The Hands-On AI Leader: Why Delegation Will End Your Career

New HBR research reveals why the best CEOs architect execution—and what that means for leading in the AI era

New Harvard Business Review research reveals the surprising secret of top-performing CEOs—and it changes everything about how you should approach AI strategy.

The conventional wisdom is clear: CEOs should focus on vision and strategy, not execution. Stay out of the weeds. Think big picture. Delegate the “how” to your teams.

The data says otherwise.

In their November-December 2025 HBR article “The Surprising Success of Hands-On Leaders,” Scott Cook (Intuit founder) and Nitin Nohria (former Harvard Business School Dean) studied four of the world’s highest-performing companies: Amazon, Danaher, RELX, and Toyota. Their finding challenges decades of leadership orthodoxy:

The best CEOs don’t just set strategy—they architect how work gets done. They obsess over execution methods. They design systems. They teach by doing. Jeff Bezos packed boxes. Eiji Toyoda ran experiments on GM factory floors. Larry Culp forces Danaher executives into week-long kaizen problem-solving sessions where they implement solutions themselves.

These leaders reject the “delegator-in-chief” model. They get their hands dirty—not to micromanage, but to model standards, build systems, and establish behavioral norms that work when they’re not in the room.

Here’s why this matters more for AI than anything else you’ll face:

In every other domain, you can delegate technical work because the capabilities are stable. Your CTO can own the infrastructure roadmap because servers and networks don’t fundamentally change every quarter. Your VP of Sales can own the go-to-market because sales processes compound over years.

AI is different. The capabilities are evolving monthly. What was impossible in January is trivial by March. What required a data science team last year can now be prototyped by a single person with the right tools in an afternoon.

When you delegate AI strategy without hands-on understanding, you create three critical blind spots:

1. You can’t distinguish between what’s genuinely hard and what’s a capability gap on your team. Is your team struggling with a complex technical challenge—or are they three months behind the state of the art?

2. You can’t architect the systems that enable everyone else to excel with AI. Just as Bezos designed two-pizza teams and six-page memos to unlock Amazon’s velocity, you need to design the workflows, processes, and methods that let your organization build with AI at scale. You can’t architect what you don’t understand.

3. You lack the credibility to lead in the AI era. Your technical teams know immediately whether you understand what’s possible. Without hands-on experience, every strategy meeting becomes theater—you’re approving plans you can’t truly evaluate.

---

---

The Five Principles of Hands-On AI Leadership

The HBR research identified five principles that guide hands-on leaders. Each one maps directly to AI leadership:

1. Obsess Over the Metrics That Customers Value

Erik Engstrom transformed RELX by coaching every employee to obsess over “customer value”—the actual benefit customers realize from products. He asks relentlessly: How does the customer measure value? How do we know? How do we measure that?

For AI leaders: You must know which AI capabilities actually move business metrics. Not which ones sound impressive in demos—which ones create measurable value. And you can’t know that without hands-on experimentation.

When you build an AI-powered analysis yourself, you feel where the friction is. You discover that the insight is brilliant but arrives too late to be actionable. Or that the accuracy is perfect but the output format is unusable. These aren’t things you learn from PowerPoint reviews.

2. Architect the Way Work Gets Done

Bezos didn’t just set Amazon’s strategy—he redesigned how product teams worked. Two-pizza teams. No PowerPoint, only narrative memos. Decoupled services. Two-way-door decisions. He built the “how” that enabled autonomous, fast-moving teams.

For AI leaders: You can’t redesign workflows around AI if you don’t understand what AI can actually do. Every major business process in your organization—sales, operations, customer service, product development—will be rebuilt around AI capabilities in the next 24 months.

If you’re not hands-on, you’re depending entirely on your teams to imagine that future. That’s not leadership. That’s hope.

3. Use Experiments to Make Decisions

Toyota’s culture is built on experimentation. Frontline workers redesign their own workflows. Plant managers test their proposals against alternatives from subordinates. Even CEOs run experiments—Toyoda didn’t declare Toyota would manufacture in the US, he persuaded GM to let him test it at a shuttered California plant first.

For AI leaders: AI strategy requires running experiments yourself, not just reviewing decks about experiments. You need to feel the difference between a well-crafted prompt and a lazy one. You need to experience the shock of an AI agent completing in 10 minutes what used to take your team two days. You need to hit the failure modes yourself—the hallucinations, the edge cases, the brittleness.

This isn’t about becoming a prompt engineer. It’s about developing the judgment to know what’s possible and what’s bullshit.

4. Lead by Teaching the Tool Kit

Danaher’s approach is radical: when executives are hired from outside, they spend two months in boot camp learning Danaher’s tools—voice of the customer, value stream mapping, kaizen problem-solving—before taking their actual roles. As Larry Culp puts it: “We force division presidents to develop a command of the how so that they can teach the how.”

For AI leaders: Your team needs to see you prompting, debugging, iterating. They need to watch you struggle with a complex workflow and work through it. They need to hear you articulate why one approach works and another doesn’t.

A few months ago, an SVP of Revenue found me via ChatGPT. She wanted to get hands-on with AI so she could lead her team to adopt AI. She knew she needed to build expertise and be the tip of the spear to lead AI adoption and create a culture for learning. She recruited her CEO and several other members of the leadership team to join the cohort. The hands-on experience quickly expanded their understanding of what was possible, and gave them the expertise to evaluate several AI tools to guide their technology infrastructure selection. This also led us to design a custom, six-week hands-on AI course for the organization.

Teaching requires doing. You can’t teach what you haven’t done.

5. Strive to Be Better, Faster, Cheaper—Every Year, Forever

At RELX, Engstrom distilled continuous improvement into a mantra: “Better, faster, cheaper—every year, forever.” Teams must show how each iteration delivers more customer value than the last. The bar never stops rising.

For AI leaders: AI capabilities evolve monthly. What’s cutting-edge today is table stakes in 90 days. Hands-on practice isn’t a phase you complete—it’s how you maintain technical credibility and strategic vision in a domain where the frontier moves constantly.

The executives who win won’t be those who “learned AI in 2024.” They’ll be those who are building with the latest capabilities every single quarter.

The Identity Shift: From Strategist to Builder

Here’s what this really requires: a fundamental transformation of how you see yourself as a leader.

For three decades, you’ve been told to stop acting like a COO. To elevate. To think strategically. That advice built your career.

And it will end it if you apply it to AI.

The highest-performing leaders don’t choose between strategy and execution—they architect the execution systems that make strategy real. In the AI era, “how work gets done” means “how we build with AI.”

You must see yourself as an AI-savvy builder who infuses AI into all business processes, products, and services. Not someone who “has an AI strategy.” Someone who builds with AI.

What This Looks Like in Practice

A few months ago, a CEO of a stealth startup joined my Hands-on AI for Leaders cohort course. He had a vision for a mobile app that would be core to his new company. But he’d internalized the belief that building a mobile app was beyond him—that was work for developers, not founders.

During a live session, I demonstrated vibe-coding with Cursor and Claude Code—using AI to translate ideas into working prototypes through conversation and iteration. Something clicked. He saw that the barrier wasn’t technical complexity. It was a self-limiting belief.

He started building during the session. Requirements. Prototypes. Iterations.

A few months later, he had a working mobile app for his company.

That would never have happened if he’d stayed in “strategy-only” mode. It required him to shed the identity of “executive who delegates technology” and adopt the identity of “builder who uses AI to manifest vision.”

The gap wasn’t his coding ability. The gap was his willingness to get hands-on.

The Delegation Trap

The most dangerous words in the C-suite today: “My team handles the AI stuff.”

That sentence means: You can’t evaluate AI initiatives. You can’t spot the difference between cutting-edge and outdated. You can’t ask the questions that expose bullshit. And critically, you can’t architect the systems that enable your organization to build with AI at scale.

You’ve outsourced your ability to lead in the defining technology shift of your career.

When you say “AI is too technical for me,” here’s what you’re really saying: “I’ve chosen to remain dependent on others to shape my organization’s AI future.”

That’s a choice. But is it the choice a leader makes?

You don’t need to become a machine learning engineer. You don’t need to understand transformer architecture. You need to build enough with AI to:

• See what’s possible and infuse that into your vision

• Ask the right questions when your team proposes AI initiatives

• Distinguish genuine technical challenges from capability gaps

• Architect the systems and processes that enable your organization to build with AI at scale

• Teach by doing rather than by PowerPoint

Your team doesn’t need another executive with AI opinions. They need a leader who can show them what good looks like.

Your Move

The HBR research is unambiguous: hands-on leadership isn’t micromanaging—it’s architecting systems, modeling standards, and teaching through doing. It’s what separates the highest-performing companies from everyone else.

In the AI era, this isn’t optional. The capabilities are moving too fast. The strategic implications are too profound. The execution details matter too much.

This week, spend two hours building something with AI. Not reviewing. Not approving. Building.

• A workflow that automates part of your analysis process

• An AI agent that handles a repetitive task for your team

• A prototype of a product feature you’ve been discussing

• A data analysis that answers a strategic question

Feel the difference between theoretical understanding and hands-on knowledge. Notice what’s surprisingly easy and what’s genuinely hard. Experience the moment when possibility expands.

That’s the difference between executives who will lead in the AI era and those who will be led by it.

Quick reflection: When was the last time you personally built something—not reviewed a deck, not approved a budget, but actually built something? If it’s been more than a month, you’re in delegation mode. If it’s been more than six months, you’re in danger.

Discussion: What’s your biggest barrier to getting hands-on with AI—time, perceived complexity, or something else? Let me know in the comments.

Want a structured path to building AI leadership capability hands-on? Join our next Hands-on Agentic AI for Leaders cohort at maven.com/james-gray/hands-on-ai-for-leaders. Save 25% during the Maven FastTrack promotion now through Monday, November 24th. We start with the fundamentals and build up to agentic workflows that boost your productivity and impact—all through doing, not just discussing.

---