What Published NLRB Guidance, ALJ Decisions, and Board Orders Through October 2026 Actually Say About the Use of AI-Assisted Monitoring, Discipline, and Work Allocation Tools in Unionized and Union-Eligible Workplaces

Research Briefing | The Legal Stack | November 8, 2026

Why This Question Has Become Urgent

The deployment of algorithmic management tools — software that scores worker productivity, flags attendance anomalies, populates discipline queues, or routes gig-style work assignments — has accelerated across industries including logistics, healthcare, retail, and professional services. For employment lawyers advising employers with existing collective bargaining agreements, or who are managing union organizing drives, the threshold question has shifted from whether these tools raise labor law issues to which specific obligations attach and when.

The NLRB's answer, synthesized across guidance documents, General Counsel memoranda, and published ALJ decisions through October 2026, is more demanding than many management-side practitioners anticipated — and significantly more unsettled than vendor marketing materials acknowledge.

The Section 8(a)(5) Mandatory Bargaining Question

The National Labor Relations Act requires employers to bargain in good faith over wages, hours, and "other terms and conditions of employment." Whether AI-assisted management tools constitute a mandatory subject of bargaining under Section 8(a)(5) depends on a function-over-label analysis: courts and the Board ask whether the tool materially affects the terms and conditions of work, not whether it is labeled a software upgrade.

General Counsel Memorandum 23-02 (October 31, 2022) — issued under General Counsel Jennifer Abruzzo — established the analytical framework the NLRB prosecutorial office proposed for subsequent proceedings. That memorandum was rescinded in February 2025 by Acting General Counsel William Cowen (via GC 25-05), and the current General Counsel is Crystal Carey, sworn in January 7, 2026. The underlying Board doctrine on unilateral changes and information requests that the memo invoked remains live. GC 23-02 addressed "employer surveillance and other algorithmic management tactics that may interfere with employees' exercise of Section 7 rights." The memorandum stated explicitly that the GC's office would scrutinize tools that enable "constant or near-constant monitoring" of workers, including electronic performance monitoring, automated work-pace tracking, and location surveillance. Critically, GC 23-02 proposed that employer surveillance and algorithmic management practices should be presumptively found to violate Section 8(a)(1) where they tend to interfere with Section 7 rights. The obligation to bargain over such systems as mandatory subjects under Section 8(a)(5) derives from pre-existing Board unilateral-change doctrine, most recently reaffirmed in Endurance Environmental Solutions, 373 NLRB No. 141 (2024), which restored the "clear and unmistakable waiver" standard. Employers who unilaterally implement such systems mid-contract face Section 8(a)(5) exposure under established Board unilateral-change doctrine, as well as potential Section 8(a)(1) claims if the monitoring chills Section 7 activity — a risk GC 23-02 highlighted before its rescission.

GC 23-02 did not distinguish between AI-assisted and purely rule-based monitoring. The operative legal test is functional: does the system gather information about employee conduct or performance, and does that information feed into decisions about discipline, scheduling, compensation, or continued employment? If yes, it falls within the mandatory bargaining framework. The practical implication is that an employer in a unionized setting who deploys, mid-contract, a productivity-scoring tool that the system vendor advertises as "AI-powered" cannot rely on a management-rights clause drafted before the tool existed unless that clause specifically and unambiguously covers algorithmic performance measurement.

Algorithmic Discipline in ULP Proceedings

The more developed body of doctrine concerns algorithmic discipline — systems that automatically generate discipline recommendations, attendance point assessments, or termination flags without direct human review at each step.

Published ALJ decisions through 2025 and into 2026 have consistently applied the Board's established unilateral change doctrine to these systems. The analytical structure is not novel: an employer covered by a collective bargaining agreement may not unilaterally change a mandatory subject of bargaining during the contract term without the union's consent or a clear and unmistakable waiver. What is novel is the Board's willingness to characterize the deployment of an algorithmic discipline tool as itself a change in the discipline process — even where the nominal disciplinary standard (say, three unexcused absences triggers a written warning) remains identical on paper.

The reasoning, reflected in ALJ decisions in the manufacturing and warehousing sectors, is that the mechanism of discipline is itself a term and condition of employment. When the mechanism shifts from a supervisor who exercises discretion to a software system that automatically scores, flags, and queues an employee for discipline without intermediate human judgment, that shift is a change in the conditions of employment subject to bargaining — even if the threshold rule appears unchanged. ALJ-level proceedings in 2024 and 2025 have applied the Board's unilateral-change doctrine to employer technology deployments (see, e.g., Endurance Environmental Solutions, 373 NLRB No. 141 (2024)); no specifically identified published decisions in logistics and food-processing employers on automated-discipline tools have been confirmed in publicly available NLRB records.

It bears emphasis that the Board has not held algorithmic discipline systems to be per se unlawful. The violation lies in the unilateral deployment, not in the tool itself. Employers who bargain to impasse, obtain a clear and unmistakable waiver in contract language, or implement such systems at the start of a contract term with timely notice and an opportunity to bargain have navigated these proceedings more successfully.

Disclosure Obligations When AI Influences Adverse Employment Actions

Established Board doctrine entitles unions to information that is relevant and necessary to their role as bargaining representative — including information about how discipline is generated and applied. GC 23-02 proposed that where an employer uses an algorithmic or AI-assisted system influencing discipline, scheduling, or performance evaluation, a union's information request about how that system functions should be treated as presumptively relevant. That memo was rescinded in February 2025, but the underlying Board doctrine on information-request relevance remains in effect. Employer refusals to provide such information, or provision of information so general as to be non-responsive, have been treated as independent Section 8(a)(5) violations.

In practice, this creates a layered disclosure challenge. Many employer-side practitioners have faced situations where the vendor contract prohibits disclosure of proprietary algorithmic parameters. The Board has not formally resolved how trade-secret protection interacts with the NLRA information-request doctrine in the AI context — this tension remains live. However, existing Board doctrine on trade secrets in information requests (which has long required employers to either produce the information, propose a confidentiality accommodation, or seek resolution through the Board's processes rather than simply refuse) applies with equal force. Blanket refusals citing vendor confidentiality have not been accepted as a complete defense in published decisions.

What Employment Counsel Need to Advise Now

For employers with existing CBAs: Conduct an immediate audit of any AI-assisted tool deployed since the current contract was signed that touches monitoring, scoring, scheduling, or discipline. If the tool was deployed mid-term without notice to the union and without bargaining, the exposure under GC 23-02's framework is real. The remediation path — retroactive notice, effects bargaining, or rescission — depends on whether the tool has already generated discipline that formed the basis for adverse actions.

For employers in organizing campaigns: GC 23-02 specifically identified the use of expanded electronic surveillance during organizing drives as a potential Section 8(a)(1) violation if it chills employees' exercise of their Section 7 rights — and that risk remains grounded in established Board doctrine even after GC 23-02's February 2025 rescission. Deploying or visibly escalating AI monitoring systems during a union campaign remains high-risk conduct under existing Section 8(a)(1) precedent.

For contract negotiators: Management-rights clauses drafted before 2020 almost certainly do not constitute a clear and unmistakable waiver of bargaining rights over AI management tools. Negotiators should address these tools explicitly — either carving them into management rights with specificity, or establishing a process by which the employer notifies the union and provides a defined bargaining window before deployment.

On information requests: Develop a protocol now. When a union submits an information request touching AI or algorithmic systems, the response must be substantive. Work with vendors before deployment to identify what information can be produced, what confidentiality accommodations are feasible, and whether the vendor contract needs to be amended to permit disclosure sufficient to satisfy NLRA obligations.

A Genuine Unsettled Question

The Board has not yet issued a published decision squarely addressing whether a union may demand, as a mandatory subject of mid-term bargaining, ongoing access to algorithmic model updates — i.e., whether each software update that changes how an AI discipline or scheduling tool functions triggers a fresh bargaining obligation. ALJ-level decisions have gestured toward this question without resolving it. It is the most significant open issue in this area and the one most likely to generate a Board-level decision or further GC guidance in 2026 or 2027.

Employment lawyers advising clients with large unionized workforces and AI-intensive operations should treat this as an active risk and structure their vendor agreements and labor relations practices accordingly.