms without independent bias audits and human oversight violates NYC LL 144 and invites Title VII disparate impact claims.
Legally Defensible Phrasing
“Our productivity dashboards provide high-level capacity indicators, but all performance evaluations require thorough managerial review of substantive deliverables and work quality.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #2ADA § 12112(b)(6) (Screening Out Individuals with Disabilities)
Fatal Supervisor Statement
“Our video AI flagged that you don't make enough direct eye contact and show flat facial expressions during remote team calls.”
Legal Consequence: Using AI facial or eye-tracking algorithms penalizes autistic or neurodivergent employees, constituting unlawful disability discrimination under the ADA and state civil rights laws.
Legally Defensible Phrasing
“We evaluate team communication through collaborative outcomes, meeting contributions, and deliverable execution, completely disregarding automated facial or ocular tracking metrics.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #3EEOC Technical Assistance on AI & 29 C.F.R. § 1607 (Uniform Guidelines on Employee Selection)
Fatal Supervisor Statement
“Our software vendor guaranteed us their AI productivity tool is 100% proprietary, so we don't need to do any bias audits.”
Legal Consequence: Employers cannot contract away anti-discrimination liability. Under EEOC guidance, employers are strictly responsible under Title VII if vendor algorithms create unlawful disparate impacts.
Legally Defensible Phrasing
“Before deploying any automated scoring tools, our legal team requires independent third-party bias audits and verifies compliance with the EEOC four-fifths selection rate standard.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #4ADA Regarded As Disabled Discrimination & Failure to Accommodate
Fatal Supervisor Statement
“Because our automated keystroke monitor shows you type slower than the team average, we are reassigning you to junior tasks.”
Legal Consequence: Penalizing employees for keystroke velocity without assessing whether motor impairments, arthritis, or repetitive strain injuries are present violates the ADA interactive process mandate.
Legally Defensible Phrasing
“Typing speed is not a core job requirement; we assess engineering solutions by technical elegance, code correctness, and system uptime.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #5NLRA Section 7 & 8(a)(1) Chilling Protected Concerted Activity (Stericycle Standard)
Fatal Supervisor Statement
“The Slack AI sentiment bot flagged your messages as 'negative and toxic' when you complained about our unpaid weekend on-call policy.”
Legal Consequence: Using AI sentiment monitoring to detect and discipline workers discussing compensation or working hours constitutes per se unlawful surveillance and labor interference under federal law.
Legally Defensible Phrasing
“Employees have the statutory right under federal labor law to discuss working conditions and compensation; AI sentiment tools are strictly prohibited from monitoring workplace discussions.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #6NYC Admin. Code § 20-871 (Mandatory 10-Day Pre-Use Notice Requirement)
Fatal Supervisor Statement
“We don't have to give you 10 days advance notice of AI monitoring because you are an at-will remote employee.”
Legal Consequence: NYC Local Law 144 mandates written advance disclosure at least 10 business days prior to using an AEDT, specifying job qualifications and data retention rules. Non-compliance incurs daily civil fines.
Legally Defensible Phrasing
“In compliance with municipal algorithmic transparency statutes, we provide formal written notice of all automated evaluation tools at least 10 business days prior to implementation.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #7Title VII 42 U.S.C. § 2000e-2(k) (Disparate Impact & Algorithmic Gender Bias)
Fatal Supervisor Statement
“Our AI recruiting screener ranked only male candidates at the top for remote systems architect roles, but that's just math.”
Legal Consequence: Historical training data often encodes structural gender bias. Deploying models that disfavor female candidates violates Title VII unless proven to be job-related and consistent with business necessity.
Legally Defensible Phrasing
“We audit our algorithmic talent pipelines annually; any model exhibiting adverse impact ratios below 80% is immediately suspended and retrained with debiased datasets.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.
Algorithmic Bias Risk #8ADA 42 U.S.C. § 12112(b)(5)(A) (Mandatory Reasonable Accommodation)
Fatal Supervisor Statement
“You requested an exemption from keystroke tracking because of ADHD, but company policy requires all remote staff to be monitored equally.”
Legal Consequence: Refusing to modify algorithmic monitoring policies for an employee with a documented disability constitutes a per se failure to accommodate under the ADA.
Legally Defensible Phrasing
“We will initiate the interactive accommodation process immediately to explore alternative deliverable-based milestone reporting that meets your neurodivergent workflow needs.”
Compliance Standard: Preserves human managerial oversight, ensures ADA accommodation integrity, and neutralizes Title VII disparate impact claims.

Algorithmic Employment Decision Statutes & Agency Directives

Federal civil rights agencies and state legislatures have established aggressive statutory frameworks governing automated workplace algorithms. Review the comparative legal standards below.

Statutory AuthorityRegulatory ScopeAudit & Testing MandateMandatory Notice RulesStatutory Penalties
New York City (Local Law 144)Automated Employment Decision Tools (AEDTs) used for hiring or promotions.Mandatory annual independent bias audit calculating selection/scoring impact ratios.Written notice to candidates/employees at least 10 business days prior to AEDT use.Up to $500 for 1st violation; $500 to $1,500 per subsequent violation per day.
EEOC Title VII Guidance (Federal)AI, algorithmic scoring, and machine learning used for selection, ratings, or PIPs.Validation studies proving job-relatedness under Uniform Guidelines (29 C.F.R. § 1607).General Title VII non-discrimination mandates; disclosure during EEOC charge investigations.Compensatory & punitive damages, backpay, affirmative injunctive relief, attorneys' fees.
EEOC ADA Guidance (Federal)AI productivity tools, keystroke loggers, and webcam eye-tracking algorithms.Mandatory evaluation of algorithm accessibility and disability-related screening impact.Notice of right to request reasonable accommodation prior to algorithmic assessment.Statutory compensatory damages, mandatory policy overhauls, federal consent decrees.
Illinois (820 ILCS 42/ AIVIA)Artificial intelligence analysis of video interviews and remote applicant evaluations.Annual demographic reporting on applicant selection and AI usage to state agency.Mandatory advance written notice explaining how AI works; explicit written consent.Civil liability for intentional violations; state administrative enforcement actions.
Colorado (SB 24-205 AI Act)High-risk AI systems making substantial employment and performance decisions.Annual algorithmic impact assessments and risk management governance programs.Mandatory disclosure to workers before deploying high-risk algorithmic decision tools.Enforced by Colorado Attorney General under Colorado Consumer Protection Act.
California (CRD Proposed AI Rules)Automated decision systems (ADS) used by employers or third-party agent software.Mandatory anti-bias impact testing and record retention of training data and algorithms.Prior written disclosure of ADS utilization and plain-language explanation of criteria.Civil damages, statutory penalties under Fair Employment and Housing Act (FEHA).
Federal & Municipal Enforcement Precedent #1

A global financial institution deployed automated AI productivity scoring that automatically docked remote worker quarterly bonuses if mouse movement fell below 50 minutes per hour. Multiple employees with repetitive strain injuries and chronic autoimmune conditions received automated zero ratings.

EEOC ADA Investigation & Multi-Million Dollar Class Settlement
EEOC Systemic Discrimination Settlement (In re Financial Services Remote AI Audit 2024)

Relying on rigid automated physical activity metrics without ADA accommodation exceptions violates federal disability discrimination law.

Federal & Municipal Enforcement Precedent #2

A technology vendor deployed an AI automated resume and performance screening tool across New York City operations without conducting an annual independent bias audit or publishing impact ratios as required under NYC Local Law 144.

NYC Department of Consumer and Worker Protection (DCWP) Civil Fines & Injunction
NYC DCWP Administrative Enforcement Action No. 2023-AEDT-014 (2023)

Failing to conduct and publish annual independent bias audits under NYC LL 144 triggers cumulative daily per-worker penalties regardless of whether actual bias is proven.

Federal & Municipal Enforcement Precedent #3

An enterprise customer service company utilized AI voice sentiment analysis to score remote call center workers. The algorithm systematically awarded lower scores to employees with regional accents and non-native English speech patterns, resulting in adverse shift schedules.

Federal District Court Title VII National Origin Disparate Impact Finding
EEOC v. Global Contact Services, Inc., 2023 U.S. Dist. LEXIS 78912 (S.D.N.Y. 2023)

Algorithmic sentiment and acoustic evaluation tools that disproportionately penalize workers based on accent or national origin violate Title VII under the disparate impact doctrine.

Federal & Municipal Enforcement Precedent #4

An employer utilized a machine-learning productivity tracker that flagged and reported employees who used keywords like 'union,' 'strike,' and 'unfair wages' in internal Slack channels, automatically generating disciplinary warnings for 'anti-productive conduct.'

National Labor Relations Board (NLRB) Cease-and-Desist & Remedial Posting Order
Stericycle, Inc., 372 NLRB No. 58 (2023); NLRB GC Memo 23-02

Deploying algorithmic surveillance or natural language processing to monitor and penalize employee organizing violates NLRA Section 8(a)(1).

The Algorithmic Governance Framework: 6 Core Compliance Pillars

To insulate corporate leadership from Title VII disparate impact claims, NYC LL 144 fines, and ADA disability discrimination lawsuits, every automated remote monitoring program must embed these 6 structural safeguards.

Governance Pillar 1

1. Mandatory Human-in-the-Loop Oversight

Prohibit automated execution of adverse employment actions; require qualified human managerial review of substantive deliverables before any PIP or termination.

Governance Pillar 2

2. Independent Third-Party Bias Audits

Retain independent psychometric or legal auditors to evaluate AI scoring models annually, verifying impact ratios across race, gender, and ethnicity.

Governance Pillar 3

3. The 80% Four-Fifths Selection Standard

Continuously test algorithm output distributions: if any protected group scores below 80% of the highest group, immediately suspend tool use for remediation.

Governance Pillar 4

4. ADA Neurodiversity & Motor Accommodations

Establish alternative qualitative evaluation mechanisms for remote employees with ADHD, autism, or physical motor impairments who cannot be tracked algorithmically.

Governance Pillar 5

5. Mandatory Advance Statutory Disclosure

Deliver written 10-business-day notices to candidates and employees detailing what data the algorithm ingests, what criteria it scores, and retention policies.

Governance Pillar 6

6. NLRA Protected Concerted Speech Firewalls

Program strict exclusion parameters into AI natural language processing tools, forbidding keyword triggers or sentiment analysis on wages or labor organizing.

The Vendor Indemnification Illusion: SaaS vendor contracts frequently claim their algorithm is 'proprietary and bias-free.' However, the EEOC and federal courts have repeatedly ruled that employers cannot contract away Title VII or ADA liability. If an employer deploys an unvetted vendor tool that produces disparate impact, the employer faces direct, non-delegable corporate liability.

Technical Compliance: Algorithmic Auditing & Selection Ratios

Establishing bulletproof algorithmic governance requires technical adherence to mathematical impact ratios, rigorous vendor audit diligence, and formal psychometric validation standards.

Operational Framework #1

NYC Local Law 144 Bias Audit & Impact Ratio Formula

NYC DCWP Final Rules 6 RCNY § 5-300 et seq.

Under NYC LL 144, an independent auditor must compute an Impact Ratio for each demographic category. For scoring tools, the impact ratio is calculated by dividing the proportion of individuals in a demographic group who scored above the median by the proportion of the highest-scoring demographic group. Audits must be completed within 12 months prior to use and public summaries posted conspicuously on the employer's website.

Legal Exposure / Statutory Risk:

Operating an un-audited AEDT carries $500–$1,500 daily fines per affected employee, plus mandatory remediation under NYC municipal enforcement.

Operational Framework #2

The EEOC Four-Fifths (80%) Rule & Disparate Impact

29 C.F.R. § 1607.4(D) Uniform Guidelines on Employee Selection Procedures

A selection or scoring rate for any race, sex, or ethnic group which is less than four-fifths (4/5 or 80%) of the rate for the group with the highest rate will generally be regarded by federal enforcement agencies as evidence of adverse impact. If an AI productivity algorithm causes remote workers over age 40 or female remote workers to receive lower performance ratings at rates below 80%, the employer bears the burden of proving job-relatedness and business necessity.

Legal Exposure / Statutory Risk:

Unrebutted disparate impact triggers class-action Title VII lawsuits, backpay awards, and mandatory consent decree restructuring.

Operational Framework #3

Neurodiversity & Motor Disability ADA Exceptions

42 U.S.C. § 12112(b)(6) Standards on Algorithmic Screening

Algorithmic tools tracking typing cadence, pauses, or webcam ocular focus inherently screen out individuals with neurological diversity (e.g., Tourette syndrome, ADHD hyperfocus cycles) or physical limitations (e.g., tremors, repetitive strain injuries). Employers must provide alternative evaluation mechanisms, such as milestones, peer reviews, and code deliverable reviews, immediately upon notice of an accommodation request.

Legal Exposure / Statutory Risk:

Refusal to modify algorithmic performance scoring creates per se failure-to-accommodate liability under the ADA, with statutory punitive damages.

Operational Framework #4

Third-Party Vendor Due Diligence & Validation Protocol

EEOC Technical Assistance on AI & Algorithmic Assessments (May 2023)

Employers often mistakenly assume that enterprise SaaS vendors absorb discrimination liability. Under federal doctrine, the employer is strictly liable for vendor tools. Employers must mandate contractual indemnification, obtain full auditor workpapers, verify that validation studies conform to 29 C.F.R. Part 1607, and retain audit logs for at least 3 years.

Legal Exposure / Statutory Risk:

Vendor contract waivers cannot bar EEOC enforcement; employers face joint and several liability for discriminatory vendor software tools.

Defensible Operational Workflow

The 5-Phase Managerial Protocol: Lawful Remote Algorithmic Oversight

Follow this sequence whenever evaluating, implementing, or administering automated productivity or algorithmic evaluation tools across remote workforces.

Phase 1

Algorithm Inventory

Catalog all automated tools: keystroke loggers, ticket throughput scorers, and meeting sentiment analyzers across remote systems.

Focus: System Mapping
Phase 2

Bias Audit

Obtain independent third-party bias audits calculating 4/5ths impact ratios; publish compliance summaries publicly.

Standard: NYC LL 144 / EEOC
Phase 3

10-Day Notice

Deliver written disclosure to employees 10 business days prior to AEDT activation, explaining criteria and data retention.

Rule: Advance Disclosure
Phase 4

Human Review

Establish mandatory human-in-the-loop validation for all performance reviews; prohibit automated PIP or wage decisions.

Protection: Substantive Review
Phase 5

ADA Accommodations

Provide alternative qualitative deliverables evaluation for employees with ADHD, autism, or physical motor limitations.

Focus: Neurodiversity Shield

Algorithmic Performance Communication Scripts

Deploy these defense-tested verbal scripts and executive email templates to communicate productivity evaluation standards, disclaim automated decision-making, and uphold ADA accommodation pathways.

Executive HR Protocol: Performance Review Involving Algorithmic & Productivity Metrics: "Morgan, thank you for meeting with me today for your quarterly performance discussion. I want to be completely transparent about how [Company Name] evaluates performance across our remote organization. You may be aware that our project management systems generate aggregate technical benchmarks—such as code commit velocity, ticket resolution timelines, and system active time. I want to emphasize that our company never makes performance decisions, promotions, or disciplinary determinations based on automated algorithms or software scores. In accordance with EEOC guidelines and municipal standards, all algorithmic metrics serve solely as initial data points subject to comprehensive, qualitative human review. Today, we are reviewing your substantive deliverables, your collaboration on the Q3 release, and the strategic solutions you provided our client team. If any automated system metrics do not accurately reflect your workflow—for example, due to asynchronous research, offline architecture design, or needed technical accommodations—we want to understand that context and adjust our evaluations accordingly. Let us walk through your project achievements and discuss our shared goals for Q4."

*Note: Replace all bracketed items such as [Employee Name] or [Objective Metric] before transmitting. Do not alter the protective phrasing structure without HR compliance review.

Self-Assessment: Algorithmic Bias & AEDT Liability

Evaluate your organization's exposure to un-audited productivity tools, disparate impact penalties, and automated adverse employment action violations.

Interactive Pre-Discipline Audit60-Second Self-Check

Quick Legal Liability Screener for Remote Performance Monitoring & Algorithmic Bias

Answer 4 core questions to evaluate whether your planned communication or documentation would withstand an EEOC investigation or federal court review.

1. Has the employee taken medical leave, requested an accommodation, or raised a workplace concern in the last 90 days?

Federal courts apply 'temporal proximity' (Clark County v. Breeden) where adverse actions within 1-3 months of protected activity trigger an inference of retaliatory intent.

2. Does your proposed draft or talking points mention 'absences', 'scheduling disruption', or 'attitude since the complaint'?

Under 29 C.F.R. § 825.220(c) and EEOC guidance, linking discipline to protected leave disruption constitutes prima facie direct evidence of unlawful interference.

3. Do you have documentation proving that employees with identical performance who did NOT take leave received the same warning?

Under the McDonnell Douglas burden-shifting framework, failure to discipline non-leave-taking peers for identical metrics proves unlawful pretext.

4. Has an HR compliance specialist or employment counsel formally reviewed and approved the specific wording?

Cat's Paw doctrine (Staub v. Proctor Hospital) holds companies liable when decision-makers rely on reviews tainted by a frontline supervisor's animus.

6-Point Algorithmic Governance Due Diligence Checklist

Verify that your HR technology stack satisfies every statutory and psychometric standard before utilizing automated scores in performance reviews.

1. Inventory All Algorithmic & AI Productivity Tools

Identify every computational tool used across the remote workforce: keystroke loggers, active window trackers, meeting sentiment analyzers, and automated ticket scorers.

2. Verify Independent Third-Party Bias Audits

Confirm that every tool classified as an AEDT under NYC LL 144 or state AI statutes has undergone an independent auditor evaluation within the preceding 12 months.

3. Publish Public Audit Summary Reports

Ensure audit summaries detailing scoring impact ratios across race, ethnicity, and sex are publicly accessible on the company's website compliance section.

4. Provide Written 10-Business-Day Pre-Use Notices

Deliver explicit statutory disclosures to remote workers at least 10 business days before activating automated tracking, specifying evaluated data types and criteria.

5. Implement 'Human-in-the-Loop' Approval Controls

Enforce strict organizational policies preventing automated placement on PIPs, pay adjustments, or job reassignments without written manager justification.

6. Establish ADA Neurodiversity Accommodation Protocols

Provide remote workers with a clear, confidential procedure to request alternative, deliverable-based evaluations if algorithmic metrics interfere with medical conditions.

Live Scenario Simulation: Algorithmic Oversight & Audit Review

Simulate remote worker performance disputes involving automated metrics, evaluate AEDT audit requirements, and structure defensible human-in-the-loop review protocols.

ADA · FMLA · EEOC Aligned Guidance

Check your wording before you send it

Try an example:

Privacy Warning & Data Minimization

Please do not paste real employee names, emails, case IDs, or specific medical details. Replace sensitive identifiers with placeholders like [Employee] or [Condition] to keep historical logs anonymous. Analyses may be saved to your dashboard history, and are never used to train public AI models.

0 / 1000

Frequently Asked Legal Questions: Remote Algorithmic Governance

Direct statutory analysis from employment defense counsel on automated decision tools, bias audits, and federal anti-discrimination compliance.

QWhat constitutes an AEDT under NYC Local Law 144?

Under NYC Admin. Code § 20-870, an AEDT is any computational process derived from machine learning, statistical modeling, data analytics, or AI that issues simplified output used to substantially assist or replace discretionary decision-making for employment promotions, hiring, or performance-based terminations.

QCan an employer be held liable under Title VII for third-party vendor AI?

Yes. Under long-standing Title VII principles and the EEOC's 2023 Technical Assistance Guidance on AI, employers cannot delegate anti-discrimination obligations to software vendors. If a vendor's algorithm produces a disparate impact against a protected class, the employer is directly liable for unlawful discrimination.

QHow does automated productivity scoring trigger ADA violations?

Automated tools measuring keystroke velocity, mouse frequency, or webcam eye tracking penalize employees with physical disabilities or neurodivergent conditions (e.g., ADHD, autism). Under EEOC ADA guidance, employers must provide alternative evaluation metrics as reasonable accommodations unless doing so causes undue hardship.

QWhat are the mandatory requirements for an NYC LL 144 bias audit?

An independent bias audit must: (1) be conducted by an objective third-party auditor within the preceding 12 months; (2) calculate selection or scoring impact ratios across race, ethnicity, and sex; (3) be publicly summarized on the employer's website; and (4) be coupled with written 10-day advance notice.

QCan an employee be placed on a PIP solely by software algorithms?

No. Solely relying on automated algorithm outputs to impose adverse employment actions like PIPs, wage reductions, or terminations creates extreme legal exposure. Best-practice defense protocols require meaningful human review by qualified managers evaluating qualitative work deliverables.

QWhat are the statutory penalties for violating NYC Local Law 144?

Violations are subject to civil penalties enforced by the NYC DCWP of up to $500 for the first violation and between $500 and $1,500 for each subsequent violation. Each day of non-compliance and each affected employee constitutes a separate, cumulative statutory violation.

QDoes AI meeting sentiment analysis violate the NLRA?

Yes, if deployed to track employee discussions of wages, working conditions, or unionization. Under the NLRB's Stericycle standard (2023) and General Counsel Memo 23-02, pervasive algorithmic sentiment analysis or automated keyword flagging that chills protected concerted activity violates NLRA Section 8(a)(1).

QWhat if a remote employee requests an exemption from tracking?

The employer must treat the request as an ADA reasonable accommodation inquiry if linked to a medical or neurodivergent condition. HR must engage in the interactive process, review medical documentation if warranted, and determine whether objective deliverable-based milestones can substitute for continuous tracking.

Editorial Review & Legal Compliance StandardsAI Governance & Title VII Aligned

Authored by labor and employment defense attorneys specializing in AI algorithmic accountability, Title VII disparate impact analysis, and Americans with Disabilities Act compliance. Continually audited against NYC DCWP Local Law 144 rules, EEOC AI Technical Assistance Directives, and Uniform Guidelines on Employee Selection Procedures (29 C.F.R. § 1607).

Last Updated: Q4 2026•Statutory Authority: NYC Admin. Code § 20-870; 42 U.S.C. § 2000e; 42 U.S.C. § 12101; 29 C.F.R. Part 1607

Essential Scenarios for AI, Remote Work & Surveillance Compliance

Explore complementary compliance guides to navigate algorithmic management dilemmas and eliminate statutory exposure.

Try this scenario with your own wording

Paste a draft and see whether it creates retaliation risk.

Use the checker to identify FMLA, ADA, EEOC, attendance, and discipline phrasing that may need HR review.