Your Next Direct Report is an AI

Don't Prompt, Lead Instead

We're all obsessed with 'prompt engineering.' And don't get me wrong, it's a useful skill. But focusing only on prompts is like believing the key to great management is just learning how to phrase an email perfectly.

It's a piece of the puzzle, but it's not the whole game. The real work is learning to lead our new AI teammates.

As I was thinking through this - a big realization for me has been that prompting is just a tactic. True leadership is the strategy that actually unlocks AI's potential. So, if your next direct report is an AI, are you ready to be its manager?

Here’s how I’ve started to think about it, and I'd love to hear if this resonates.

Part 1: Thinking Through the AI Employee Lifecycle

If we're really going to treat AI like a teammate, it helps to think about its entire journey with us, just like we would for a new hire. You wouldn't skip the interview or onboarding for a human, so why would we do it for our AI? It's about setting them - and us - up for success.

1. "Hiring" (Model Selection & Evaluation) - This is your first leadership test. You wouldn't hire a senior engineer based on a single trivia question, instead, you "dig wells" in your interviews. So why pick an AI model based on one generic benchmark? That's like hiring based on a SAT score alone. The real "interview" (evals) means testing it against your actual business problems. Give it a case study. See how it handles the nuances of your world before you bring it on board.

2. "Onboarding" (Integration & Training) - A leader’s job is to ramp up their team effectively. For an AI, this means giving it the right context (ie: using RAG/CAG to access your internal knowledge) and the right training (ie: fine-tuning it for your specific language and workflows). This is the equivalent of giving a new hire a laptop, access to the wiki, and a proper orientation (think new hire Launch Plan or Ramp-up). Without it, you can't expect them to thrive.

3. "Assigning Tasks" (Effective Delegation) - This is where prompting fits in—not as a technical trick, but as the leadership act of clear delegation. A great prompt is simply a great project brief. It provides context, defines what success looks like, and clarifies constraints. It’s management 101. In the world of prompt, it comes with a good combination of system prompts and user prompts, along with any context you can provide to support them getting the work done well.

4. "Performance Management" (Coaching & Growth) - As a leader, you provides feedback to make the entire system smarter over time. Every "thumbs down," every correction, every time you have to rephrase a prompt—that's a coaching session. It's how you develop your teammate.

5. "Career Growth" (Increasing Responsibility) - As an AI proves itself reliable, a good leader gives it more responsibility. You graduate it from simple, low-stakes tasks to more complex, autonomous workflows. You wouldn't have a star performer just answer emails forever, right? You find ways to leverage their growing skills.

6. "Succession Planning" (Strategic Transition) - Now the reality is, a better model is always just around the corner and we should always have a great succession plan for the models. When it arrives, a leader doesn’t just unplug the old one and create chaos. They manage the transition, ensuring workflows are migrated, knowledge is preserved, and the team doesn't miss a beat.

Part 2: The Situational AI Leadership Ladder

Okay, so you've 'hired' and 'onboarded' your AI, and thought through their lifecycle. Now what? The day-to-day management is where things get interesting. It's not a one-size-fits-all deal.

This reminds me so much of the classic Situational Leadership model from Hersey and Blanchard. Just like with people, the most effective AI leadership is situational.

via Situational Leadership® and Performance Readiness®

A good leader adapts their style based on the AI's "development level" for any given task. The mapping is surprisingly direct:

Level 1: The AI Novice (You are a DIRECTOR)

• The AI's State: It has low competence for your specific task but high "willingness." It's a blank slate, ready for a command.

• Your Style: Directing (S1). This is where you are highly directive. You give clear, step-by-step instructions. This is the home of the simple, straightforward prompt: "Tell me X," "Format this Y."

Level 2: The AI Apprentice (You are a COACH)

• The AI's State: It has some competence but is unreliable, making mistakes and needing refinement. It's trying, but it's not quite there.

• Your Style: Coaching (S2). Here you're still highly directive, but you add a strong layer of support. You provide detailed instructions but also engage in that crucial back-and-forth feedback loop to guide it toward a better outcome.

Level 3: The Capable Contributor (You are a SUPPORTER)

• The AI's State: It's highly skilled at the task but might lack the business context or confidence to operate without a final check.

• Your Style: Supporting (S3). Now you can back off on the direction. You trust the AI to do the heavy lifting but provide that crucial layer of review and encouragement. This is your classic "trust but verify" stage.

Level 4: The Autonomous Partner (You are a DELEGATOR)

• The AI's State: The AI is fully trusted, reliable, and capable of handling the task from start to finish.

• Your Style: Delegating (S4). You can now be low on direction and low on support. You delegate the outcome, trusting the system to manage the process on its own. Your job is to check the final result, not every step along the way.

Now, I'm not a doubter on whether AI will eventually get to Level 4, but hey we all heard from Sam Altman that AGI is coming.

Part 3: The AI Leadership: From Prompting to Leading

This brings us to the heart of the matter. The difference between tactical prompting and strategic leadership isn't just semantics; it's a fundamental shift in how we create value. It's the difference between being an operator and being an orchestrator.

Here’s what I mean, principle by principle:

1. Set Clear Goals & Define Success - A prompter asks for an answer. A leader drives toward an outcome. A prompter asks, "What were our sales last quarter?" A leader commands, "Analyze our last four quarters of sales data, identify the top three drivers of growth, and draft an email to the sales team highlighting these wins and proposing one action item to double down on." The first is data retrieval; the second is strategic action.

2. Coach for Performance - A prompter gives a command. A leader provides feedback. A prompter treats AI like a vending machine: you put in a coin, you get a candy bar. If it's the wrong one, you might shake the machine and try again. A leader treats AI like an apprentice. They know that every correction and clarification is an investment that makes the apprentice more capable for the next task.

3. Master Delegation (Trust but Verify) - A prompter uses AI to do the thinking. A leader designs systems that help them think. A prompter uses AI like a calculator. A leader uses it like a whiteboarding partner, and write to think. They design workflows where an AI can generate a first draft, run it against a compliance checklist, and then flag the key areas that require human judgment. That’s building a process, not just getting an answer.

They don't zero-shot the prompts, they create auditable processes and output step by step to help their team of AI tools and agents get better without blindly trust them.

4. Adapt Your Approach - A prompter uses one style for all tasks. A leader adapts their management style. Just as we discussed, they know when to be a Director (for a simple task), a Coach (for a complex draft), or a Delegator (for a trusted, autonomous agent). This situational awareness is the hallmark of effective leadership.

5. Cultivate a Partnership - A prompter has transactional conversations. A leader builds a relationship. For a prompter, every chat is a blank slate. For a leader, every interaction builds on the last. By intentionally building context over time, you create a partner that remembers, "Last time we talked about this project, we decided to focus on the European market first." This transforms a generic tool into a hyper-personalized expert.

6. Empower Autonomy - A prompter manages one task. A leader orchestrates a team. A prompter focuses on a single AI for a single task. A leader thinks like a systems architect, asking, "How can my research agent feed insights to my marketing copy agent, which then hands off drafts to my legal review agent?" This is where you start conducting a symphony, not just playing one instrument.

Your Call to AI Leadership

The future doesn't belong to the best prompters; it belongs to the best leaders. Research from places like Harvard is already showing that an individual with AI can be as productive as an entire team without it. The multiplier effect is massive.

But here’s the thing I keep coming back to: the best AI won't save a bad manager.

The companies that thrive will be those who recognize that the real skill gap isn't technical, it's managerial. They will invest in developing AI leadership.

Your new digital direct report is waiting for its first 1:1. The question is no longer what you will ask it, but how you will lead it.

So, what's the first thing you'll do as an AI leader?

#YourNextDirectReportIsAnAI #AI #Leadership #Management #FutureOfWork #GenerativeAI