AI can replace teams in US offices, but here’s why Harvard says that would be a mistake

On a quiet spring morning in Boston, sunlight filtered through the tall windows of Baker Library at Harvard Business School. The building, long associated with case studies on management, labour and power, has become an unlikely symbol of a new question confronting the US workforce. On May 27, 2025, researchers inside the Harvard University campus were not debating whether artificial intelligence should be used at work. That question, by and large, has already been answered. They were asking something harder: where does AI actually help, and where do humans still do better?

Recent research conducted by the Digital Data Design Institute at Harvard Business School is probing exactly that divide. Its work is investigating where AI most effectively increases productivity and performance and where human judgment, creativity, and collaboration continue to outperform machines.

The findings echo a growing body of evidence reshaping how American companies think about work.

When one worker starts to resemble a team

Across US workplaces, AI adoption has surged. Data from Anthropic shows that workplace AI use is at an all-time high, particularly in white-collar roles. The early promise has been speed. Drafts appear faster. Analysis takes minutes, not days. One person can now complete tasks once assigned to entire teams.

That promise was put to the test in recent research by D³ (Data-Driven Decision-Making), conducted in partnership with Procter & Gamble. The study found that AI-equipped individuals performed at levels comparable to teams that did not have access to AI.

For American firms under pressure to cut costs and increase output, the implication is obvious. If one AI-enabled employee can replicate the output of a group, the logic of team structures begins to wobble.

The researchers themselves did not shy away from the conclusion. AI, they argued, can reproduce certain benefits traditionally gained through human collaboration, forcing organisations to rethink how teams are built and how resources are allocated.

Why collaboration refuses to disappear

Yet the story does not end there. While AI-equipped individuals gained speed and baseline performance, the most innovative and highest-quality solutions emerged elsewhere.

The D³–P&G experiment showed that strategically designed, AI-enabled teams consistently produced better outcomes than individuals working alone with AI. Notably, the tools used in the study were not optimised for collaboration. Researchers suggest that AI systems purpose-built for team settings could amplify these gains even further.

This distinction mirrors debates taking shape at Harvard Business School. Scholars at the Digital Data Design Institute have stressed that productivity is not the same as value creation. AI can accelerate execution, but it does not automatically generate insight.

In practice, the research suggests, AI works best when it augments teams rather than replaces them.

AI as a great leveller inside organisations

Another finding carries particular weight for large US corporations. AI integration sharply reduces performance gaps between departments and roles.

In the D³ study, access to AI-driven knowledge bases allowed teams beyond research and development or engineering to produce more universally useful outputs. Expertise travelled faster. Silos weakened.

For American organisations long criticised for internal fragmentation, AI offers something close to an equaliser. Knowledge no longer sits with a few specialists. It becomes accessible, searchable and reusable across the enterprise.

But this levelling effect comes with trade-offs.

The entry-level advantage, and its hidden cost

In a separate D³ experiment conducted with Boston Consulting Group, researchers uncovered a striking imbalance. Lower-skilled and early-career workers benefited far more from AI than top performers.

Participants in the lower half of the skill spectrum saw performance gains of 43 percent when equipped with AI. Those in the top half improved by 17 percent. Both figures are substantial. But the asymmetry matters.

The same experiment also found that AI use leads to more homogenised outputs. For US companies competing on originality, sameness is a strategic risk.

More worrying is what this means for training. If AI can do junior-level work better and faster, senior staff may stop delegating foundational tasks. The apprenticeship model that underpins leadership development in American companies quietly erodes.

Harvard researchers studying workforce development have flagged this as a long-term vulnerability. Productivity gains today may weaken talent pipelines tomorrow.

Delegation becomes the real decision

The evidence emerging from Harvard, Procter & Gamble and Boston Consulting Group points to a shared conclusion. AI is not a substitute for people. It is a force that reshapes how people work together.

US employers now face a more complex challenge than simple adoption. They must decide what to automate, what to protect, and where humans still need space to struggle, learn, and create.

Back inside Baker Library, that question feels less abstract. The future of work will not be defined by how much AI companies use, but by where they choose not to use it.

The most resilient American organisations may be those that treat AI not as a replacement for collaboration, but as a tool that makes human teamwork more deliberate, not less.Ready to navigate global policies? Secure your overseas future. Get expert guidance now!