Think about what happens when you hand someone Excel.
In the hands of someone new to it, Excel adds and subtracts. Give it to someone with a bit of technical comfort, and they’ll build pivot tables that summarize a semester of data in seconds. Give it to a true power user, and they’ll build a fully automated back end that runs an entire operation.
Same tool. Very different outcomes. The difference is the skill of the person using it.
AI works the same way, with one important twist: the output looks good no matter who’s driving.
When it looks finished and isn’t
Recently, someone I know shared the fourth draft of a web app they’d built with AI. At first glance, it looked finished. Every button was there. Every text field was in place. It had the polish of a real product.
Then I started using it. The flow felt disjointed. The questions didn’t line up with each other. Moving forward through the app didn’t translate into the progress I expected as a user. The pieces were all present, and the experience didn’t hold together.
This was their first time designing an app, and that’s exactly the point. In the past, building even a rough prototype forced you through a discovery process first. You had to understand the goal, think through the experience from the user’s perspective and map out the flow before you could write a line of working code. By the time you had something clickable, the thinking behind it was already done.
AI has removed those barriers. Today, anyone can produce something that looks right in an afternoon, and that’s genuinely exciting. It also means it’s easier than ever to build something that looks right and doesn’t pass the smell test.
The rocket problem
I hear a version of this question often: “Can’t we just give everyone AI and let it help them figure things out?”
Handing someone AI with no training is like handing a kid a rocket and saying, “Okay, go to the moon!” The rocket is incredibly capable. The kid has no idea what a rocket is, how to steer it or where the moon actually is.
AI is a tool, and like any powerful tool, its value depends on the person operating it. AI makes polished-looking output available to everyone, and the quality underneath still depends entirely on the skill and judgment of the person directing it.
Ideas are cheap. Execution sets teams apart.
With AI, a new idea can become a draft, a prototype or a plan in minutes. When ideas are that cheap to produce, execution is what sets teams apart. And you can only execute well when everyone is moving to the beat of the same drum: solving the right problems, using shared standards, and knowing what good looks like.
That takes investment in people. Here’s where I’d start:
- Train before you roll out. Give your team a shared method for working with AI before you hand out licenses. I use a four-step approach — Persona, Context, Detailed Prompt and Validation — and I walk through it in Stop Boiling the Ocean.
- Keep the discovery step. Before anyone builds, ask: What’s the goal? Who is this for? What should their experience feel like from start to finish? Those answers come from people who understand the work, and they shape everything AI builds next.
- Define what good looks like. Agree on standards up front, so people can tell polished from sound.
- Validate with real users. Put early versions in front of the people who will use them. Some of my most useful lessons came from exactly that, which I wrote about in What Building a Free Interview Tool Taught Me.
Back to the fundamentals
There’s no easy button here. The teams that get real value from AI will be the ones that pay attention to the fundamentals: clear goals, thoughtful design, shared standards and people who are trained to use the tool well. That has always been true of good operations. With AI, it matters more than ever.
Helping teams build those fundamentals is the heart of my AI Enablement Workshops. Every participant leaves with a shared method, clear guardrails and practice on their own work. If your team is ready to move from experimenting with AI to executing with it, let’s talk.
