Between Excitement and Concern: Leading Software Teams in the Age of AI
With AI, software development is solved. Anyone can build their own software today and the model of choice takes care of the implementation. So why do we still need software developers?
There is a lot of discussion, philosophizing and speculation about this right now. I keep coming back to a different question. What does it mean for leadership? Does everything just carry on as before?
When I founded my own company, I took a step back, from leadership to development. Not every leader can afford to do that. Most are stuck in the middle of their day-to-day work and responsibilities. At the same time, they see software development changing fundamentally. And with it, the daily work of their team.
This switch gave me an insight I would never have had as a leader. Simply because I would not have had the time. Today I experience hands-on how software development with AI works. How far you can push it. And where its limits are today.
Two things are very clear to me. A software developer with AI is incredibly powerful. And at the same time, software development is still software development. The same questions as before still matter. It is not just about the initial build, not just about the surface. It is about making sure a solution runs reliably.
How do you, as a leader, notice what is changing right now? What part of it is good? And how do you lead a team through a time in which their own profession is being reshaped?
Three camps in one team
In many teams I see the same patterns. Again and again.
On one side are those who are all fired up. They are in the middle of a Sturm und Drang phase, discovering new things and building at an enormous pace. Day after day, sometimes hour after hour, they present something new. That moves the team forward. And sometimes risks are knowingly accepted along the way.
On the other side are the skeptics. They wonder how this is supposed to work in the long run when nobody understands in detail how an individual step was implemented. Anyone who builds and runs software knows that the initial build is only one part. Running a large application that has grown over years is a challenge of its own. Without really understanding how it is built, that becomes hard.
Then there is the question of whether the profession will exist in this form at all. A question every role faces today. And wherever so much is changing, fear of the future is never far away.
In between, there is the middle. People who are already building with AI, feeling their way forward step by step and finding their own path.
The task of leadership is to resolve the tension between these camps in a meaningful way.
When everything suddenly looks finished
As if the dynamics within the team were not enough, the environment around it is changing too. Many people around the team are convinced that with AI, everything is easy now.
Where there used to be one request, there are now ten. Where there used to be a simple question, there are now ten-page PDFs. Finished prototypes. Things that look finished. On the surface.
And with them comes the obvious question. “The prototype works. It does what the application is supposed to do. Why can’t you just take it into production and run it?”
The answer is rarely simple. But the team has to deal with this expectation. And the team’s leader even more so.
Leadership therefore has to work in two directions. Inside the team, people need orientation for a changed way of working and someone who brings the different types together: those who lead the way and those who increasingly feel left behind. Outwardly, leadership has to explain what lies between a working prototype and software that can be run reliably.
Nobody knows where this journey leads
Behind all this are big questions. If AI makes so much more possible, will there be enough work for a team of today’s size? Or will the same topics require far fewer people in the future? And what does that mean for each individual on the team?
Every company is facing these questions right now. There is no single right answer yet. Everywhere, people are experimenting, trying things out, learning. It would be wrong to claim anyone already has the answer.
In my view, three things matter most for leaders in this phase.
1. Set a framework that holds
With AI, some control is lost. You know less and less about the details of the implementation. Yet the software being built still has to work sustainably. It has to be maintainable. It has to be something the team can run. Not just on the surface.
The first task is therefore to create conditions in which a team can develop software with AI. Questions like these help me:
- Which quality criteria apply, regardless of who or what wrote the code?
- How do we make sure someone on the team can explain and run a solution?
- Where do we deliberately set limits and where do we deliberately let things run?
I described what this can look like in practice in my first article at Enqraft (in German). Within a team, guardrails like these become a shared foundation everyone can rely on.
In my own development work, I notice every day how important these guardrails are. Coding agents will take on more and more tasks on their own. That makes the infrastructure and platform around them all the more important: guardrails within which an agent can operate. So that reliable quality emerges, reproducibly and not by chance. This is exactly what software teams now have to solve for themselves.
2. Clarify responsibility instead of multiplying input
With AI, it takes no time to put a thesis out there. And to write a long treatise on why something has to be exactly this way. But in the end, someone has to decide. And stand behind that decision.
I am convinced by the principles behind Extreme Ownership: every person on the team carries responsibility, for their own results and their own decisions. That was true before AI. And it is true today. But with AI, it becomes all the more important that people on the team take responsibility for individual parts. Because each individual now makes more far-reaching decisions.
This has consequences for leadership, too. The team will produce more and more. In the past, you could still review every line of code. Today you can no longer check that much input and vouch for every detail. That means a loss of control. And it requires growing into a role in which you trust your team.
Trust was always good. Now it is urgently needed. It is the only way a team moves forward and benefits from the new speed.
What does a person need for this? Above all, room to shape things. A topic they own and shape. Without the same topic being worked on in parallel from all sides.
I see exactly this in many places right now and I experience it in my own development work too. Many people contribute ideas and impulses on the same topic. With AI, this input multiplies and every aspect comes with countless options. In the end, nobody can evaluate it anymore. “Perfect is the enemy of good enough” has always been true. Today, without a decision, you simply never reach the end. Everyone is busy and it feels productive. But you only move forward through decisions.
The person who is supposed to decide does not need even more information. They need the ability to act.
3. Provide security and room for new things
Third, it is about showing the team honestly where things stand. I expect AI to become a natural part of software development. Every individual can achieve much more with AI. What this means for the size and tasks of individual teams is still an open question.
What is clear to me, though, is that even with AI someone has to judge whether a solution works and what makes a good solution. That takes software development know-how. Whoever combines it with AI builds different, better software.
This uncertainty weighs on many people. This is exactly where a team needs security. What does that mean in concrete terms? I distinguish between two kinds of security.
Security about the future is something a leader can hardly promise. Nobody knows where this journey will take us. Anyone who promises it anyway loses credibility.
Security in how we treat each other, on the other hand, can be shaped:
- Name the uncertainty. Say openly what is known and what is not.
- Make concerns discussable. Create a space where skepticism has as much room as enthusiasm. Where a concern points to a concrete risk, the team looks for ways to reduce it together. Not every uncertainty can be resolved this way. But it can be addressed openly.
- Make time for learning. Deliberately give people time to experiment with AI. Without every experiment having to deliver a result right away.
For leaders, this is a particular challenge. Many originally come from development themselves. That helps enormously in understanding the team and knowing its challenges. Today they have to relearn, just like every developer, how software development works in the age of AI. But they often lack the time. For a team lead who can still work hands-on, it is manageable. The further removed from the daily work, the harder it gets.
What I observe in leaders who do this well is that they understand for themselves how AI is best used day to day and where its limits are. That only works if you work and experiment with AI a lot yourself.
At the same time, they do not join any of the three camps. They are not the person who constantly raves about the next innovation. But they are also not the one who gives innovation no stage and no motivation. Their role is to bring the camps together. To recognize the different strengths in the team. And to create a space for open discussion, so the team achieves the best result together.
Security and freedom belong together. Freedom, so that innovation can emerge.
Without it turning into careless, indiscriminate behavior. It is not about producing as much as possible, no matter what. It is about keeping the consequences of your own decisions in view.
What remains
Leadership in the age of AI does not mean having all the answers. It means creating a framework in which a team can find them for itself.
Part of that is learning again yourself. Today, leaders have to understand anew how software development works, just like their teams. Those who take the time can help shape the change rather than merely manage it.