Casting an AI Team Is Only the Beginning

I’ve spent much of my career figuring out how to get different people, with different strengths, to work together effectively.

In the agency world, we called it casting. You looked at the assignment, considered the skills and perspectives needed, and assembled the right team for the work.

But casting is only the beginning. A great team doesn’t guarantee a great outcome. Whether you’re assembling a team from scratch or inheriting one, team leadership is about understanding each person’s strengths, bringing complementary capabilities together, and creating the conditions to enable the best work to get done. It’s also about recognizing when something isn’t working and knowing what to change.

Now that I’m building my own ventures through WeAreDestico, I’m applying much of that same thinking to something new: building and leading my own AI team. What’s surprised me is how much of what I’ve learned about leading people applies to leading AI. The players are different, but the responsibility for making the team work well is remarkably familiar.

Casting my AI team

My team consists of three AI tools, each with a defined role. I didn’t start with an organizational chart or a predetermined belief about which technology was best. The roles emerged as I got to know each tool better through working with them, discovered strengths and limitations, and adjusted responsibilities around what I needed to accomplish. It’s much the same way I get to know and understand a human team.

Now, ChatGPT is my strategist and creative partner. Through conversations ranging from business strategy to travel planning and personal interests, it’s accumulated context about how I think and what matters to me. That makes it particularly useful for shaping ideas and creative direction.

Gemini is my researcher. It started as my strategic collaborator on one of my ventures, but as I experimented with different tools and ways of working, I found value in separating research from strategy. Today, Gemini focuses on gathering information that informs my decisions.

And the hub that brings it all together is Claude, my chief of staff and execution partner. I’ve found it particularly effective at maintaining documents, preserving decisions and translating direction into tangible work. It helps manage my projects, priorities and documentation, while Claude Design handles visual production from creative briefs.

At a recent Women + AI networking dinner in Los Angeles, I was talking with another attendee about my AI team. Her primary strategic collaborator is Claude, while mine is ChatGPT. Different casts, but neither one is necessarily better than the other. We’re each organizing our teams around our own objectives and ways of working. The strongest person for one assignment isn’t necessarily right for another, and the same applies to AI.

Knowing when to recast

A good example of recasting my AI team came while developing WeAreDestico’s brand identity. I worked with Claude to create a creative brief, then passed it to Claude Design to develop visual concepts. The results were competent, but something was missing. The work didn’t have the emotional pull or personality I was looking for. It reflected the business, but it didn’t quite capture me.

I’ve encountered that situation too with human-led creative teams. The instinct can be to keep pushing the people doing the execution, asking for another round of concepts or more variations. But sometimes the problem isn’t the execution, it’s the direction. In this case, I realized I’d cast the wrong collaborator for developing the brief.

I shifted that responsibility to ChatGPT, which had a richer understanding of my professional experience, my judgment and creative sensibilities, and the mix of candor and playfulness that comes through in how I express myself. Together, we developed a brief that captured more of the character I wanted the brand to express. The execution stayed with Claude Design, but the creative direction was now informed by a collaborator who understood me better.

When work isn’t meeting expectations, I look at where the problem might be. Is it the capability, how the assignment was defined, the direction provided or the process connecting the work? The answer helps me decide what to change, rather than simply asking for another round.

Familiar leadership principles, different dynamics

The parallels between leading human teams and my AI team have become increasingly apparent. There are responsibilities to define, strengths to recognize, context to share and standards to establish. Work needs to move between team members without losing important information. And the quality of the outcome still depends on whether I’ve provided a clear vision and made sound decisions along the way.

One of the biggest differences is what happens when the cast needs to change. With human teams, reassignment is rarely just an operational decision. People bring aspirations, pride, relationships and their own perspectives to their work. Moving a responsibility from one person to another can affect confidence, trust and team dynamics. Sometimes the better leadership decision is to coach someone through the challenge, give them room to develop, or accept an approach that differs from what you initially envisioned.

With AI, recasting is easier. If a tool isn’t delivering what I need, I can refine the instructions or move the assignment to another tool without navigating those interpersonal considerations. I can experiment with different combinations, evaluate the results and change course quickly.

That flexibility is valuable, but it doesn’t replace the need for good judgment. I still need to understand what’s going wrong, recognize which capabilities are best delivered by which tool, and decide how to bring it all together.

I’ve had AI tools confidently give me inaccurate information, lose important details when revising documents and produce beautifully articulated responses that didn’t push my thinking forward. When something goes wrong, I still have to distinguish whether it’s a limitation of the technology or a problem with how I’ve organized or briefed the work.

Whether I’m leading people or AI, the objective isn’t to build a team that simply follows instructions. It’s to bring together different capabilities and perspectives, provide clear direction, and create the conditions for the team to do its best work.

The cast is still evolving

I’m not claiming that I’ve figured out the perfect operating model. The biggest friction is still the handoffs. I move between ChatGPT, Gemini and Claude, carrying context, decisions and instructions from one to another. Sometimes the meaning gets lost, documents drift out of sync, or I have to reconstruct where a project left off.

I’ve started addressing that by using Claude as a chief of staff, with a central task board, documented project instructions and a process for capturing decisions. It’s already made a meaningful difference in how I manage multiple ventures, even though the coordination is still more manual than I’d like.

I’ll keep experimenting with the lineup as the tools and my objectives change, adapting responsibilities and finding better ways of working. What’s interesting to me is how familiar that process feels. After years of building and leading human teams, I’m finding that the same instincts for recognizing strengths, matching capabilities to the work, and bringing them together effectively are serving me well with AI.