AI Multiplies Output, But Labor Law Hasn't Caught Up

The labor policy framework for American workplaces rests on a basic assumption that no longer holds: human output scales with headcount. AI has broken that link, and regulators haven't noticed.

Nate B Jones, an AI strategy consultant, argues that AI-boosted workers now produce output that once required whole departments. He points to companies like Midjourney -- which reportedly runs with about 100 employees while serving millions of users worldwide. His claim: the old link between team size and output has snapped. But these revenue-per-employee figures need checking. Midjourney doesn't publish headcount or revenue. Neither does Lovable, another often-cited example. The actual productivity boost remains uncertain.

What we can verify: A Harvard Business School study published in 2025 tracked 776 professionals at Proctor & Gamble working on innovation challenges. Teams using AI were three times more likely to produce ideas in the top 10% of quality. Not three times more volume -- three times higher odds of getting it right. That study tested idea generation, not sustained output. But it hints that AI's effect on knowledge work may be about quality as much as quantity.

The policy effects of even a modest productivity rise are real. If individual output climbs while coordination costs stay flat or grow, the incentive to keep teams small gets stronger. Jones frames this as a chance for mission growth rather than headcount cuts. But labor markets rarely work that way. Companies aim for margins, not ambition.

The Regulatory Vacuum

Current employment law wasn't built for five-person "strike teams" pulling in millions per head. The Fair Labor Standards Act sorts workers as exempt or non-exempt based on job duties and pay levels set for 20th-century roles. The National Labor Relations Act protects bargaining in units shaped by old-style supervisor setups. Workers' comp, unemployment insurance, anti-bias rules -- all of these systems expect a clear structure with managers, direct reports, and divisions.

Jones describes a future where "three to four strike teams share a domain coordinated by a single person focused on inter-team coherence." That person isn't a standard manager. They don't assign tasks or run reviews in any normal sense. They maintain what he calls "the quality of relationship required for coordination." Try fitting that role into FLSA exemption buckets.

The misclassification risk alone should worry any general counsel. If AI lets one person do the work of five, and you restructure around that, have you created five new contractor roles? Or one employee with four AI tools? Employment status drives tax withholding, benefits, and liability. The IRS common-law test looks at behavioral control, financial control, and relationship type. None of those factors considered AI tools.

European regulators are at least raising these questions. The EU's proposed AI Act includes rules on algorithmic management and worker monitoring. It doesn't directly address team structure, though. The UK's Employment Rights Bill, now in Parliament, tries to extend protections to platform economy roles that blur old categories. Neither framework tackles the structural changes Jones describes.

What Coordination Costs Actually Mean

Jones cites Dunbar's number -- the cognitive cap on stable relationships -- to argue that five-person teams maximize "correctness." Larger teams drown in coordination overhead. The U.S. military builds infantry around fire teams of four to five soldiers for the same reason. Communication paths scale fast with group size. Five people create 10 relationship lines. Twenty people create 190.

This isn't a metaphor. It's math, and it has legal weight. Every communication path means potential liability. Every decision point creates record-keeping duties. Every coordination meeting produces material that could surface in future lawsuits. If AI multiplies output per person, it also multiplies the compliance burden per person.

Consider the Americans with Disabilities Act and its requirement for reasonable accommodation. In a 20-person team, adjusting one employee's schedule or remote setup spreads the load across many colleagues. In a five-person strike team where each member holds one of 10 key communication links, that same adjustment may weigh much more heavily on the other four. The legal duty hasn't changed. But the practical reality has.

Jones quotes Toby Lutke of Shopify describing a "10x loss of productivity with each addition beyond five." If that's true -- and Shopify hasn't shared data backing the exact claim -- it means growing a team past the sweet spot doesn't just slow things down. It creates a measurable harm that could fuel shareholder suits alleging waste.

The Weak Link Problem

The most concerning policy angle in Jones's framework is what he calls "the AI slop tax." This is the outsized cost of a mediocre performer on a small, AI-powered team. He argues that in a team of five, where AI amplifies each person's judgment, "a mediocre contributor doesn't just underperform, they consume a coordination slot without providing the judgment that justifies their cost."

This is employment discrimination litigation waiting to happen. If AI makes individual performance matter more and team tolerance for gaps shrinks, the push to remove underperformers grows. That pressure peaks right when the legal risks of firing someone are highest. Protected class members already face extra scrutiny in reviews. Add AI-driven productivity metrics that measure each person's contribution with new precision. You've built a recipe for disparate impact claims.

The Age Discrimination in Employment Act becomes especially relevant. If older workers adopt AI tools more slowly, and if team results depend on all five members running at AI-augmented speed, age becomes a stand-in for the "correctness" Jones prizes. Employers can defend performance-based firings. But they need consistent, documented standards applied evenly. Small teams make that documentation harder because the sample size is smaller.

What Policymakers Aren't Debating

Congress right now focuses on AI safety, algorithmic bias, and content moderation. The White House Executive Order on AI covers discrimination in hiring and federal procurement. None of this attention lands on how organizations are structured.

The questions that matter: If AI drives radical productivity gains, does employment law need to tell human output apart from AI-boosted output for wage and hour rules? Should bargaining units be redrawn around coordination costs rather than job functions? Do fiduciary duties to shareholders require restructuring around optimal team size -- even when that triggers WARN Act obligations?

Jones argues that companies should "keep your people, restructure how they work together, and go after something bigger." That's the humane path, and maybe the smart one. But labor markets don't usually produce that result without guardrails. The policy void around AI-driven restructuring will get filled. Either by thoughtful legislation, or by years of lawsuits setting precedent case by case.

Right now, we're headed for the lawsuits. Companies are reshaping around AI without legal clarity on classification, liability, or duties. Workers face productivity demands set for AI-augmented output but lack protection against unfair standards. Regulators draft AI bills that focus on model training and deployment while ignoring the organizational shifts those models drive.

The five-person strike team may be the optimal structure. Whether it's legally viable under current employment law is a separate question. It deserves far more attention from policymakers who think AI regulation means governing algorithms -- rather than governing how those algorithms reshape work itself.

Samira Okonkwo-Barnes covers technology policy and regulation for Buzzrag.