TL;DR
- I built (and now manage) 21 AI agents over the past year. It’s changed how I lead, both people and machines
- The gap between your thinking and your output is never more visible than when you’re directing an AI agent
- The military figured this out decades ago. They have precise technical orders, demand mastery before progression, and do mission planning from multiple angles. Those are elements in the AI Agent Playbook.
- Leadership clarity, accountability, and delegation discipline are the same muscle, whether your direct report is human or artificial
- The leaders who win this era won’t be the most technically proficient. They’ll be the ones who use AI to get sharper, not just faster.
I want to tell you something personal.
Over the past year, one thing has driven more of my growth than any book I’ve read, any coach I’ve hired, or any conference I’ve attended.
I built and now manage 21 different AI agents.
Not because it was trendy. Not because someone told me to automate everything. I did it because I was drowning. Too many systems. Security gaps. Juggling requests that don’t impact the bottom line. Too many demands for my time. I needed leverage. And I found it.
But here’s what I didn’t expect.
My AI agents didn’t just give me leverage. They handed me a mirror.
AI Isn’t the Threat. The Mirror Is.
Leadership isn’t a technology problem. It never was.
But leadership clarity has been hidden for most careers behind a layer of polite effort and human goodwill. Your best team members compensate for your ambiguity. They read between the lines. They ask clarifying questions. They cover your drift with their loyalty.
AI agents don’t do any of that.
Vague backplane or prompt. Vague output.
Fuzzy strategy. Fuzzy execution.
There’s no covering for you. There’s no effort-based goodwill. There is only the clarity (or the lack of it) that you bring to the work.
That’s not a technology problem. That’s a leadership problem that technology just made impossible to hide.
What 21 Agents Taught Me About the Gap
When I sat down to build my first agent, a deal research stakeholder & alignment agent, I expected a technology challenge. What I got was a self-assessment I hadn’t asked for.
The agent kept producing output I couldn’t use. Not because the model was bad. Because my instructions were over generalized and vague.
I knew what I wanted. I thought I’d communicated it. The agent showed me I hadn’t.
That’s the gap. And it’s the same gap that lives between you and your team right now.
A landmark National Bureau of Economic Research (NBER) project found that leadership performance with AI agents strongly predicts effectiveness with human teams, with a correlation of 0.81.
The same skills that make you good at directing AI make you good at directing people.
Asking the right questions.
Giving context.
Synthesizing information into decisions. Closing the loop.
If your AI agent can’t execute your strategy, your team probably can’t either.
Here’s a quicksand trap sales managers across the country are deep in right now.
They say to their sales teams
- You better use AI. Your job depends on it. Then…
- “Use [your favorite model here…maybe its ChatGPT, Copilot, Claude, Gemini…] AI to help you prepare for your sales meetings.”
No kidding…
It’s a sales leader giving a lazy edict.
Of course you should use AI. But how?
Trust me. I was living this and getting meh answers.
Then, I started learning from prompt engineers, and behavioral psychologists.
I told AI places to look for deeper insights. Tools to interpret visual data I screen scraped from images found online. Sources not every sales person is tapping into.
I looked for correlation data by bridging disparate data sources most sellers never look at, or know about. Then I look for subtle cues in between that help me shortcut a lot of effort. Then I built resources, skills, tools, and instructions fine tuned to a bank of ICP’s dynamically updated weekly. Then I automated it.
Trust me, when I am selling today, I have more potent firepower to help me boost my win percentage.
With your human sales team, they’ve just been too loyal to tell you.
The Five Leadership Muscles AI Exposed in Me
1. Clarity of Vision. Not the Vision Statement. The Actual Vision.
When I write a prompt for an agent, I can’t just say “research this market.” I have to define what great looks like. What format. What depth. What the output will be used for. What I’m not looking for.
That’s not a prompting skill. That’s a leadership skill.
The leaders who thrive in this era are building something like an updated variation on EOS (Entrepreneurial Operating System) documenting every process, player, tool, outcome, KPI, SOP’s, and more. It’s so much more than a prompt library, or skill.md instructions. Rather, it’s a living document how they actually think strategically. It becomes a leadership operating system that works for AI and for people.
(BTW, I’m going to share next week the gritty details in a webinar and a new article that’s shows more of my journey. Tools like Braintrust clarifies like a $500/hr McKinsey consultant in this the AI arena).
The discipline of writing tight, context-rich direction for AI forces you to articulate your strategy with a precision that transforms how you also lead humans. You develop clarity that is unmistakable.
2. Decision Ownership. Insights Without Owners Are Just Noise
I’ve watched leaders pour AI into their workflows and then wonder why nothing changes. The reason is almost always the same: nobody owns the output.
AI can produce recommendations. It cannot own an outcome. The moment you treat an AI-generated insight as “the team’s” or “the system’s” instead of naming a human who owns the decision, you’ve created expensive exhaust.
Every one of my 21 agents has a human attached to its output. No ambiguity. No diffusion of responsibility. I partner with a technology capability that goes more granular than that, showing you things you should be asking of people but never had a need.
<aside> 💡
Radical ownership isn’t just a veteran value. It’s the asymmetric edge in an AI-driven world.
</aside>
3. Delegation Discipline. Cognitive Triage Is the Real Skill
Managing AI agents forced me to ask a question I’d avoided when working with my human team: What is this task actually worth, and who should own it?
Researchers call this Cognitive Triage. Deliberately deciding which tasks demand human judgment and which should be delegated to machines. The skill transfers directly to managing people.
How many tasks on your plate right now could be done by someone else if you just had the discipline to hand them over?
I redesigned my week around this question. The tasks I kept are the ones only I can do: relationship intelligence, ethical judgment, vision-setting, and being a present father and husband. Everything else is a delegation opportunity.
4. Emotional Intelligence — The One Thing AI Can’t Replicate
Here’s the asymmetric leadership opportunity hiding inside all of this.
AI handles execution. It processes faster than any team member. It doesn’t need rest, motivation, or recognition.
What it cannot do is inspire. It cannot read the room. It cannot hold a grieving employee, celebrate life’s joys, navigate a conflict, or earn the kind of trust that makes a team run through walls for a mission.
As AI absorbs execution, leadership becomes less about having answers and more about creating the conditions in which people thrive. That’s an emotional intelligence game. And it’s the most undervalued skill in the room right now.
Invest disproportionately in EQ. The leaders who win in this era won’t be the most technically proficient. They’ll be the ones people choose to follow.
5. Strategic Persistence . Adoption Without Reckoning Is Wasted Potential
A 2025 AI Readiness study found that nearly 80% of business leaders now personally use AI tools. But fundamental attitudes about its transformative potential have barely changed. Leaders are adopting the tools without truly reckoning with what they mean.
That’s the persistence gap. And it’s widening every week.
I see it constantly. Leaders download AI tools and use them for email drafts. Meanwhile, BCG data shows AI-leading organizations are already generating double the revenue growth and cost savings compared to laggards.
Don’t just use AI. Study what it reveals about your organization. Every friction point is a diagnostic. Every failed prompt is a leadership signal.
What the Air Force Taught Me About Building AI Agents
A lot of people look at the military and see a fallback. A place for people who didn’t have other options.
That couldn’t be more wrong.
The military is one of the most sophisticated mastery systems ever built. And when I started building AI agents, I kept thinking: I’ve seen this before.
The Air Force takes 18-year-olds. No experience. No credentials. And it trains them to maintain, operate, and troubleshoot the world’s most complex equipment. Aircraft worth hundreds of millions of dollars. Systems where a single error costs lives.
How? Not through inspiration. Not through culture decks and vision statements.
Through three disciplines that most organizations have completely abandoned.
1. The Technical Order
Before any maintainer touches an aircraft, there is a document. Precise. Sequenced. System-aware. It says: here’s the task, here’s why it matters, here’s how it connects to every related system and subassembly, here’s exactly what you do and in what order.
Nothing is assumed. Nothing is implied. Nothing is left to interpretation.
That’s a prompt library. That’s what a well-built AI agent instruction set looks like. And that’s what’s missing in most organizations when they try to delegate, to a person or to a machine.
Your team is guessing at your technical order right now. Most of them have been guessing for years.
2. The Mission Plan
Before a mission, you study the forward edge of the battlefield from multiple perspectives. You don’t plan from one vantage point. You look at the terrain, the threats, the interdependencies, the contingencies. You anticipate what can go wrong before it does.
Most leaders deploy AI and people without doing that reconnaissance. They hand someone a task and expect execution. What they get is a ground-level view of a problem that required an aerial one.
Building agents forced me back into that discipline. Before I deploy any agent, I map the mission area: What is this agent trying to accomplish? What adjacent systems does it touch? Where are the failure points? What does success actually look like from the outside?
That’s not AI best practice. That’s military planning applied to modern operations.
3. Mastery Before Progression
This one is the hardest for civilian organizations to accept.
The military doesn’t let you advance until you’ve demonstrated the skill. Period. You don’t move to the next task until you’ve proven you can execute the current one. No exceptions for confidence. No shortcuts for time-in-seat.
Most organizations do the opposite. They promote on potential and manage on hope.
AI agents expose this instantly. An agent that hasn’t mastered its core task doesn’t get more responsibility. It gets more failure. You see the gap immediately. No soft landing. No grace period.
I’ve applied this same standard back to my human team. Not harshly. Honestly. What have we actually mastered? Where are we progressing on hope instead of proven capability?
The answers changed how I build, hire, and develop people.
Here’s what I want leaders to understand, especially those who dismiss the military as a relic or a last resort.
The military built a human development system that takes raw, untested people and produces disciplined, high-performing operators in months. Not years. Months.
It did it through clarity of instruction, mastery-based progression, and mission-level perspective.
That’s not a military advantage. That’s a leadership advantage that belongs to anyone willing to apply it.
AI agents just gave me a reason to remember it.
Similar. Different. Here’s the Honest Breakdown.
After a year of managing 21 agents alongside a human team, here’s what I know.
What’s the same:
- Both require clear direction and context
- Both perform in proportion to the quality of your thinking
- Both reflect your accountability structure back at you
- Both get better when you invest in the feedback loop
What’s different:
- Your team brings initiative, judgment, and emotional labor. Your agents bring consistency and scale.
- Your team needs inspiration, recognition, and trust. Your agents need well-designed instructions and iteration.
- Your team will eventually tell you when something’s wrong. Your agents will silently produce the wrong thing forever, until you notice and fix them.
- Your team can surprise you with creativity. Your agents will optimize inside the boundaries you give them.
The leaders who understand both relationships and invest in leading both well are the ones building compounding advantages right now.
The Asymmetric Math
Here’s what I want you to take from this.
A clear thinker with an AI agent outperforms a vague thinker with an entire department. That’s not an exaggeration. That’s what I’ve watched play out in my own company over the past twelve months. BCG confirms it as well.
And yet clarity, judgment, and empathy become scarcer as AI handles more of the execution. The human element doesn’t diminish. It becomes the differentiator.
21 agents later, I’m not better because I have more automation.
I’m a better leader because the mirror showed me the gaps I didn’t know I had. And I stopped looking away from them.
That’s asymmetric math. And it’s available to any leader willing to look.
How is AI revealing your leadership gaps — and what opportunities are you finding inside them?
Reply to this issue or connect with me on LinkedIn. I read every response.
—Joe
