An AI President Would Still Need to Answer to Workers
I remain unconvinced that AI would make better presidents. Oakland's negotiations suggest a more useful test: whether people can influence objectives, challenge decisions and identify who must answer for them.

On November 12, the Guardian reported that Oakland's tentative labor agreement would require retraining or reassignment for workers displaced by AI. [1] I think that development offers a sharper test of an AI president than a promise of superior intelligence: would the people affected have power over its decisions?
I remain unconvinced that replacing a human leader with AI would improve workers' prospects. My standard is whether people can influence the goals, challenge harmful decisions and identify an authority that must answer for the consequences.
Oakland negotiated a choice
Oakland's October 31 announcement confirms a tentative three-year agreement with IFPTE Local 21 that includes AI protections. Member ratification and City Council approval were still ahead when that announcement appeared. The city's notice does not reproduce the specific retraining and reassignment provisions described by the Guardian. [2] [1]
I take that bargaining gain seriously. But neither account establishes completed retraining, successful placements or how a disputed case would be resolved. Announced protections cannot yet demonstrate jobs saved. [1] [2]
What interests me is the choice being negotiated. If technology reduces the labor needed for a task, an employer could cut positions, change responsibilities or expand services. Deciding among those options involves priorities about costs, service quality and people's livelihoods. Greater analytical ability cannot make those competing priorities disappear.
A system instructed to minimize payroll could perform its assignment brilliantly while making workers' lives worse. That is a hypothetical example, but it exposes my objection to assuming that intelligence produces good leadership. Workers need influence over the assignment itself.
I see Oakland's negotiations as a concrete attempt to secure that influence. An AI leader would need to operate within comparable constraints if its decisions were to deserve public trust.
Authority cannot hide inside software
Albania's Constitutional Court addressed AI and authority in an October 8 decision notice concerning challenged presidential decrees. It stated that technology cannot alter the constitutional and legal allocation of public powers, responsibility and accountability, or create or transfer those powers. [3]
The majority dismissed the application after examining the available constitutional review routes. This was a decision notice, with a fuller reasoned judgment to follow. [3] I would not read the dismissal as an endorsement of autonomous AI government.
From that Albanian notice, I draw a practical leadership principle: using software must leave someone answerable for its consequences. An employee needs an identifiable authority able to explain a decision, review evidence and change an outcome.
A human signature would offer little reassurance if the signer lacked the time, information or authority to disagree with the system. I would judge oversight by the reviewer's ability to intervene. Otherwise, responsibility could become ceremonial while the consequences remain very real.
An appeal must be able to change the answer
The UC Berkeley Labor Center's bossware summary draws on existing research and 25 interviews with California workers. Interviewees described discipline involving disputed driving infractions, scanning errors and delays outside their control. The authors also describe barriers to reviewing data and challenging automated decisions. [4]
These interviews illustrate how disputes arise; they cannot establish how frequently those problems occur. Bossware also includes electronic monitoring and algorithmic management beyond generative AI. The Guardian draws on this research too, so its coverage is not an independent replication. [4] [1]
I think these accounts expose a weakness in imagining that an apparently impartial AI would automatically govern fairly. A decision can apply a rule consistently while relying on incomplete information. Even an accurate account of events leaves room to disagree about whether the rule is reasonable.
Consider a hypothetical public employee who completes fewer cases because the cases assigned are harder. A system might count completions accurately and still produce a misleading assessment. Meaningful review would let the employee challenge both the information and the decision to use that measurement.
That distinction matters for careers. In my view, improving someone's skills is an inadequate answer when the obstacle is an assessment they cannot inspect or correct. Competence offers limited protection if the institution refuses to hear evidence of it.
Better analysis still leaves political choices
I see a plausible role for AI in comparing staffing plans, revealing assumptions and helping officials examine more alternatives. Those are possible benefits, not improvements demonstrated by these sources. I would welcome experiments that made public decisions easier to inspect.
Even if an AI avoided a particular official's favoritism or vanity, someone would still have to decide how much weight to give employment, expenditure and service quality. I would want those priorities visible and open to revision. Hiding them behind a technical recommendation would make government less accountable.
Workers' preferences cannot settle every public decision either. Leadership must also account for service users and people hoping to enter public employment. A proposal preserving existing positions may carry costs for those groups. I would expect those tradeoffs to be explained and debated, with affected people able to participate.
That is why I find the claim that AI would make better presidents premature. Analytical competence could improve government. It does not settle who should have authority or whose interests should prevail.
Accountability belongs in career planning
For someone comparing public employers, I would treat the ability to correct a damaging assessment as part of job quality. Look for a written process identifying who reviews contested decisions, what evidence employees can submit and whether the reviewer can reverse an outcome.
That is a specific feature to investigate alongside the work itself. A general promise of responsible AI tells me little about what happens when an employee and a system disagree. A documented process gives an applicant something concrete to evaluate, although it cannot guarantee fair treatment.
My prognosis is cautiously hopeful about negotiated protections and skeptical of promises of benevolent AI rulers. Oakland shows that AI safeguards can reach a tentative bargaining agreement; approval, implementation and employment outcomes still require evidence. [1] [2]
The danger I foresee is a government retaining its power while attributing unpopular choices to a machine. The better possibility would use AI to make decisions more transparent and easier to challenge. I would trust that government more because people could hold it responsible.
The sources behind this perspective
- US public workers are fighting to protect their jobs from AI – and some are winning | AI (artificial intelligence) | The Guardian
www.theguardian.com · Nov 12, 2026
- Oakland Reaches Tentative 3-Year Agreement with IFPTE Local 21 | City of Oakland, CA
www.oaklandca.gov · Read on Nov 13, 2026
- NJOFTIM VENDIMMARRJEJE – GJYKATA KUSHTETUESE
www.gjykatakushtetuese.gov.al · Read on Nov 13, 2026
- Bossware Comes to California: The Impact of New Management Technologies on Workers and the Case for Policy Regulation - UC Berkeley Labor Center
laborcenter.berkeley.edu · Read on Nov 13, 2026