Wunderdog Blog

From Problem Solver to Manager of Agents: The Real Shift in Agentic Development

Written by Ville Komulainen CEO & Founder | Sep 16, 2026, 10:58:46 AM

Key takeaways

  • The real change in agentic development is not speed of code, but the shift from solving problems to defining them.
  • Engineers are trained for a tight solve-and-verify feedback loop. Agentic work replaces it with framing, delegating and reviewing — managerial work.
  • The skill that now matters most: defining a problem well enough that it can be implemented autonomously, and defining what "good" looks like.

Agentic development is changing what engineers are actually paid to do. The biggest change in software development is not about AI writing code faster. It is about the shift from solving problems to defining them. From doer to manager.

 

What changes in the engineer's day-to-day work

Most of us engineers are trained to solve problems. We are good at it. We punch holes into plans, find edge cases, and make things work. That feedback loop is tight and rewarding. You write code, it works, your brain gives you endorphins.

But with agentic development, the job starts to look different. You define what to do. You set the intent. You decide the ambition level for quality. You review output instead of producing it. That is managerial work, and most of us never signed up for that.

And that is the cultural and way-of-working revolution that needs as much attention as the technical and methodological one.

 

How do you make the shift from doer to manager?

Why were these pages performing so poorly? During our audit, we identified several causes that were all affecting the website’s load times, and thus the overall user experience:

  • Learn to love the problem as much as the solution.

  • Embrace the uncertainty of the messy problem space.
  • Make the framing of the problem the problem you want to crack.

The hardest part is not learning new tools or methods. It is accepting that you are now a manager and you need to tell your teammates (the agents) what to do and why. And let them sort out the how, within your engineering guidelines and constraints.

Luckily, this managerial work can also be turned into a solvable problem. The non-trivial problems to solve are: How to define the problem so well that anyone can implement it autonomously? How to define success criteria for the agent, what good looks like in all its relevant dimensions?