Imagine a fully remote company where someone has a problem to solve. A while ago, they might have asked a colleague for a perspective, set up a call with someone in another function, or sent an unfinished idea to a teammate and asked, “What am I missing?”
Now they can take the problem to AI. They can get access to information, work through questions, produce a draft and come back with a polished recommendation. They can do a lot of that on their own, and they can do it fast. When they present it, everyone else sees the result. They may not have been part of the thinking that led there.
AI is changing work. Subscribe to The Human Part to explore what that asks of us as leaders.
I don’t think the question is whether a person should ask a colleague. The question is whether, as organizations foster the habit of doing work with AI, people will turn to each other as often. If they don’t, what kind of organization are we building?
The individual gets faster. What happens to the organization?
This is a precarious space for leaders. We want to get things done faster. AI can help us do that. It can give someone access to knowledge and a way to work through a problem when the alternative might have been finding the right person and waiting for time on their calendar.
There is a behavioral pull to that. You feel accomplished. You feel like you are doing more work. AI gives you a response when you want it, in the form you ask for, at a speed collaboration often can’t match. The response still needs to be checked, but the experience of getting something back immediately is hard to ignore.
Now repeat that choice across an organization. A question that once went to a colleague goes into a prompt. An early thought that might have been shared stays on one person’s screen until it becomes a finished recommendation. A cross-functional discussion happens later, if it happens at all.
That doesn’t mean every question deserves a meeting. It doesn’t mean a quick answer from AI is a bad decision. I am concerned about the pattern that can develop when we add those individual choices together. An organization can become more individualistic without ever deciding that is the culture it wants.
My worry is that the effect goes beyond whether people like working together. I believe some of the best organizational growth comes through collaboration: people bring different knowledge to a problem, challenge each other’s assumptions and discover options they might not have reached alone. When we stop needing each other for more of the work, we risk losing some of the relationships and habits that make that possible.
That is a concern and a prediction, not a finding that AI has already fractured every team. The evidence is more complicated than that, and it is worth looking at before we make the story too comfortable.
AI doesn’t always mean less interaction
A 2026 working paper from INSEAD researchers studied 316 employees in a European technology services company. Members of 42 teams were assigned either an AI assistant grounded in organization-specific information or their usual way of working. After three months, employees with access to the assistant were more central in the reported collaboration and knowledge-sharing networks.
That is not the result you would predict if AI automatically made people retreat from one another. It is also one company, one implementation and a three-month window. The researchers relied largely on survey measures of working and advice ties, and they say the longer-term pattern could change. The study doesn’t tell us what will happen in every remote company. It does challenge the idea that isolation is inevitable.
An October 2026 preprint examined digital collaboration data from more than 12,000 employees in one multinational organization during a staged Copilot rollout. The researchers estimated more collaboration time and interaction with more colleagues after adoption, while also finding that workers concentrated more of their collaboration on people with whom they already had stronger ties. The data came from a highly interdependent business unit and measured digital interactions such as meetings, calls and email. It cannot tell us everything about the quality of those relationships or prove what happens in a different organization.
And a study of software developers found another possibility: people sometimes turned to AI instead of colleagues for routine technical questions, while still seeking people out for context, mentorship and connection. Some developers reported that conversations could shift toward clarification, joint reasoning and alternative perspectives. This was a relatively small mixed-method study of developers, not a measure of every workplace. But it suggests that the kind of conversation may change even when conversation doesn’t disappear.
I take those findings seriously. Maybe AI gives people more time and better information for collaboration. Maybe it creates new questions that require colleagues to help interpret the work. Maybe, in some organizations, the tools and routines make people more connected. I don’t know that we can declare a single outcome.
What I do know is that leaders shouldn’t assume that individual productivity tells them what is happening to the organization. More messages or meetings wouldn’t settle the question either. We need to understand whether people are actually bringing different perspectives into the work, whether they are learning from one another, and whether the relationships exist when the next difficult problem arrives.
What we may lose between the prompt and the presentation
Some knowledge is written down. Some of it lives with people.
A person in customer support knows where customers get confused. Someone in operations knows which handoff consistently breaks. A new employee may ask a question that everyone else stopped noticing. A colleague in finance may see a cost that didn’t occur to the person designing the solution. None of those people has a monopoly on the right answer, but each may see something the others don’t.
If one person uses AI to produce a recommendation and then asks the team to approve it, those perspectives can still enter the room. But the invitation is different. People are responding to a solution that already has a shape. They may not see which options were discarded or which assumption sent the work in that direction. It can be harder to challenge a polished answer than an unfinished question.
Imagine a remote team deciding how to change a customer onboarding process. One person asks AI to research approaches and prepare a recommendation. The result looks thorough, and the team can review it in a fraction of the time it might have taken to build a proposal together.
That speed has value. But suppose the support team knows customers regularly misunderstand one step, operations knows the proposed handoff is difficult to staff, and a newer employee can describe where the process was confusing from the inside. Those are hypothetical details, not a client story. My point is that a polished proposal is not the same thing as having those people contribute while the problem is being defined.
Perhaps their insights would change the recommendation. Perhaps they would confirm it. Either way, the discussion gives the organization something more than a document: it gives people a chance to understand how their work connects and how their colleagues think.
There is also the relationship side. I worry that, over time, people will spend less time getting to know colleagues and less time in the conversations that happen around the work. Call them water-cooler conversations if you want. In a fully remote company, culture is already challenging enough. If AI makes it easier to work without reaching out, those moments may need more deliberate attention.
I expect this to be a greater challenge for remote organizations than for companies where people regularly see each other in person. That is my expectation, not a comparison the studies above have established. Being in an office doesn’t guarantee that people talk across functions, and a remote team can make real connections. The question is what habits the organization is creating.
The wisdom of the crowd is not automatic
When I think about collaboration, I keep coming back to the wisdom of the crowd: what becomes possible when people with different perspectives contribute to a decision. AI can bring a lot of information into the conversation, but it doesn’t know what a particular colleague has learned in a conversation with a customer unless that knowledge is available to it. And a generated perspective isn’t the same as that person being there to challenge, clarify and take responsibility for what happens next.
The research also gives us a useful warning. A classic 2011 experiment on crowd judgment found that, on the factual estimation tasks studied, seeing other people’s answers could narrow the range of estimates without improving the collective error. People became more confident as their answers converged. That study was about numerical estimates, not a remote team designing an AI workflow. I wouldn’t use it to claim that discussion makes teams less accurate in general.
But it does remind us that putting a group in a room doesn’t automatically produce wisdom. People can follow the most confident voice. They can defer to a leader. They can converge around the first answer, including the first answer an AI system produces. More people talking is not the same thing as hearing genuinely different information.
The opportunity is to be deliberate about both parts: protect the independent knowledge people bring, then give them a way to examine it together. AI might help the group organize ideas, surface options or identify a question they haven’t considered. It shouldn’t become the only voice the group responds to before anyone has had a chance to think.
And we shouldn’t pretend a human group always does better. A 2026 product-development experiment at Procter & Gamble found that individuals with AI matched the evaluated performance of teams without AI on the tasks studied; teams with AI also participated in the experiment. The study involved 791 professionals in a one-day virtual workshop using real product challenges. That’s meaningful counterevidence to the idea that human collaboration is guaranteed to produce a better evaluated output. It doesn’t measure what happens to relationships, learning or culture after years of working together or apart.
That distinction matters to me. If the only thing we measure is the quality of one finished recommendation, we may decide the individual with AI did everything the team would have done. But a team also carries knowledge and relationships forward to the next problem. Those are not proven benefits of every meeting. They are things worth paying attention to when we decide how to work.
Slowing down to go fast
The art of collaboration is slowing down and thinking through the permutations of what could and could not be possible, distilling that down to the best set of options, and then creating an action item from it.
That doesn’t sound like an argument for having a meeting about everything. It is a challenge to leaders: can you articulate and facilitate an environment where slowing down to go fast is a better approach than just going fast?
Consider the onboarding example again. Instead of asking one person to arrive with the answer, the leader could start with the problem and invite support, operations and the people closest to implementation to describe what they know independently. Where do customers get stuck? What constraints are real? What would each person want the group to avoid?
Then the group could use AI to organize those contributions, identify gaps and suggest possibilities. The people involved could challenge the output, decide which options deserve attention and leave with an owner and a next action. That is a proposed process, not a tested intervention or a guarantee of a better decision. Its value depends on the question, the people in the discussion, the quality of the information and whether the leader actually makes room for disagreement.
I am not suggesting that a colleague’s idea is automatically better because it came from a human. A group can waste time, reinforce the boss’s view or hold back someone who already knows what needs to be done. There are decisions an individual should make. The work of leadership is to be clear about the result you want and design a process that gives the right perspectives a chance to improve it.
The decision is not simply whether to let people use AI. People are going to see how much they can do with it, and they should be able to benefit from that. The question is whether the organization is also building the capacity to think together.
Connection has to be designed
For a remote leader, some of this means creating cultural connection moments. That might be virtual coffees. It might be bringing people together in person for activities that don’t necessarily involve AI, including time to think and collaborate as a group. Neither is a magic fix. A virtual coffee cannot repair a culture where people are punished for disagreeing, and an in-person event doesn’t guarantee that different functions will keep working together afterward.
I would want leaders to look at the work itself as well as the calendar. Who is included when a problem is still open? Are people learning from colleagues in other functions, or only seeing the final output? Is there room to ask a question without feeling that you should have already found the answer from AI?
There is a second question here: what happens when AI is in the middle of the group conversation? Organizations are experimenting with group chats that include AI, and in my experience they do an okay job. But behaviorally, people are still struggling to use that setting to contribute their own thoughts and reflections on how to do things better.
That is the million-dollar question. How do we collaborate through AI in a way that brings together the wisdom of the crowd and the intelligence of AI?
I don’t have a finished answer. It is a significant opportunity for organizations to work on. If we leave the process to chance, the easiest individual choice may become the culture. If leaders pay attention to how people bring their knowledge together, AI may become part of stronger collaboration rather than a reason to avoid it.
Speed matters. So does what the organization can do together after everyone has become faster on their own.
Prepared with AI assistance from my interview answers.
Dr. James Kelley
drjameskelley.com
Get The Crucible’s Gift on Amazon
Check your AI fluency
If this gave you something to think about, I’d love to have you back. Subscribe for more on AI, leadership, and the human part of work.
Sources
Büchsenschuss et al., “GenAI Adoption Increases the Density of Knowledge and Collaboration Networks,” INSEAD working paper, 2026. One firm, organization-grounded tool, three-month follow-up and largely survey-based network measures.
Soda, Iorio and Rizzo, “The Effects of Generative AI on Collaboration and Social Networks Among Knowledge Workers”, preprint submitted October 23, 2026. Digital collaboration records from one multinational business unit; results do not establish relationship quality or universal effects.
Salomon et al., “From Disruptions to Discussions: How GenAI Impacts Human Interactions in Software Development”, first public January 27, 2026. Mixed-method study of developers; findings are reported perceptions and do not establish outcomes for all jobs.
Lorenz et al., “How Social Influence Can Undermine the Wisdom of Crowd Effect”, 2011. Historical experiment on factual estimates; not a test of an AI-assisted organizational meeting.
Dell’Acqua et al., “The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork”, Organization Science, 2026. One-day virtual product-development workshop; not a longitudinal study of culture.


