Stop Using AI Like a Magic Wand and Build Your Dream Team Instead (Here's How)

Transform how you think, write, and solve problems by matching the right AI tools to the right tasks.

Credits: Anni Roenkae

The first time I saw what AI could do, it felt like a cheat code, and that intoxicating thought hit me: "This changes everything."

For a while, it did. I took the courses on prompt engineering and got good at it (and shared the quintessence of how to write great prompts here). I also went deeper and figured out how to build more complex agentic workflows (and yes, there’s a tutorial for that here too).

But then, the magic wore off.

All those rabbit hole AI explorations, all that deep learning, were helpful. They built a necessary foundation. But I’ve come to believe that the biggest leverage, the thing that unlocks the most significant benefits with AI, isn't just about technical skill. It’s about a fundamental mindset shift.

There are two parts to this. First, as I explored in last week’s edition, it’s about reframing AI as your thinking partner, not your thinking replacement. Second, and this is what we’re digging into today, it’s about understanding how to build your perfect AI team.

A few weeks ago, I changed my approach. I stopped looking for that elusive magic wand. Instead, I started building a team.

You’d never hire one person to be your master writer, your strategic analyst, your empathetic coach, and your creative muse all at once. It’s an impossible job.

So, if this sounds familiar, this is for you:

• You’re currently trying to make one LLM do everything, and you’re finding the outputs for specialized tasks are often… mediocre.

• You’re not quite sure which LLM truly excels at the specific tasks relevant to your work.

• You don’t just want another damn list of tools, but you want to understand how to think about these AI partners in relation to your own cognitive strengths and workflow.

Here’s what you’ll walk away with:

• Clear strategies for using different LLMs as intelligent scaffolding for your own learning and thinking, not just as glorified output generators.

• An understanding beyond "use X for Y." We’ll get into why a particular AI might be better suited for a specific task, sometimes even touching on its underlying "cognitive style."

• The ability to achieve genuinely elevated impact: higher quality outputs, deeper insights, and more effective communication in your mission-driven work.

Let's get started and build that AI dream team that actually helps you do your best work.

---

Steering Your AI with Integrity

Before we dive into the "who does what" and the "how-to," there are a few caveats we shouldn’t skip. Because using AI isn’t just about moving faster but about being intentional, and staying the one steering the ship.

Every LLM, from ChatGPT to Claude, Grok to Gemini, carries a distinct vibe, shaped by its training data and design. ChatGPT often plays the agreeable generalist. Claude is ethical and cautious. Grok goes edgy.

These are baked-in design choices that, if you’re not paying attention, you’ll start to mirror without noticing. Just like I suddenly sprinkle in stylistic elements depending on the authors I read, this happens with AI, affecting your decision-making, tone, and logic.

While these tools can generate language that feels thoughtful—even emotionally intelligent—it’s not empathy, but pattern-matching. If your work involves coaching, teaching, or supporting others, use AI to help shape sensitive communication, but never let it replace your judgment, instinct, or lived experience. Your AI has access to insane amounts of data, but it can’t embody wisdom the way you do.

Moreover, as a recent Stanford report highlighted, even the best models are built on biased, messy, inequitable data. They often reproduce stereotypes, reinforce harmful assumptions, or steer your thinking in subtle ways you didn’t choose.

If you’re leading change, working with people, or shaping stories, then you have the inherent responsibility to be the gatekeeper. You’re the editor, the filter, the final word. No AI output should pass through unchallenged, just as no chief doctor would pass the judgment of a doctor in training unchecked.

Think of AI as a brilliant co-pilot, maybe even a whole crew. But you’re still the captain. You decide the direction, ethics, and integrity.

That’s the lens I use when testing new tools. I don’t ask, “Is it fast?” I ask:
“Does this help me think better? Does it align with the kind of deep, ethical work I stand for? Is it amplifying my best or making me lazy?”

Now, with these guidelines in mind, let’s move to the fun part.

---

Recruit and Work With Your AI Dream Team

This is written and double-checked on June 9th, and I assume it will be valid for at least another month.

The key is to be strategic before you even touch a tool. First, ask yourself: What am I trying to achieve here with my own brain? The tool is secondary. The type of thinking required is primary.

Once that's clear, then, and only then, do we ask: "Okay, which AI, if any, can act as a useful partner in this specific thinking process?"

I've structured my AI team around specific roles. For each, I'll outline its designated LLM, the core tasks I assign it (the use cases), and my go-to prompts.