If you've worked with me, you know I've always believed in leadership at all levels. The best engineering teams I've built didn't wait for permission. ICs made decisions, took ownership, and led from wherever they sat.
But that was a philosophy. A preference. A "nice to have" for high-performing teams.
In the age of AI agents, it's no longer optional. Every person on your team is now a leader, whether they realize it or not.
What Changed
The old model was simple: an engineer receives a task, executes the task, and ships the code. The new model is different: an engineer receives a task, delegates pieces to AI, reviews and directs the AI's output, then integrates the result.
That second workflow is management. Task scoping. Delegation. Review. Quality control. Judgment calls about what's good enough and what needs another pass.
Even your most junior engineers are now "managing" AI agents. The moment someone started using Claude Code, Copilot, or Cursor, they became a manager of non-human workers. Their title didn't change, but their job did.
This matters because the skills that make someone good at working with AI are the same skills we've always associated with leadership. And now everyone needs them.
Five Leadership Skills That Used to Be Optional
Clear and specific communication. When you delegate to a human teammate, they can ask clarifying questions. They can resolve ambiguity through conversation. AI agents can't. They take your instructions literally and run with them. Every vague prompt produces garbage output. Every unclear requirement produces code that technically works but misses the point. The ability to communicate with precision, which we used to develop in senior engineers over years, is now table stakes for anyone using AI tools. (Hot take: the next wave of LLM coding tools will have this built in. The best vibe-engineers are already having agents ask clarifying questions before writing code.)
Judgment and taste. AI generates options fast. It can produce five different implementations in a fraction of the time it takes to prototype one. But knowing which option is right? That requires judgment. Understanding what "good" looks like for this specific context, this specific codebase, this specific user need. This used to be a senior skill, something you developed after years of seeing what worked and what didn't. Now it's required from day one.
Scope management. AI will happily over-engineer a solution or under-deliver on requirements. It has no sense of "good enough." Someone has to decide what's in scope, what's out, when to stop iterating. That someone is now every engineer on your team, making these calls dozens of times a day.
Accountability without control. Here's an uncomfortable truth: you're now responsible for code you didn't write. You didn't type those lines, but you shipped them. When something breaks in production, "the AI wrote it" isn't an answer. Taking ownership of outcomes you didn't directly produce is a leadership skill. And it's now required for everyone using AI tools.
Strategic prioritization. With AI, the constraint isn't "how fast can we code this?" The constraint is "what should we code?" When you can delegate unlimited tasks to AI agents, the bottleneck becomes deciding what matters. Every engineer now needs to think about leverage and impact, not just execution.
If Everyone's a Leader, What Are Managers For?
This isn't about eliminating management. It's about elevating expectations across the board.
The manager's job shifts from directing work to developing leaders. Instead of telling people what to do, you're teaching them how to delegate effectively. Instead of reviewing code, you're reviewing how they reviewed the AI's output. What did they catch? What did they accept? Why?
Senior ICs become force multipliers in a different way. Not just through the code they produce, but through the judgment they model. When a senior engineer explains why they rejected an AI suggestion, they're teaching the entire team what good looks like.
The org chart doesn't flatten. But the leadership expectation does. Every level is now expected to exercise judgment, make decisions, and take ownership. That was always true for the best teams. AI just made it true for everyone.
What This Means for Hiring
The interview questions that mattered five years ago are becoming less predictive. "Can you solve this coding problem?" tests something AI handles well. What you need to know is different: Can they scope a problem? Can they delegate effectively? Can they judge quality?
The traits we used to screen for at the Staff+ level now matter at every level. Ownership. Communication. Systems thinking. Judgment. These aren't "senior" skills anymore. They're baseline requirements for working effectively with AI.
Here's a thought that might be uncomfortable: the best junior engineers of the AI era will look more like product managers than the junior engineers of five years ago. They'll spend less time writing code and more time defining what code should be written, reviewing output, and making judgment calls. That's a fundamental shift in what we're hiring for.
What Actually Helps
If leadership at all levels is now mandatory, how do you build it?
Make delegation explicit. Before engineers start coding (or start AI coding), have them articulate their task breakdown. What are they delegating? What are they keeping? Why? Practice the skill of scoping and delegating before it becomes invisible habit.
Review the review. Don't just review the code. Ask how they reviewed the AI's output. What alternatives did the AI suggest? Why did they pick this one? What did they change? This surfaces the judgment that's now the core of the job.
Create ownership, not assignments. Give people problems, not tasks. Let them figure out how to use AI to solve them. The engineer who decides how to break down the work is exercising leadership. The engineer who just executes assigned tasks isn't learning the skills that matter.
Reward judgment, not just output. When someone says "the AI suggested X, but I knew Y was right because..." that's the skill. Celebrate it. Make it visible. The teams that develop strong judgment will outperform the teams that blindly accept AI output.
Model it yourself. Share your own AI delegation decisions with your team. Show them how you scope problems, how you review output, when you accept suggestions and when you push back. Leadership skills are learned by watching leaders.
The Question Isn't Whether Your Team Can Code
Producing code is no longer the constraint. Producing the right code still is. But the leadership bar just got higher. Every engineer making dozens of judgment calls a day, taking ownership of output they didn't directly produce, deciding what's worth doing and what isn't.
You can't hire your way out of this. You can't just recruit people who already have these skills. You have to develop them. The companies that figure this out will build teams where trust and ownership are the default. The companies that don't will wonder why their AI investments aren't paying off.
The best teams I work with aren't the ones with the best AI tools. They're the ones where everyone, from the newest hire to the most senior architect, operates like a leader. That was always true. AI just made it impossible to ignore.
AI won't replace your engineers. But it will expose which ones can lead and which ones were just executing. The age of AI didn't eliminate the need for engineers. It made every engineer's leadership skills visible.
Frequently Asked Questions
Ready to build leadership at all levels in your engineering organization?
Let's Talk →