California just wrote into law something most organizations have never written down at all: when a decision about someone’s job is on the line, a human has to own it. That is the real story behind this week’s news, and it is a story about AI workplace policy far beyond one state. The question is not whether AI can help make the call. It is who stands behind the call once it is made.
In my work with leadership teams, I see plenty of organizations that have approved AI tools. Far fewer have defined what those tools are allowed to decide.
That gap is now visible in statute.
Key takeaways
- On September 30, 2026, California Governor Gavin Newsom signed a set of AI workplace laws, including SB 947, which bars employers from relying solely on automated systems to discipline or fire workers.
- The laws set a legal floor. They do not answer the leadership question underneath: where should AI help, where should it not, and who is responsible for the result?
- In The Next Turn, I argue that AI can participate in the work without inheriting responsibility for it. Organizations that define that line early avoid drift, hidden AI use and quiet distrust.
- Leaders anywhere, not just in California, can use this moment to write a plain-language AI standard for decisions that affect people.
What’s happening
According to KQED’s report on the bill signings (October 1, 2026), Newsom signed a group of labor-backed laws on September 30 aimed at protecting workers as AI spreads through the workplace. KQED notes that he had vetoed a previous version of the “robo bosses” legislation last year and signed a narrower version this time.
HR Dive’s breakdown of the new California AI protections (October 1, 2026) lays out what employers will face. Under SB 947, effective July 1, 2027, employers cannot rely solely on an automated decision system for discipline or termination. If such a system is the primary tool, a human must verify the decision using supporting documentation, such as evaluations and personnel files. Employers must notify workers when one was used, and workers can request a description of the data involved. SB 951 requires notice of mass layoffs caused “in whole or in substantial part” by AI. Two other bills, AB 1331 and AB 1883, limit workplace surveillance, including tools that collect neural data or try to detect emotional states. HR Dive also reports that California Privacy Protection Agency regulations taking effect January 1, 2027 will require pre-use notices, opt-out opportunities and risk assessments for AI used in hiring and compensation decisions.
KQED adds AB 1979, which bars health care entities from using AI to replace the clinical judgment of licensed professionals. It also notes who is not covered: rideshare drivers, who are classified as independent contractors under Proposition 22. In the bill author’s announcement of the No Robo Bosses Act (September 30, 2026), State Sen. Jerry McNerney put the intent simply: “AI must remain a tool controlled by humans, not the other way around.”
Whatever you think of the policy itself, the direction is clear. Lawmakers are starting to define AI’s role in work because many organizations have not.
Why AI workplace policy is a leadership question, not a legal one
In Chapter 2, I describe two turns happening at once. The first is the definition of work: what people believe work is for and what it should provide. The second is the structure of work: how work actually flows now that AI is built into it. This news sits squarely in the second turn. When software can rank, flag, score and recommend, the structure of a decision changes even if the org chart does not.
In Chapter 12, “Defining Work Again,” one of the seven moves is to define AI’s role in the work before employees define it informally. The central line of that chapter is the one I keep coming back to:
AI may participate in the work. It does not inherit responsibility for the work.
The Next Turn
SB 947 is, in effect, a legal version of that sentence for one category of decision. But most of the decisions that shape someone’s working life never reach discipline or termination. They live earlier. Who gets flagged as underperforming. Whose application gets screened out. Whose output gets called “low quality” by a dashboard. If AI is shaping those moments and no one has said where human judgment must stay in charge, the organization is not neutral. It is drifting.
Ambiguity is not neutral.
The Next Turn
In Chapter 8, “The Rise of Judgment,” I suggest that leaders set explicit standards for where judgment must stay human-led. Not because AI output is always wrong, but because polished output can create the illusion that judgment has already been exercised. A recommendation that looks thorough is not the same as a decision someone has thought through. A human “verifying” a decision by glancing at a score is not oversight. It is a signature.
That is where standards and signals matter. The standard is a fair, accountable, well-reasoned decision. A clean automated report is a signal. Leaders who confuse the two will meet the letter of a law like this and miss its point.
The surveillance question is a trust question
Two of the bills deal with surveillance, including tools that try to read emotional states. In Chapter 9, “Invisible Work,” I argue that as work gets harder to see, many leaders reach for more monitoring. That gives them more data about activity, not more clarity about judgment.
It is not just a workflow issue. It is a trust issue.
The Next Turn
People notice what an organization chooses to measure. If the system watches faces and keystrokes but nobody can explain how a performance decision was reached, the message to every generation on the team is the same: we track you, but we do not explain ourselves. That is a hard foundation for trust, whether you work in an office, a clinic or a warehouse.
Your next turn
You do not need to be in California, and you do not need to wait for a law, to do this work. This week, try one or more of these:
- List the people decisions AI touches. Hiring screens, scheduling, performance flags, productivity scores, promotion inputs. Most teams are surprised by how long the list is.
- Mark each one: assist, recommend or decide. Then ask where a human must own the final call and what that human needs to see to own it well.
- Define what real review means. Name the evidence a manager must look at, and expect them to explain the reasoning in their own words.
- Tell people when AI is part of a decision about them. Disclosure is not only a compliance step. It is how you keep trust from turning into suspicion.
- Check your monitoring against your purpose. For each tool, ask whether it gives you clarity about judgment or just more data about activity.
If you operate in California, talk with your employment counsel about what these laws require. I am not a lawyer, and this post is not legal advice. The leadership work, though, is the same everywhere.
Frequently asked questions
What does California’s No Robo Bosses Act require of employers?
According to HR Dive, SB 947 bars employers from relying solely on an automated decision system to discipline or fire a worker. If such a system is the primary tool, a human must verify the decision with supporting documentation, and workers must be notified. It takes effect July 1, 2027.
Should companies outside California update their AI workplace policy?
The law applies to California, but the leadership question applies everywhere. Any organization using AI in people decisions benefits from defining where AI assists, where it recommends and where a human must own the outcome. Doing that before regulation arrives is simply clearer leadership.
How can managers show meaningful human oversight of AI decisions?
Oversight means the manager can explain the decision in their own words, name the evidence they reviewed and say what they checked rather than trusted. If the only answer is “the system flagged it,” the oversight has not happened yet. Building that habit is part of manager readiness for AI.
The question underneath
Laws like these will keep coming, in more places and in different forms. Leaders can treat each one as a compliance checklist. Or they can treat it as a prompt to do what the moment already asks of them: make explicit what used to be assumed.
AI can inform the decision. Someone still has to own it.
So here is my question for you: if someone on your team asked tomorrow how AI was used in a decision about their job, could you give them a clear answer?
More on the book behind these ideas at The Next Turn.
Sources
- California revamps AI protections for workers in flurry of bill signings — HR Dive, October 1, 2026
- Newsom Signs Slate of AI Workplace Laws, Barring ‘Robo Bosses’ and Surveillance — KQED, October 1, 2026
- Newsom Signs McNerney’s No Robo Bosses Act of 2026 Requiring Human Oversight of AI in the Workplace — Office of State Sen. Jerry McNerney, September 30, 2026