Stop Prompting Your AI, Start Project Managing It
The biggest mistake people make with AI isn't misunderstanding the technology. It's misunderstanding their role in relation to it.
For many people, there is still something almost mystical about AI.
Like an oracle sitting atop a mountain, we can ask our favorite AI almost anything and receive a meaningful response within seconds. And not unlike that mountain-sitting oracle, some of those responses might have been hallucinated thanks to the availability of local poppy or mycelium... or maybe just because it sounded plausible and fit the vibe.
Nobody fully understands what's happening inside these models, but that doesn't make them mystical. Even as they can occasionally feel like real-life magic, AI is ultimately responding to the information we provide it and drawing upon patterns it has learned from the information it was trained on. The more I work with AI, the more convinced I become that the biggest mistake people make is not misunderstanding the technology itself, but misunderstanding their role in relation to it.
Daily, I get questions asking, "Can my AI do this?" Most of the time, my honest answer is that they could have received a much better answer by simply asking the AI directly. At the same time, when people do engage with AI, they often approach it as though it should already understand exactly what they need. They provide broad prompts, vague instructions, and leave enormous room for interpretation. Then they are disappointed when the results miss the mark.
The deeper I get into the world of AI, the more I realize that my years as a Project and Program Manager at Adobe prepared me for this moment.
As a PM, I learned very early that poorly defined projects rarely succeed.
If I don't know what we're delivering, my team can't know what they're building.
If success hasn't been clearly defined, nobody knows whether we've achieved it. Ambiguity compounds. Assumptions multiply, teams spin their wheels, customers get frustrated. The project drifts without ever knowing if the finish line was crossed because nobody bothered to draw it.
AI behaves in remarkably similar ways.
That realization crystallized for me during a customer call not long ago.
I joined expecting to discuss a Salesforce Marketing Cloud integration. The customer expected a demo. Those are not the same thing.
After a brief half-second of panic, my PM instincts kicked in. If I had joined that call with a Business Consultant and Technical Consultant sitting beside me, I know exactly what I would have done. I'd have started furiously Slacking them.
"What's a good page we can use as the basis for a demo?"
"Can we quickly generate a few emails from that page?"
"What can we build that demonstrates the value of this integration?"
Then I remembered that the "P" in FDPM stands for Product now, not Project or Program. But I wasn't quite ready to leave those other muscles behind. Time to activate that Voltron-like Power Delivery Machine spanning all those Ps living in my experience.
So while introductions were happening, I quietly got to work with my new team.
First, I confirmed that Gradial was connected to SFMC in the environment we were using. Then I asked a simple question:
"Can you find me a page that would make a good starting point for this demo?"
Moments later, I had a solid suggestion.
Foundation established.
As the conversation evolved, the customer clarified that their primary use case wasn't really the integration itself. They wanted to understand how content living in AEM could be transformed into email campaigns in SFMC.
No problem.
"Can you find me a page in AEM that would work well for this use case?"
Done.
"Can you help me turn that into a campaign with multiple email touch points?"
Done.
Within minutes, we had transformed a real landing page into a three-email journey that demonstrated the value of the workflow they cared about. The customer got the demo they wanted. Nobody got frustrated by a last-minute pivot. Nobody had to scramble after the call to clean up the mess.
More importantly, something clicked for me.
Thinking about AI as a member of your team is far more valuable than thinking about it as a glorified search engine.
As a PM, I never knew everything my team was capable of.
So I asked.
I didn't personally build every technical solution.
So I asked.
I worked with customers to understand what they were trying to achieve. I worked with Business Consultants to connect that work to business value. I worked with Technical Consultants to understand constraints, possibilities, and implementation details. My job was rarely to have all the answers. My job was to create enough clarity that the right people could help find them.
That process feels remarkably similar to working effectively with AI.
The people who seem to get the most value from AI are rarely the people with the most technical knowledge. They aren't necessarily the best prompt engineers, either. More often, they're people who know how to clarify goals, expose assumptions, break down problems, and ask better questions.
In other words, they're acting like project managers.
Great PMs don't succeed because they know everything. They succeed because they've learned how to navigate uncertainty. They understand that most problems become easier once they're defined clearly enough. They know that a team's effectiveness is often constrained less by talent than by a lack of clarity around what success actually looks like.
AI rewards those same behaviors.
The difference between a good PM and a bad PM isn't intelligence. It's a willingness to ask questions and an ability to drive toward clarity.
I suspect the difference between individuals and teams that succeed with AI and those that struggle will be much the same.
Ask questions.
Provide context.
Define success.
Your AI doesn't need you to be more technical. It doesn't even need you to be smarter.
It simply needs you to do what great leaders, project managers, consultants, and operators have always done: create clarity, ask better questions, and help your team succeed.
The fact that your newest team member happens to be artificial is mostly beside the point.
Ryan Yepsen
Ryan writes about enterprise AI, customer reality, systems thinking, and the operating conditions that make meaningful execution either possible or painful.