The Baker Library of the Harvard Business School on the Harvard University campus in Boston, Massachusetts, US, on Tuesday, May 27, 2025. Recent research conducted by the Digital Data Design Institute at Harvard Business School is investigating where AI is most effective in increasing productivity and performance — and where humans still have the upper hand.
Bloomberg | Bloomberg | Getty Images
Workplace AI adoption is at an all-time high, according to Anthropic data, but just because organizations use AI doesn’t mean it’s effective.
“Nobody knows those answers, even though a lot of people are saying they do,” said Jen Stave, chief operator at the Digital Data Design Institute (D^3) at Harvard Business School. While much of the business world tries to figure out where AI can be best deployed, the team at D^3 is researching where the technology is most effective in increasing productivity and performance — and where humans still have the upper hand.
Workplace collaboration is a long-held standard for innovation and productivity, but AI is changing what that looks like. AI-equipped individuals perform at comparable levels to teams without access to AI, D^3’s recent research in partnership with Procter & Gamble finds. “AI is capable of reproducing certain benefits typically gained through human collaboration, potentially revolutionizing how organizations structure their teams and allocate resources,” according to the research.
Think AI-enabled teams, not just AI-equipped individuals.
While AI-equipped individuals show significant improvement in factors like speed and performance, strategically curated teams with AI have their own advantages. When factoring in the quality of outcomes, the best, most innovative solutions come from AI-enabled teams. This research relies on AI tools not optimized for collaboration, but AI systems purpose-built for collaboration could further enhance these benefits. In other words, simply replacing humans with AI may not be the fix businesses hope for.
“Companies that are actually thinking through the changes in roles and where we need to not just lean into it but protect human jobs and maybe even add some in that space if that’s our competitive advantage, that, to me, is a signal of a super mature mindset around AI,” Stave said.
The D^3 experiment at P&G also shows that AI integration significantly reduces gaps that exist between an organization’s pockets of domain expertise. For example, having a knowledge base at hand could make any one team’s outputs more universally beneficial beyond sole teams like human resources, engineering and research and development.
Lower-level workers benefit more, but it is a double-edged sword.
Another experiment D^3 conducted with Boston Consulting Group showed AI leads to more homogenized results. “Humans have more diverse ideas, and people who use AI tend to produce more similar ideas,” Stave said, recognizing that companies with goals of standing out in the market should lean into human-led creativity.
Performers on the lower half of the skill spectrum exhibit the biggest performance gains (43%) when equipped with AI compared to performers on the top half of the skill spectrum (who get a 17% performance surge). While both outcomes are substantial, it’s the entry-level workers who get the biggest perks.
But for the less-skilled workers, it’s a double-edged sword. For instance, if AI can do junior work better, the senior-level workplace might stop delegating work to their junior counterparts, creating training deficits that negatively impact future performance. Bearing a company’s future in mind, businesses will want to carefully consider what they do and don’t delegate.
Human managers are not prepared to oversee AI agents. They need to learn
While Stave says humans serving as managers to a suite of AI agents is “absolutely going to happen,” the scaffolding to do so both effectively and with minimal adverse harm is simply not there. Stave herself has had this experience, and it contrasted with all her managerial and leadership education. “You learn how to manage according to empathy and understanding, how to make the most of human potential,” she said. “I had all these AI agents that I was personally trying to build and manage. It was a fundamentally different experience.”
Moreover, while Grammarly CEO Shishir Mehrotra said entry-level workers could be the new managers (with AI agents — not people — in their charge), the junior workforce has not actually proven to be enterprise AI-native or managerially equipped. “We want to see AI giving humans more opportunity to flourish. The challenge I have is with assuming that the junior employees are going to step in and know how to do that right away,” Stave said.
She added that the companies truly getting value from their AI deployments are the ones undertaking process redesign. Instead of relying on AI notetaking to save time, lean into where AI helps and where humans are the winners. “It’s very easy to buy a tool and implement it,” she said. “It’s really hard to actually do org redesign, because that’s when you get into all these internal empires and power struggles.”
But even so, she says, the effort is worth it.
