California draws a line on automated workplace decisions
Artificial intelligence is becoming part of the machinery behind hiring, performance reviews, restructuring and layoffs. California has now moved to limit one of the most consequential uses: allowing software to effectively decide whether someone keeps a job.
Governor Gavin Newsom recently signed SB 947, referred to as the “No Robo Bosses Act.” The law prohibits employers from fully outsourcing disciplinary and termination decisions to automated decision systems. It takes effect July 1, 2027.
The policy does not ban employers from using AI in workplace processes. Instead, it establishes a threshold that matters in practice: an automated output cannot be the final authority for discipline or dismissal. Employers must review AI-driven recommendations, and workers must receive a description of the reasons behind a disciplinary or termination decision.
That explanation includes the data used by the system, such as personnel files, evaluations, work product, peer reviews and witness interviews. If an employer cannot corroborate the automated output, or if a human reviewer finds it inaccurate, incomplete or misleading, the employer may not use it to make a disciplinary or termination decision.
Why the legislation matters
Employment decisions are already being shaped by algorithmic tools. A survey of 1,000 HR professionals conducted by MyPerfectResume found that 73% said they use AI to make hiring decisions. Fifty-two percent said AI informs restructuring and role planning, while 51% said they use it to flag candidates considered risky. Only 26% said they do not use AI to make layoff decisions.
Those figures do not show that every decision is fully automated. They do show that AI is moving beyond routine back-office assistance and into systems that can determine who receives an opportunity, a warning or a termination notice.
That shift creates a difficult accountability question. AI systems can process large volumes of information, but their outputs depend on the data they receive and the assumptions built into their design. A recommendation may appear objective because it arrives as a score, ranking or flag. Yet it can still be incomplete, misleading or unsupported by the facts of a specific employee’s work.
SB 947 addresses that risk by making human review more than a ceremonial sign-off. The law’s standard is tied to corroboration. A reviewer cannot simply accept an automated recommendation when it cannot be supported, or when the reviewer concludes that the system’s output is wrong or materially incomplete.
A practical change for employers
For employers, the measure creates a need to examine how automated tools are used across the employee lifecycle. The most obvious systems include tools that score performance, identify conduct risks, recommend discipline, rank workers during restructurings or flag employees for possible dismissal.
Organizations using such tools will need a process for documenting the non-automated basis for a decision. They will also need to identify the information that influenced the automated system and be prepared to describe it to an affected worker.
The law’s focus on data sources is significant. Personnel files and evaluations may reflect formal workplace records, while work product, peer reviews and witness interviews can be more contextual and subjective. An automated system may combine those inputs in ways that are difficult for an employee to see or challenge without disclosure.
Providing reasons and identifying relevant inputs does not settle every question about fairness. But it gives an employee and a human reviewer a clearer starting point for examining whether a recommendation was grounded in the individual’s actual circumstances.
A model other jurisdictions may watch
Regulation of AI in employment remains largely a state-level project in the United States. California’s action could nevertheless influence the debate elsewhere because it focuses on a straightforward principle: systems may assist people, but they should not independently make the ultimate decision to discipline or fire them.
The approach is narrower than a sweeping prohibition on workplace AI. It recognizes that employers may use automated systems while reserving the final judgment for a person who can assess context, verify the evidence and reject an unreliable result.
That distinction may become more important as AI tools are embedded more deeply into human-resources platforms. The technology is often marketed as a way to make workplace decisions faster and more consistent. California’s new law emphasizes that speed and scale do not remove the need for accountable judgment when a person’s livelihood is at stake.
What changes in 2027
When SB 947 takes effect, the central question for employers will not simply be whether AI was involved. It will be whether an actual human review occurred, whether the automated recommendation could be corroborated and whether the employee was given a meaningful description of the decision’s basis.
For workers, the law offers a protection against being subjected to a purely automated disciplinary or termination outcome. For employers and technology providers, it raises the bar for how workplace AI must be governed. And for policymakers, it offers an early test of whether enforceable human oversight can keep pace with the expanding role of automated decision systems at work.
This article is based on reporting by ZDNET. Read the original article.
Originally published on zdnet.com
