AI doesn't know how to work as a team

—We decided to bring AI into the company. Each employee, their own account. Let everyone figure out how to use it. A reasonable investment, almost obvious for 2026.

—And?

—Three weeks later, each person is a project that diverges.

I heard it at a coffee meeting recently. And I've been hearing it, in different versions, in every conversation with clients and colleagues for months now.

The story always starts the same way. The decision was good. The intention was clear. The employees are using the tool. Individual metrics improved. Nobody is complaining.

But when you look at the company as a whole, something doesn't add up. There's no new process. There's no collective learning. There's nothing you can point at and say "we did this with AI and we couldn't before".

Each employee a little more efficient. The company, exactly where it was.

And the question that lingers in the room: what exactly did we invest in?

What's happening

Something is going on in how companies are adopting AI that's worth naming:

AI doesn't know how to work as a team.

This isn't a technical limitation. It's a structural feature of the product. The AI tools the market made available these past years are extraordinarily good at one thing: augmenting an individual. Turning a person into a more capable version of themselves. That's real and not up for debate.

But augmenting an individual is not the same as improving an organization. Sometimes it's exactly the opposite.

What happens when every person has their own AI

Three things occur at the same time, without anyone naming them:

Decisions fragment into individual chats. A person used to think through a problem out loud with a colleague. Now they think it through silently with their AI. The conclusion arrives at the meeting, but the process that produced it doesn't. What used to be collective thinking becomes the negotiation of pre-cooked positions.

Actions get executed in isolated sessions that leave no trace. AI accompanies the person doing their work, but the work stays on their screen, in their context, in their session. There's no shared state. There's no record the team can review. There's no learning the organization can capitalize on.

The knowledge each interaction produces stays trapped in the private surface of the individual who initiated it. Every conversation with an AI produces an insight, a decision, a new articulation of a problem. All of that disappears when the tab is closed. The organization doesn't learn — only the individual does, and only for a while.

Three weeks later, the company has twenty different versions of how to use AI. None of them wrong. None of them coordinated. None of them adding up.

The bias holding this up

There's a dominant narrative we're not questioning: AI is a tool for individual productivity. That narrative is legitimate — for many things, it's true. But it became so dominant that it covered another question that should have been just as important:

What does AI do for us as an organization?

Not what it does for me. What it does for us.

Nobody is asking this because the answer is uncomfortable. Augmenting an individual is trivial — you do it with an account. Improving an organization is hard — it requires thinking about how work connects, how knowledge gets preserved, how shared state becomes visible. No mass-market product solves that, so the conversation drifts toward what can actually be bought: more accounts, more tools, more individual assistants.

And then what we're seeing happens. Companies that invested in AI. Employees who use it. Individual metrics that improve. And a growing sense that something isn't quite working, that the whole doesn't add up, that each part performs but the collective stays where it was.

The consequence almost nobody is naming

The collective intelligence of an organization is not the sum of individual intelligences. It's what emerges when those intelligences brush against each other, contradict each other, modify each other, validate each other. It's the space between people, not the people themselves.

That space is dissolving in silence.

When each person thinks with their AI before thinking with their team, the team stops being a space for thinking. It becomes a space for coordinating thoughts produced somewhere else. Meetings no longer produce ideas — they align pre-cooked ideas. Workspaces no longer show work in progress — they show outputs of processes that happened on private screens.

A company thought it was adopting AI. What it was actually adopting, without realizing, was the replacement of collective thinking by twenty sessions of private thinking.

What we're building

We're building individual AI tools. That's fine and it will continue.

What we haven't built yet are the collective tools. And until we do, what's happening will keep happening: companies that invest in AI and discover, three weeks later, that each employee is a project that diverges.

The individual chat is not enough for an organization.

This isn't a critique of the individual chat. It's the recognition of a limit the market hasn't named yet.

Thinking as an organization requires more than twenty private chats running in parallel. It requires shared state. It requires cross-visibility. It requires that the knowledge produced in one conversation can be seen, modified, and capitalized on by the other people on the team.

That's what we're working on with our clients at OpenNube.ai. Not replacing the individual chat — completing it. Making sure the intelligence each interaction produces doesn't get lost in a tab that closes, but instead becomes part of the organization's shared state. Giving the team back its place as a space for thinking.

In the meantime, there's a question worth asking before giving every employee their AI account:

Are we adopting AI so each person thinks better, or so the organization thinks better?

If the answer isn't clear, this isn't a budget problem. It's a problem of what we're buying, and why.

---

AI doesn't know how to work as a team. Yet. Whoever teaches it will define how organizations work for the next ten years.

---