I disagree with a lot of what comes out of California’s legislature. Some of their laws seem to create solutions for problems that don’t necessarily exist. But every now and then, they get something right, and their recently enacted No Robo Bosses Act is one of those occasions.

In fact, I think every state should adopt something similar. The law addresses something I’ve been concerned about as more organizations put artificial intelligence into their decision-making. Just because a human signs off on a decision doesn’t mean that person actually made it.

On September 30, Governor Gavin Newsom signed SB 947, the No Robo Bosses Act. Beginning July 1, 2027, the law will prohibit California employers from relying solely on automated decision systems to discipline or terminate employees. When an employer primarily relies on one of these systems, a human must corroborate the decision using supporting information.

If the decision can’t be corroborated, or the reviewer determines that the system’s output is inaccurate, incomplete, or misleading, the employer can’t use that output to make the disciplinary or termination decision.

I agree with what the law is trying to accomplish. But requiring somebody to review a decision doesn’t necessarily mean they’re doing anything more than signing off on it.

Consider a supervisor who receives an automated report identifying an employee as a disciplinary problem. The system has already evaluated the employee, weighed whatever information it was programmed to consider, and recommended action. The supervisor reads the report, agrees with the recommendation, and signs the paperwork.

Technically, a human reviewed the decision. But what did that supervisor actually do? Did they examine the underlying information, consider what the system might have missed, and determine whether the recommended action was appropriate? Or did they simply assume that a system capable of producing a detailed report must have reached the right conclusion?

There’s a reason supervisors are supposed to review employee performance, complaints, and disciplinary recommendations before action is taken. They’re expected to exercise judgment. An automated system might identify a pattern that deserves attention, but it doesn’t necessarily understand the circumstances behind it.

An employee could have legitimate reasons for missing a performance target, or the information being evaluated could be incomplete. Those are things a supervisor needs to examine before deciding what happens next.

That doesn’t mean every automated recommendation is wrong, or that supervisors should disregard technology because it isn’t human. These systems can help identify problems that might otherwise go unnoticed. But the more useful and reliable the technology becomes, the easier it may be for the person reviewing its conclusions to stop questioning them.

California’s law requires corroboration, but it doesn’t guarantee that every reviewer will approach the job with the independence or skepticism the situation deserves. That still depends on the people responsible for making the decision and the expectations their organization places on them.

If I’m going to hold a supervisor accountable for approving a disciplinary action, I expect that supervisor to understand what they’re approving and be able to defend the decision. Telling me that the computer recommended it doesn’t answer that question. Neither does showing me that somebody signed the paperwork. The technology can help make the case. The person responsible for the decision still has to make it.