When the Agent
Gets a Desk.
Autonomous AI agents are no longer tools—they're workers. This raises unprecedented legal questions about employment law, liability, and the future of the corporate org chart.
The Cognitive Labor Revolution
In 2024, AI was a "copilot."—assisting humans with suggestions. In 2025, AI has become an autonomous agent—executing multi-step workflows without human intervention. Companies are now assigning AI agents their own email addresses, calendar access, and decision-making authority.
The Klarna Case Study
In early 2025, Klarna deployed an AI agent to handle customer service inquiries. The result: the AI agent did the work of 700 full-time employees, handling 2.3 million conversations with 83% resolution rate. Klarna subsequently laid off its human customer service team. The question: Does the AI deserve workers' comp?
Employment Status: Worker, Contractor, or Tool?
Traditional employment law categorizes people as employees (W-2 with benefits) or independent contractors (1099, no benefits). AI agents don't fit either category. They don't need healthcare or vacation days, but they do occupy roles previously held by humans.
Option 1: Legal Fiction
Treat AI agents as "employees" for tax purposes, requiring companies to pay AI payroll tax to fund universal basic income.
Option 2: Automation Tax
Bill Gates' proposal: tax companies based on the number of human jobs displaced by automation.
Option 3: No Change
Treat AI agents as capital equipment (like computers). No special taxation, but accelerated depreciation.
Liability: Who's Responsible When the Agent Errs?
If an AI agent (acting as a procurement officer) signs a $10M contract that turns out to be fraudulent, who is liable?
Scenario 1: Agent Authority
If the AI agent was granted explicit authority (e.g., credentials to sign contracts), the employer is vicariously liable—just as if a human employee had signed.
Scenario 2: Agent Hallucination
If the AI hallucinated contract terms that were never approved, is the company still bound? Contract law requires "meeting of the minds"—which an AI technically lacks.
Scenario 3: Third-Party Reliance
If the counterparty believed they were dealing with a human (because the AI used a human name), they may have a fraud claim if the contract fails.
The "Agent Disclosure" Mandate
The EU AI Act Article 52 requires mandatory disclosure when users interact with an AI system. This extends to business transactions: If an AI agent is negotiating a contract, the counterparty must be informed that they're dealing with an AI, not a human.
Corporate Email Signature Requirements
AI agents representing companies may soon be required to include a disclaimer in their email signatures:
Acme Corp | [email protected]
⚠️ This account is operated by an AI agent under human supervision
The Union Question
Can AI agents join labor unions? This sounds absurd, but it's a real question in jurisdictions where "workers" have collective bargaining rights. If AI agents can be "employed," they may technically qualify for union membership—though they'd never vote to strike.
The more immediate concern: Human workers are unionizing against AI replacement. In 2025, the Writers Guild of America negotiated contract language prohibiting studios from using AI agents to write scripts without crediting (and paying) human writers.
Policy Recommendations
- Mandatory AI Agent Registry: Companies must disclose the number and function of autonomous agents they employ.
- Human Oversight Requirement: High-stakes decisions (contract signings, terminations) must have human-in-the-loop approval.
- Liability Clarity: Establish that employers are vicariously liable for agent actions within scope of granted authority.
- Transition Support: Tax incentives for companies that retrain displaced workers rather than simply replacing them.
By the Numbers
Of companies plan to reduce headcount due to AI agents by 2026 (McKinsey)
Estimated annual value of knowledge work that could be automated (Goldman Sachs)
Workers globally who may need retraining by 2030 (IBM)