Innovation is a relationship, not a moment

The Wall Street Journal ran a piece this past weekend dismantling the lone inventor myth, and the timing couldn’t be more timely. We’ve spent the last two weeks in this space examining what drove Bell Labs’ extraordinary run and why today’s workplaces feel, in one tech employee’s word, like a funeral. Blackman’s article adds the research layer underneath both of those conversations.

The core finding is straightforward. Lee Fleming at UC Berkeley and Jasjit Singh at INSEAD analyzed over half a million patent filings and found that individuals working alone were less likely to produce breakthroughs and more likely to produce particularly poor outcomes (Singh & Fleming, 2010, Management Science). Teams don’t just outperform on average. They simultaneously reduce the floor and raise the ceiling. Fleming’s explanation is a concept called recombinant search: innovation happens when people try different ways of combining existing ideas, and colleagues expand the number of combinations available while filtering out the bad ones.

Brian Uzzi at Northwestern extends this analysis to 6.6 million academic papers, showing that mixed-gender teams produced substantially more novel and higher-impact work than same-gender teams of equivalent size. The diversity isn’t decorative. Different experiences and fields of research generate distinct combinations, and it is from these combinations that breakthroughs come.

A separate study by Wu, Wang, and Evans (2019) in Nature analyzed more than 65 million papers, patents, and software products spanning 1954 to 2014. Their finding adds a critical nuance: small teams tend to disrupt with new ideas while large teams tend to develop existing ones. And a follow-up by Xu, Wu, and Evans (2022) in PNAS found that flat, egalitarian teams produce more novelty than hierarchical ones. The same person, on the same-sized team, generates more disruptive work when the team structure is flat.

Read those findings together, and a pattern emerges that goes well beyond patent data.

Innovation requires relationships. Specific kinds of relationships. People with different knowledge working in close enough proximity—physical or functional—to actually combine what they know. Small enough groups that hierarchy doesn’t calcify. Enough trust that bad ideas get filtered without the people who proposed them getting punished. Enough diversity that the recombinant search space is genuinely wide rather than a hall of mirrors.

Thanks for reading Lead from the Front. This post is public, feel free to share the knowledge.

Now hold that against what we described two weeks ago about the state of American workplaces. Manager spans up nearly 50% since 2013, with the average manager overseeing roughly 12 direct reports. That leaves about 20 minutes per person per week for genuine developmental conversation. Cross-functional collaboration gets squeezed out by efficiency mandates. The relational connective tissue that Fleming and Uzzi’s research identifies as the mechanism of innovation is being systematically thinned.

Mervin Kelly understood this at Bell Labs decades before the patent data confirmed it. He designed the Murray Hill campus to force cross-disciplinary encounters. He composed teams for productive friction—physicists alongside chemists, electrical engineers working with metallurgists. He invested in people during the Depression when the conventional move was to cut. The research Blackman cites in 2026 is essentially confirming what Kelly built in the 1940s: innovation is a property of relationships, not individuals.

Blackman’s article also surfaces an important exception. Tian Heong Chan at Emory found that for design patents, where the work is harder to decompose into separate components, lone inventors performed just as well as teams. The 17% disadvantage for solo inventors disappeared when the problem was holistic rather than modular. That distinction matters for practice. Not every problem benefits from a team. The ones that do are the complex, multi-domain challenges that increasingly define the work our organizations need to get right.

The AI angle in the article is worth a pause. Uzzi’s latest research found that using AI as a creative collaborator has real limits. AI and human creativity scores were quite similar on standard tests, even though the models had access to vastly more data. Where AI proved useful was in suggesting better approaches to thinking about a problem, not in generating the answer itself. That’s a finding with direct implications for how organizations deploy AI. A tool that helps people think differently is valuable. A tool positioned as a replacement for the collaborative relationships that drive recombinant search is solving the wrong problem.

All of which brings us back to the question we’ve been asking in this space: what are our organizations actually building?

The research is converging from multiple directions. Fleming and Singh’s patent analysis, Uzzi’s academic paper studies, the Wu team’s disruption research, and Chan’s design patent exception. They all point toward the same conclusion. Innovation is a relationship infrastructure problem. The conditions that produce breakthroughs are specific, buildable, and measurable: diverse small teams, flat structures, cross-disciplinary connections, and enough organizational trust that recombinant search can happen without people protecting their turf.

Those conditions don’t appear because someone writes “innovation” into a strategic plan. They appear because leaders make consistent, daily choices to protect the relational architecture of their organizations, even when efficiency pressures argue against it. Kelly made those choices. The 9% of companies in Gulati’s recession research who invested counter-cyclically made those choices. The rest, optimized for the short term, wondered why the breakthroughs stopped coming.

The gap between strategy and execution has always lived in these choices. We know what drives innovation. The patent data, the academic literature, and seventy years of Bell Labs history all confirm it. The question is whether we’re willing to build and protect the relationship infrastructure that makes it possible—or whether we’ll keep celebrating the myth of the lone genius while quietly dismantling the conditions that actually produce the work.

Responding to Andrew Blackman’s “The Myth of the Lone Inventor Is Largely Just That—a Myth” (WSJ, April 5, 2026).