Answers Start With the Right Questions
Artificial intelligence is changing the way we build software, but above all it’s changing how we lead teams, make decisions, and shape organizations. Today it’s not enough to know the latest model or the framework of the moment: the real challenge is understanding how to integrate AI into processes, skills, and company culture.
This is exactly the value of a conference like Codemotion Milan 2026. Not just a chance to discover new technologies, but a place to compare notes with people already tackling topics like AI adoption, technical leadership, coding agents, team organization, and the transformation of developers’ work. The goal isn’t to walk away with a list of new tools to try, but with ideas, methods, and concrete approaches to apply starting the very next day.
Because ultimately, the organizations that will adopt AI best won’t be the ones with the most advanced tools, but the ones capable of asking themselves the right questions.
Here are five questions every technical leader should start asking in 2026.
1. Am I really measuring my team’s productivity?
For years we’ve used relatively simple metrics: lead time, throughput, delivery speed, completed stories.
But what happens when an AI agent generates in a few minutes what used to take hours?
Productivity no longer equals the amount of code produced. If anything, the opposite might be true: more automatically generated code means more time spent on review, validation, and architectural decisions.
So the question changes radically:
Am I measuring how much code we produce, or how much value we manage to deliver?
This is one of the topics Alfonso Graziano addresses in his session, Scaling AI Adoption: The Real Challenges of Transforming 300 Engineers, where the focus shifts from tools to the organizational change needed to truly adopt AI at scale.
Takeaway
The most important metric might not be “how fast do we develop,” but “how well do we make decisions.”
2. Am I still hiring developers… or people capable of working with AI?
Hiring processes have stayed almost identical, while developers’ day-to-day work has changed radically. Yet in many interviews we still evaluate candidates with tests designed for a world where AI didn’t exist.
Does it still make sense to ask someone to write code from memory or solve an algorithm with no support at all? Today, in everyday work, a developer works alongside AI tools, questions them, evaluates their suggestions, corrects them, and above all, decides when to trust them and when not to.
Perhaps the real skill to look for is no longer the ability to remember perfect syntax, but how a person collaborates with artificial intelligence: how they frame requests, how they verify answers, how they handle uncertainty, and when they consciously choose to ignore a model’s suggestion because it isn’t the best solution.
That’s the shift in perspective Dennis Nerush proposes in his session, How to Interview Engineers in the AI Era: What I Changed and What Actually Works, which calls for evaluating not just technical skills, but the ability to collaborate with AI while keeping a critical mindset.
Takeaway
In 2026, the competitive edge won’t belong to whoever uses AI, but to whoever uses it with judgment.
3. Which tasks should agents handle… and which must stay human?
The most common mistake when talking about AI is thinking the goal is to automate everything. The most costly one, though, is the opposite extreme: automating nothing out of fear of losing control.
The real challenge isn’t deciding whether to use artificial intelligence, but where it can actually make a difference. There are tasks where AI tools are already proving to be valuable allies: from prototyping to test generation, from documentation to refactoring, all the way to code analysis. These are all tasks that can be accelerated, freeing up developers’ time to focus on what generates the most value.
That said, this doesn’t mean the human role becomes less important. On the contrary, the more AI is able to handle operational tasks, the more weight falls on decisions that require experience, vision, and responsibility. Defining an architecture, weighing technical trade-offs, setting business priorities, or managing project risk remain tasks no model can handle on its own.
This is exactly the reflection Alfonso Fuggetta offers in his session Why Coding Agents Need You More Than Ever, showing how coding agents don’t reduce a developer’s value, but shift it toward higher-value strategic and decision-making work.
Andrey Breslav touches on similar ground with Future of Programming: What’s our place in the world of coding machines? His question isn’t whether programmers will be replaced, but how their role will change in a world where writing code is increasingly handed off to machines. The answer, once again, seems to point in the same direction: less time spent on execution, more time spent designing, validating, and deciding.
Takeaway
The goal isn’t to replace developers, but to free up time for the decisions no model can make in their place.
4. Am I building skills… or dependency on tools?
Many companies are investing in artificial intelligence platforms, convinced that adopting a new tool is enough to accelerate transformation. Far fewer, though, are investing in building an actual adoption method. And that’s exactly where the difference lies between companies that experiment with AI and companies that truly integrate it into how they work.
Adoption isn’t a project that ends with a license purchase — it’s an ongoing journey that requires training, experimentation, and a willingness to question established processes. It means creating spaces where teams can share experiences, compare notes on what works, and learn from mistakes too, turning AI use into a collective skill rather than something left to individuals.
This theme also comes through clearly in Alfonso Graziano’s session, Scaling AI Adoption: The Real Challenges of Transforming 300 Engineers, which focuses less on technology and much more on the organizational change needed to bring hundreds of people along into a new way of working.
Ultimately, tools will change quickly. The real strategic resource will remain an organization’s ability to learn, adapt, and transfer knowledge at the same pace technology evolves.
5. Are we designing AI products… or experiences people actually want to use?
Artificial intelligence is often framed as a technological challenge. In reality, many of the difficulties show up long before the model: when people don’t trust the tool, don’t understand how to use it, or simply stop using it after a few days.
Adoption doesn’t depend only on the quality of the algorithm, but on the experience built around it. An unclear interface, missing feedback, or poorly managed expectations can undermine even the technically best solution.
That’s the point of view Carolina Pinto, UX Designer at King, brings in her session What Candy Crush Can Teach AI Products About Human Behavior. Drawing on her experience with one of the most-used games in the world, she shows how concepts like motivation, progression, feedback, and engagement can make a difference in AI-based products too.
The message is simple: AI adoption isn’t just a matter of technology, it’s above all a matter of people.
Takeaway
An AI product’s success isn’t measured only by what the model can do, but by how much people actually choose to use it.
Conclusion
Perhaps the most important question a technical leader should ask in 2026 has nothing to do with LLMs, coding agents, or the next framework.
It’s this:
Is my team simply using AI tools, or is it learning a new way of doing engineering?
Because the tools will change. The models will evolve. The platforms will multiply.
The difference will be made by organizations capable of building a method: learning fast, experimenting with discipline, and developing the judgment needed to collaborate with AI without delegating what truly matters.
And this is probably the most valuable takeaway that technical leaders, Tech Leads, CTOs, and Engineering Managers can bring home from the conference.
