Lessons from Phased Management
In small teams, external push and self-drive both matter. Between hard constraints and no management, adjusting by phase often beats a single architecture.
I have a friend who has been a serial founder for years and has managed small teams for a long time. We had a deep conversation about managing small teams, and it gave me plenty of insights.
Before we talked, he wasn’t clear on the concept of an AI Native Team architecture. He asked what this new term, “AI Negative Team,” was. I told him it meant a no-management or low-management team architecture, and then I asked what he thought of that architecture.
After a long silence, he didn’t tell me whether he judged the architecture good or bad. He told me his team’s story instead.
His team stayed at around 10 people year round. Recently, because of the market, they were forced to cut down to 5. Those 10 were never fixed. Over the years, he told me, people came and went — around 20 people passed through — and in the end all of them left for one reason or another. The last 5 who stayed walked with him from start to finish.
His first reaction: a no-management team architecture is essentially like remote work during the pandemic — it lacks both external push and internal push. External push is the force a leader or KPIs bring. For internal push he gave me a detailed example: the daily morning standup. If you did no work at all the day before, then at the next morning’s standup you have no way to report to your leader. Others will look at you strangely. That sense of embarrassment makes you generate your own drive to work from the inside.
During that stretch of remote work in the pandemic, because there were no face-to-face meetings, everyone’s floor for shame was lowered, and their internal drive decreased.
His description was very apt, and very simple. I then told him that AI Native Team member selection really is the same as remote work: you only pick people with strong execution and long-lasting drive.
A no-management architecture is far harder to sustain than a remote-work architecture, and there’s another very important reason: a no-management architecture requires people to claim their own tasks, while remote work, like traditional architectures, assigns tasks. How to avoid human laziness became a very important question.
Only when that problem is thoroughly solved can you start building the team.
He told me his team can work remotely right now. But if a new person joins at this point, you have to re-evaluate his habits and ability. Early on you need to sell the vision and pretend to get angry to correct the new member’s bad habits. You can’t put him on remote right away, because you don’t yet know his style, his ability, or whether his self-drive is sufficient.
Strong management early on is meant to guarantee a floor for members’ working capability. For work that chases efficiency, hard constraints are always the best solution. But for work that requires innovation, a no-management architecture is instead the best option.
The will of the group affects each individual’s development. If you hold a standup every day, or a review every week, the group’s will gets spread onto every person and buries personal ideas.
After hearing his thoughts, I said his team already had the AI Native precondition: every member trusts the others and their self-drive is extremely strong.
He answered that he couldn’t judge whether their current work was strong on innovation or needed efficiency — in reality they need both, because every client’s needs are different. Some clients ask for things that can directly reuse previous projects or experience, while others want fully customized work that can’t reuse any past experience. And their work keeps jumping back and forth between innovation and efficiency. So they’ll just run on the most traditional management model.
When designing an AI-native team architecture, you can’t take it for granted that one particular architecture is the most suitable. Instead, amid dynamic demand, you have to find a team model that can keep coordinating. That adds to the complexity of designing a team model.
How to effectively manage an AI Native Team has become a very complex question worth digging into — and every founder has to face it. Between managing and letting go, finding a model you can adjust by phase or allocate on demand may be the best solution.