The fastest way I've found to get non-technical people comfortable with AI is to take the blank canvas away from them.
Our company aligned on ChatGPT. Early on, I realized that the blank chat window was the biggest barrier to adoption. People would open it, stare at the cursor, and think, "What do I even ask?" So I started building custom GPTs. Not because I wanted to show off what was possible. Because I wanted to create experiences where someone could type a plain, simple question and get something useful back without needing to be good at prompting.
That's the key insight. A well-built GPT already has the context it needs for a specific task. It knows the domain. It has instructions. It has shared knowledge baked in. So the person using it doesn't have to front-load every prompt with background and constraints. They just ask the question. The GPT handles the rest.
I built a Copywriter GPT that knows our brand voice and content guidelines. A User Research Advisor that can help frame interview questions and synthesize findings. A Career Coach for my team's development conversations. Each one is scoped to a specific part of our workflow. Each one is designed so that even someone who has never touched ChatGPT can get value from it on the first try.
That last part matters more than the technology. As I learn and experiment with these tools, I stay connected to what it felt like to be new to all of this. It's easy to forget how disorienting AI is when you've been deep in it for months. Every time I build a GPT for my team, I test it by imagining the most skeptical person on the roster using it. If they'd roll their eyes at the instructions, I rewrite them. If the output requires too much follow-up prompting, I restructure the knowledge base.
The best part is what happened next. A couple people on my team started building their own. They saw the pattern. They understood the concept. And they had ideas I never would have had because they know their own workflows better than I do. That's the whole point. You don't coach people on AI by teaching them everything you know. You build something useful, put it in their hands, and let them take it from there.
The guardrails aren't limitations. They're the thing that makes experimentation feel safe enough to actually try.