Beyond Solo AI: Harnessing Collective Intelligence in Teams

This past week, Tobi Lütke, CEO of Shopify, introduced a policy requiring teams to explore AI solutions before filling new roles, effectively making a thorough search for automation or generative AI tools a prerequisite to expanding headcount.

The rationale is that if existing or emerging AI technologies can handle certain tasks, hiring for those functions may be unnecessary, thus streamlining Shopify’s operations.

Cutting to the chase, he is saying that if you can solve a problem with AI, then you don't need to hire someone.

I get where he is trying to go, but does he have it wrong?

Almost uniformly, when I get asked to talk with a company about their AI tool set, the CEO has already taken action.

Most commonly there has been an attempt at implementing AI tools such as creating an AI use policy where employees can use ChatGPT, or, my all-time favorite: implement Microsoft Copilot since they are already running other Microsoft tools.

Somehow, this is magically supposed to solve the AI problem, and I think it momentarily does for the board and maybe for stockholders.

Although I think both are figuring out that just adding individual tools does not make a company AI literate or more productive.

In fact, in this essay, I'll argue that it might make things worse.

Everywhere I look in organizations today, I see individuals brandishing AI tools like ChatGPT as personal superpowers.

Workers are cranking out reports, strategies, and code faster than ever, often working solo alongside their AI assistants. This surge in individual productivity is exciting—who wouldn't want a personal genius on call?

But as impressive as this is, I can’t shake a nagging question: what happens to the team?

In our rush to turn every employee into an AI-augmented powerhouse, we risk undervaluing something far more powerful: collective intelligence.

The magic of a great team isn’t just the sum of each person's output; it’s the collaboration, the clash and blending of ideas, the shared context that ties work together.

If everyone retreats to work with their AI in isolation, we may inadvertently chip away at that magic. Instead of AI making our teams smarter as a whole, we could end up with brilliant individuals who no longer gel into a brilliant team.

The real promise of AI in organizations isn’t to create a few star performers, but to elevate how we think and work together.

The Hidden Cost to Teamwork

Individuals are increasingly working one-on-one with AI tools.

In many workplaces, using AI has become a solo endeavor. An analyst might quietly use a chatbot to draft a client proposal without looping in her team for input. A developer might debug code with an AI assistant instead of asking a colleague.

On the surface, this seems great: each person is faster and more self-sufficient. But when everyone is doing this, an unintended side effect creeps in—teamwork can fray around the edges.

When every employee becomes an island powered by their own AI, we risk weakening the connections that make a team more than a group of individuals.

Consider how idea-sharing works now: rather than bouncing thoughts off a teammate, someone might just trust the answer their AI gives them. The casual hallway discussions or impromptu brainstorming sessions may happen less often when answers are a click away.

Over time, the team’s shared understanding — the context everyone used to build together through dialogue—can become fragmented. Each person might now have slightly different information or approaches influenced by their personal AI assistant.

The result is fewer “a-ha” moments born from debate, less alignment on decisions, and a decline in team cohesion.

In short, the very strengths that make a team special—collective problem-solving and mutual learning—get diluted.

The Unique Magic of Team Collaboration

Even before AI, we’ve known that a well-functioning team can accomplish things no lone genius can.

Why? Because there are qualities of group collaboration that amplify thinking and creativity.

Consider what truly great teams do:

• Creative Collisions: When people with different ideas and perspectives bounce off each other, it sparks innovation. A casual comment in a meeting or a question from a colleague can trigger an insight that no one would’ve arrived at alone. This spontaneous “idea collision”, or collaborative collision, is often where breakthrough solutions are born.
• Shared Context and Alignment: Through discussion and debate, teams build a shared understanding of the problem and the goal. Each conversation layers more context that everyone absorbs. That means when action is finally taken, everyone is on the same page about why and how—something that’s hard to achieve if each person has been working separately with their own AI-generated information.
• Distributed Intelligence: In a team, knowledge is distributed. One person knows the client’s quirks, another knows the technical constraints, and someone else remembers past mistakes. By pooling these pieces, the group’s intelligence covers a much broader field than any single member’s. This distributed cognition allows teams to tackle complex problems by leveraging each member’s strengths.
• Collective Learning: Teams don’t just solve immediate problems; they also learn and improve together. In the process of collaborating, people pick up new skills and insights from each other. Over time, the team develops an institutional memory and a way of doing things that is richer than any manual or database. The expertise lives in their interactions and shared experiences.

Augmenting Collective Intelligence

How can we get the best of both worlds—AI’s efficiency and the magic of teamwork?

The answer is to refocus our approach: integrate AI in a way that augments group intelligence, not just individual smarts. In other words, use AI to make the team as a whole more intelligent.

This means leveraging AI to support distributed cognition—the way knowledge and thinking are spread across people and tools—so that the collective brainpower of the organization grows.

Imagine an AI that acts as a collaborative partner to the entire team. Instead of each person getting different answers from different AI assistants, the team shares an AI system that everyone feeds with information and draws insights from. It could summarize the team’s meetings, highlight connections between ideas raised by different members, and ensure that knowledge learned by one person is available to all.

Such an AI becomes a sort of communal brain, enhancing memory and awareness across the group.

The real promise here is an AI that amplifies communication and coordination: flagging when one department’s update applies to another, or capturing the nuances of a decision so no context is lost. Think of it as adding a super-intelligent team member whose key role is to make everyone else more informed, aligned, and capable.

When AI is woven into the fabric of team workflows like this, it stops being a threat to teamwork and starts becoming a catalyst for it.

AI could improve brainstorming sessions by offering relevant data and experiences to everyone at the most opportune moments.

Decision-making could improve when an AI helps aggregate everyone's input and suggests options that reflect the team's combined thinking.

In short, AI can elevate the group's collective intelligence—the capacity of a team to perform at a level higher than any member could achieve alone.

That is the real game-changer: not one person doing more with AI, but the entire team doing more together with AI’s help.

When Team Knowledge Is Lost: A NASA Story

In 1970, it was the tight coordination of NASA’s Mission Control team that brought Apollo 13 safely home.

Flight directors, engineers, and astronauts all pooled their expertise in real time, each contributing a crucial piece to solve the puzzle of that damaged spacecraft.

This incredible rescue wasn’t the result of one person’s knowledge; it was the triumph of a network of people, each relying on others, sharing information, and thinking together under pressure.

And it wouldn't have been the work of a single engineer using ChatGPT.

NASA later learned a poignant lesson about the value of that shared context.

Years after the Apollo missions ended, NASA found it impossible to easily resurrect the capability of returning to the Moon.

It’s not that they lost the paperwork—the schematics and blueprints were all stored away in archives. But without the original teams around, those documents were like the stones of an ancient monument: present, but silent.

The engineers who had lived and breathed the Apollo program had moved on, and with them went the unwritten understanding of how everything really worked.

The know-how wasn’t fully captured in manuals; it lived in the conversations, decisions, and collective experience of those teams. Once the teams disbanded, NASA’s institutional memory for moon travel faded. They had to relearn and rebuild much of that expertise almost from scratch.

This story highlights a critical point: true expertise in an organization lives in the interactions and shared context of its people.

It’s a reminder that no matter how sophisticated our tools or documentation, they are only as meaningful as the team context around them. If we allow teamwork and collective learning to fade—if we rely on AI or any tool to replace human interaction—we risk creating a future “lost archive” of knowledge in our own companies.

The goal should be to use AI to strengthen team knowledge, not to let the team’s knowledge disappear trusting that “the system” remembers.

As NASA’s experience shows, a library of information means little without the living collective intelligence to give it meaning.

When AI Isn’t a Team Player

Picture a company that deploys a powerful AI chatbot on its customer support website.

This chatbot is brilliant—it instantly provides customers with detailed answers to technical questions. In fact, it knows the company's products inside out, drawing from a comprehensive knowledge base. At first, customers liked the new system, and management congratulated themselves for using advanced AI to answer questions.

Now imagine that same customer, after chatting with the bot, calls the support hotline to follow up or clarify a question with a human agent. The human support rep picks up the phone, unaware of what the AI chatbot told the customer. The customer says, "Your bot just told me X, can you help me with Y?" and the support agent responds with a confused silence–they might not even know the information the AI provided. Or worse, the agent gives an answer that contradicts the chatbot.

The customer experience quickly sours.

The incredible knowledge held by the AI isn’t in the heads of the support team, and the team’s processes haven’t adapted to include the AI’s inputs. A "virtual colleague" provides information the support staff never received, making them feel undermined and out of the loop internally.

So be cautious.

The AI may be extremely capable, but it’s operating as a lone genius, not as part of a coordinated team. The result is a disjointed experience—much like a team where one member acts independently of the others.

It shows that simply adding an AI solution without integrating it into the fabric of teamwork can create new silos. Knowledge needs to flow through the team, whether it comes from a human or a machine.

If the AI isn’t a team player, the entire organization can end up looking fragmented and inconsistent.

Rethinking Collaboration

Integrating AI into our teams requires a conscious effort to keep people connected and informed. It’s not just about adopting new tools, but about developing our collaboration habits and workflows.

Here are a few forward-looking ways to begin:

• Design Collaboration-Focused AI Workflows: Rather than each person using AI in a vacuum, set up workflows where AI outputs are shared and discussed. For example, if someone uses an AI to draft a plan, have them bring that draft into the team meeting for critique and improvement. Make AI a step in your collaborative process, not a substitute for it.
• Create a Shared Knowledge Hub: Ensure that the knowledge AI is using (and generating) is feeding into a common repository that the whole team trusts. This could mean training your AI tools on a centralized knowledge base that’s updated by the team, and letting everyone see and refine what the AI learns. When AI provides an answer, everyone should be able to access that same information, preventing divergent silos of data.
• Train Teams Together with AI: Invest in team training sessions on how to use AI collectively. This might involve workshops where teams solve problems with AI input in the room, learning how to collectively vet and integrate AI suggestions. Encourage team members to share useful prompts or techniques with each other, turning individual AI successes into team best practices. Consider developing a collective "prompt library" to share among teams.
• Keep Humans in the Loop: Establish practices that keep human judgment and team alignment at the center. For instance, if a customer-facing AI is deployed (like the chatbot scenario), schedule regular sync-ups where the AI’s most frequent answers or updates are reviewed with the entire support team. Make it standard that no AI operates completely standalone; there’s always a human overview and a feedback loop into the team’s decision-making.
• Create a Culture of AI Acceptance: Encourage teams and individuals to use AI tools and to not feel reluctant to share that content or outputs were created with the help of AI. Instead, encourage use and applaud the integration of advanced tools and new found efficiency.

Ultimately, embracing AI as a catalyst for collective intelligence means never losing sight of the human element.

Leadership should promote a culture where sharing knowledge is just as important as generating it.

Encourage teams to experiment with AI in their group workflows and openly discuss what works or doesn't.

The companies that will thrive with AI are those that reimagine their teamwork, making sure that “we” remains stronger than any one “I”—even if that “I” is an AI.