When Productivity Is No Longer the Bottleneck, How Should Teams Collaborate?

#ai #reflection #teamcollaboration

I've been thinking about a question lately. For quite a while, I haven't figured it out.

How to lead teams, how to divide work—these were things I thought I had figured out. Now I realize the playbook doesn't work anymore. AI has shattered many assumptions that used to hold true. Honestly, I still haven't found a new approach.

Functional Division Is Breaking Down

The premise of traditional division of labor is simple: you specialize in frontend, I specialize in backend, they specialize in design. Everyone does their own thing, and together we assemble a complete product.

This premise is crumbling. AI is rapidly expanding everyone's capability boundaries. One person with AI can do more than an entire small team used to. So does it still make sense to divide functions based on "who's good at what"?

What's more troubling is that within the same team, some people have used AI to 10x their productivity while others haven't touched it at all. This isn't traditional skill disparity—it's differing acceptance of new tools. You simply can't evaluate and allocate work using the old methods.

Building Things Is Faster Than Having Meetings

The bottleneck of collaboration used to be communication. Meetings, aligning goals, discussing plans—these consumed vast amounts of time. But back then, production itself was slow, so spending time talking was worthwhile.

Now it's the opposite. While you're still sitting in meetings discussing, the thing has already been built, deployed, and is serving users. So why would you still sit down and talk with someone for half a day?

Here's the tricky part: the speed of absorbing information can no longer keep up with the speed of creating and validating. People don't want to talk anymore; they want to go straight to building. This isn't an attitude problem—the efficiency structure has changed.

Only Validated Ideas Are Valuable

When everyone has high productivity, what becomes scarce?

Not ideas, and not output. Both are depreciating in value. What's valuable are validated hypotheses. You propose an idea, test it, and find that it works and solves the problem. This process of "validation" is what matters.

So what should the team do? Accelerate validation speed.

I have an immature hypothesis: run horse races under shared goals. Everyone explores and validates independently; no need to integrate from the start. Good ideas that emerge get extracted to form shared infrastructure that everyone can use. Seek common ground while preserving differences—allow personalization under one platform. For example, deploy some AI agents specifically to run analysis and see which directions are more likely to yield good results.

Passion May Matter More Than Skills

Traditional division of labor looks at skills. But once AI has leveled the skill differences, what makes people create different things is what they care about.

Some people are passionate about creating new things and want to try everything. Others care more about whether the system is stable and secure. Some care about making the product look better and the experience more pleasant. Others want good ideas to be seen by more people—not just building, but also promoting. You can't just build, right? Spreading validated ideas is itself important.

These concerns differ, but a good product precisely needs to pass muster across many dimensions. Standing out in only one aspect while being rough elsewhere won't get you far.

Diversity Might Be the Way Forward

I have a vague feeling that diversity might be the answer.

Not diversity as a slogan. I mean teams no longer need a bunch of similar people, but rather very different ones. Different passions, different perspectives, different aesthetics—even using different AI models. These differences themselves create value.

You add a brick, I add a brick. Some modify it to make it look better, some to make it run faster, some to keep it from crashing. Together we build an engine, each person contributing different dimensions.

Conversely, similar people gathered together actually struggle to form synergy. All focusing on the same dimension, they end up competing with each other more than collaborating.

No Conclusion

Writing this far, I still have no conclusion.

AI is triggering organizational-level transformation—I'm certain of that. But how to respond, I'm still figuring out. Sense of purpose has become more important; if the goal itself doesn't inspire people, no division of labor will work. Passion is replacing skills as the new basis for division. Diversity has become important.

But how do these become a functioning collaboration system? I don't know.

I think this is a problem every team will have to confront. Productivity is high, old collaboration methods don't work well anymore, and new ones haven't grown yet.