A creative and productive company is, at its core, a socially healthy company. We are standing at the edge of a new era of efficiency powered by Artificial Intelligence, and while the easy path is to use these tools to optimize the individual, our greatest opportunity lies in using AI to supercharge the collective. We can foster stronger, more creative teams than ever before, but it requires a conscious, strategic choice. The future is not one of hyper-efficient, isolated spokes, but a world where thoughtfully designed AI tools serve as a “Serendipity Engine,” actively promoting the highest quality human-to-human interactions and ensuring that the most efficient companies are also the most connected.
Most often we think of the companies as sets of boxes, hierarchies with the CEO at the top, working its way down through layers of leaders, managers, and individual contributors. That organizational chart helps us see how decisions are structured, but it is only one way to understand how a company works. Like a human body, a company has many systems that overlap and interact to make it run. There’s the financial view, organized as cost and revenue centers that describe how money flows. There’s also the project view, which shows how cross-functional teams are organized to execute initiatives. But the most elusive system, and arguably the most predictive of a company’s performance, is its social structure [1].
This structure isn’t formalized anywhere; it’s not an official hierarchy, nor can it be easily documented or managed. It is simply who you talk to, rely on, build friendships with, and turn to when you have questions or need advice. Tiffany McDowell, in ‘Strategies for Organization Design,’ demonstrates that while we focus 100% of our design efforts on the formal hierarchy, that hierarchy only accounts for about 20% of how work actually gets done. The other 80% of value, including critical decision-making and information flow, occurs in the informal network [2]. Leading People Analytics teams invest tremendous effort trying to understand this structure through what’s called an organizational network analysis. These analyses reveal that the most important people in a company are the ones that promote serendipitous interactions between colleagues. They are the super-connectors who bring people together, pass information, and act as hubs. Sometimes they are a leader, but sometimes they are Debbie from finance, who keeps a candy dish on her desk. People with the weakest connections to this social structure are the most likely to feel negatively about their company, be seen as lower-performing, and ultimately leave. That may sound obvious, but it can often be hard to identify who those people are.
This social fabric is so delicate that the layout of an office building can impact the creativity of its teams. In the 1970s, Thomas Allen developed what is now known as the “Allen Curve,” showing that communication drops off with physical distance [3]. He showed that putting teams on separate floors of the same building could harm their collaboration as much as putting them in different cities. His work led to corporate campuses designed with atriums and open spaces to foster serendipitous “water cooler” moments meant to spark new ideas. Now, in a world of hybrid work and powerful AI, we are experiencing an emerging new kind of campus that lives partially in the cloud. Arena, Hines, and Golden characterized this new “campus” using organizational network analysis and showed that, like Allen’s architectural recommendations, companies could identify and address the stifling effects of employee isolation [3].
The most creative and productive teams work in environments where they not only share information efficiently but also have strong social interactions. This isn’t just about feeling like you belong; random, serendipitous interactions also fuel creativity. You may not realize it, but the casual conversations in the cafeteria or the comments you overhear in the hallway accumulate, resulting in those light-bulb moments hours later when you’re in the shower. As much as work-from-home proponents point out that people are more efficient at specific tasks without the “distractions” of a busy office, there is good evidence to show that if you need your team to be creative, there’s no replacement for whiteboards and watercoolers. The more we talk to each other face-to-face, the better we coordinate and create [3].
So what does it mean for a company when employees discover the convenience of large language models? When they realize they can get ideas, opinions, and help drafting important emails much more efficiently from an AI than from any of their colleagues? When, without thinking, they turn to an AI with their question instead of a person? When the bulk of an employee’s interactions are already online, what happens to the social fabric of a company when Slack feels cluttered and confusing compared to asking an AI for help?
The Hub-and-Spoke Trap
As an enterprise AI becomes more efficient and appealing, it will naturally centralize information flow. This isn’t a new phenomenon but the radical acceleration of a trend that began with tools like Google Search. For decades, organizations functioned on a “transactive memory” system—a collective awareness of “who knows what.” As researcher Daniel Wegner described, to get an answer, you had to interact with a person, reinforcing the social fabric with every query [4]. The rise of search technology introduced a phenomenon known as the “Google Effect”, in which search engines began to replace humans in our transactive memory [5]. Instead of asking a colleague, we started searching keywords. This shift, while efficient, caused a subtle degradation of the interpersonal pathways that formed our collaborative culture. LLMs are only accelerating this evolution—capable of not just recalling information but synthesizing it, making the bypass of human colleagues almost irresistible.
To understand the potential impact of an LLM on how we experience work we can simulate workplace collaboration using an agent-based model. In this simulation each employee can choose between talking to a person or an AI. Each time they make a choice we can look at the strength of their connection to all the other employees in the simulation. In this way we can explore the competition between the pull of AI Efficiency (the quality of the information it provides) and Human-to-Human Synergy (the value of the human interactions). What becomes immediately apparent is that if the perceived value from interacting with other humans does not outweigh the value of getting a quick answer, a tipping point is reached where people are more likely to turn to the AI than a peer.
This simulation visualizes the “substitution effect” of Generative AI on workplace topology. The black line tracks the density of peer-to-peer connections (strong ties) as the AI tool becomes more efficient.
Stage 1 (Distributed Mesh): When AI is less convenient than asking a colleague (<0.8), the organization remains a resilient, interconnected mesh.
The Tipping Point: Once the AI crosses a critical threshold of convenience, a feedback loop is triggered. Employees bypass colleagues for the bot, causing social bonds to decay. This raises the “social friction” of future interactions, making the AI even more attractive.
Stage 2 (Centralized Hub): The network collapses into a “Hub-and-Spoke” structure. While individual productivity may rise, the organization loses the lateral connectivity required for complex, tacit problem-solving.
Simulation based on a stochastic agent-based model (N=400). The “Tacit Floor” represents the minimum viable network required for non-codifiable tasks.
For fully distributed or remote teams, where the primary interaction with colleagues is through digital channels and focused on information exchange, this tipping point could represent a social collapse. The organizational network of such a company would look less like a network and more like a hub with spokes, where the AI is the hub and each employee sits alone on their spoke. Based on Allen’s model, this represents disaster. Imagine an office building with exactly one employee per floor. Each person might be highly efficient with very few distractions, but their opportunity to serendipitously bump into a colleague is precisely zero.
Mark Granovetter’s work on the “strength of weak ties“ showed, novel information and opportunities often come from infrequent, arms-length connections—not our close collaborators [6]. Said differently, highly creative companies are highly connected companies. Conversely, the severing of those weak ties within a company would result in a significant drop in creative output. A highly effective AI could compromise weak ties by soaking up a critical mass of employee queries. This creative collapse isn’t without parallels and could be described as a sort of corporate “filter bubble,” a term coined by Eli Pariser [7]. Without meaningful human-to-human interactions, a highly effective LLM would show employees what is most relevant to their questions and their current role, isolating them in a bubble of convenient information. The very cross-pollination of ideas we designed our buildings to create would be challenged by the software we’ve created to make work more efficient.
This internal corporate dynamic is a microcosm of a larger societal shift. In his seminal work Bowling Alone, political scientist Robert Putnam documented the decline of “social capital”—the networks, norms, and trust that enable a society to function. As we have replaced community leagues and local clubs with isolated, on-demand entertainment, our collective bonds have frayed [8].
The shift to an AI-dominated hub-and-spoke structure within a company represents a corporate version of “bowling alone.” By optimizing for the friction-free efficiency of the individual, we risk eroding the collective social capital of the organization, leaving it less resilient, less innovative, and ultimately, a less creative place to work.
Avoiding Creative Collapse Through Human Connection
Is creative collapse inevitable? Absolutely not. There is a future where we foster a workplace that is more efficient, creative, and fulfilling. This future will be dependent on our ability to promote the highest quality human-to-human interactions at work.
Using sociometric badges, Alex Pentland’s Human Dynamics Lab observed that the most significant predictor of a team’s success was not intelligence or skill, but the pattern, density, and quality of their real-time, face-to-face communication [9]. He suggests that these patterns of communication can account for 30%-50% of a team’s performance. “Collective intelligence” is the innovation output that fuels human collaboration. A company that thoughtfully uses technology to foster meaningful human interaction could enter a virtuous cycle of creativity rather than spiral into creative collapse.
When people collaborate on complex problems, they don’t just solve the task at hand; they build trust, share tacit knowledge, and generate novel ideas. Innovation is an inherently social process, invention rarely springs from an isolated genius but from a connected community of thinkers.
AI as a Super-Broker
The future is not a binary choice between human connection and AI efficiency. The tipping point is not a cliff we must avoid but a dynamic frontier we must manage. For People Analytics and People Technology teams this represents a call to action, a shift in how we measure success and design people-centric digital environments. We must actively monitor the impact of these tools on the company’s social structure and pay just as much attention to what they are doing to the people as to what they are doing to their output [10].
The most important shift we must make is to redefine the AI’s job description. Instead of being a simple hub for answers, the AI becomes a “Serendipity Engine”. Instead of narcissistically positioning itself as the expert, it could respond with, “Here is the information you asked for, but you should really talk to Sarah in marketing. She is the expert on this.” This means fostering what Tomas Chamorro-Premuzic identifies as the truly irreplaceable human qualities: curiosity, emotional intelligence, and adaptability [11]. In this model the AI acts as a connection broker.
An agent-based simulation of a 200-person organization over a 40-week period was used to move beyond speculation. The baseline projection (the dashed dark orange line) confirms the “substitution risk”: as AI becomes a frictionless oracle, employees naturally bypass their peers, causing the organizational network to decay. But this collapse is not inevitable. Four distinct interventions, or “Knobs”, were modeled that could be used to reverse this trend. These include AI Design choices (like a “Gatekeeper” function that deliberately refers complex questions to human experts) to HR Policies (work-from-office requirements). The results, visualized below, offer a mathematical blueprint for resilience: they demonstrate that by combining physical proximity with socially-aware AI, we can achieve the efficiency gains of the technology without sacrificing the structural integrity of the human network.
To prevent organizational collapse, the technology and leadership will need to pull multiple levers. The dashed orange line represents the “Default Decay” caused by high-efficiency AI. The colored lines show the impact of specific interventions:
• Knob 1: The Gatekeeper (green): The AI deliberately refuses to answer 30% of context-heavy questions, forcing human-to-human interaction.
• Knob 2: Warm Intros (dark blue): The AI uses social proof (”Mike trusts Sarah”) to increase the acceptance rate of referrals.
• Knob 3: Double-Sided Nudge (yellow): The AI primes both the seeker and the expert, doubling the relationship value of the interaction.
• Knob 4: Work From Office (light blue): Physical proximity (The Allen Curve) reduces the friction of asking a neighbor, lowering reliance on AI for local problems.
• All Interventions (Orange): By combining policy (WFO) with “Social Physics” design, the organization maintains a network density that could support both a highly creative and highly efficient workforce.
Measure What Truly Matters. This ideal state will not be achieved by default. We must use Organizational Network Analyses (ONAs) more often to monitor the creative health of the company. As Rob Cross’s research has shown, identifying and empowering the hidden “brokers” and “connectors” in your organization is key to resilience and innovation [12]. Going forward, we must include the enterprise AI super-broker and other tools in that network analysis to understand and manage their impact on the company’s social health. This represents a new body of work. Typically software is treated solely as a medium of communication. Building an environment of resilient creativity will require that we understand technology’s role as both a collaborator and a broker and use that understanding to guide organizational design, site strategy, tech strategy, and work-from-office policies.
Work-From-Office. We should be pushing for more people in the office more often and designing an experience that makes it worth their time. The imperative for teams that must innovate is not about accountability, it’s about creativity and team health. When it comes to fostering highly productive, creative, and healthy teams there’s no replacement for in-person interaction. While we may be able to reduce the impacts of hyper-efficient AI on a remote workforce through technological interventions, there is no panacea for the reduction in collaboration that inevitably follows geographical distance.
We have an opportunity to foster stronger, more creative teams than ever before, but it won’t happen on its own. If we do nothing, we will end up with AI tools that serve the individual task, creating a world of hyper-efficient, isolated spokes. Or we can take advantage of this moment to create a whole new Allen Curve, an understanding of the modern “campus” built on the knowledge that the health of the connections between us is the greatest predictor of our long-term success.
Works Cited
Michael Arena, Scott Hines, John Golden III. The three Cs for cultivating organizational culture in a hybrid world. Organizational Dynamics Volume 52, Issue 1, 2023
McDowell, T. (2023). Strategies for organization design: Using the Peopletecture model to improve collaboration and performance. John Wiley & Sons.
Allen, Thomas J., and Gunter Henn. The Organization and Architecture of Innovation. Elsevier/Architectural Press, 2007.
Wegner, Daniel M. “Transactive Memory: A Contemporary Analysis of the Group Mind.” In Theories of Group Behavior, edited by B. Mullen and G. R. Goethals, Springer, 1987, pp. 185-208.
Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips. Science, 333(6043), 776–778.
Granovetter, Mark S. “The Strength of Weak Ties.” American Journal of Sociology, vol. 78, no. 6, 1973, pp. 1360–80.
Pariser, Eli. The Filter Bubble: What the Internet Is Hiding from You. Penguin UK, 2011.
Putnam, Robert D. Bowling Alone: The Collapse and Revival of American Community. Simon & Schuster, 2000.
Pentland, Alex. Social Physics: How Good Ideas Spread—The Lessons from a New Science. Penguin Press, 2014.
Hao-Ping (Hank) Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ‘25).
Chamorro-Premuzic, Tomas. I, Human: AI, Automation, and the Quest to Reclaim What Makes Us Unique. Harvard Business Review Press, 2023.
Cross, Rob, and Andrew Parker. The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations. Harvard Business School Press, 2004.
Great read. Love the reframe of AI as a tool that reinforces human connection.
"I appreciate your focus on AI as a tool for enhancing productivity and, by extension, job satisfaction. I’ve found that a great way to maximize AI assistance is to utilize multiple platforms simultaneously.
By prompting the four major free AI models with the same question, then asking them to merge and synthesize the collective results, you create a much stronger output. Scaling this to a team level—where 20 people merge their individual AI-assisted findings into a single project document—can be incredibly efficient.
However, it’s vital for teams to remember that AI isn't just a fancy search engine; it’s an engine prone to 'garbage-in, garbage-out' errors despite its linguistic polish. Human oversight is the 'secret sauce.' Employees must apply their own accumulated wisdom to tweak and refine these outputs. The intersection of team social interaction and AI is where we’ll see solutions build exponentially."