What Disney World Taught Me About When I Should (and Shouldn’t) Trust AI
There are just some things AI will never replace. Image of author and the reason Disney is worth the hassle.
Imagine you just moved into a new house. The furniture is placed. The boxes are gone. And as you sit and enjoy your home, you notice the wall looks a little bare. You decide that a beautiful work of art would be the right final touch. Now how are you going to get that perfect piece?
For some reason, this is the image that came to mind when I started learning about ‘Agentic AI’. I think it’s because I’m trying to make all of this AI stuff as tangible and real as possible for my brain to understand. I like analogies and visuals more than code and lectures.
So I took that scenario and looked at what options I’d have in real life. I saw it as a spectrum, ranging from me doing everything myself to simply defining what I wanted and trusting someone else to take it from there.
For instance, let’s say that I wanted a painting of families at the park (I’m probably thinking of this one by Seurat, to be honest). I’d have the image of the painting and I’d do one of the following:
Level 1: Paint it myself. I’m doing all the work, reliant on my abilities alone.
Level 2: Hire someone to paint it, staring over their shoulder the entire time telling them what looks right and what to fix. They’re better than me. But I still have to supervise.
Level 3: I find three master painters with different expertise. One does scenery and background, another paints the people, and the third is an expert in light and shading. They don’t know each other, so I’d haul the canvas from workshop to workshop until it’s done. I’ve gotten away from staring over their shoulders, but I’m still driving all over town.
Level 4: I find three masters that sit in the same workshop. I’d bring the canvas, they’d paint it together, and then I’d pick it up. My responsibilities have dropped to simply showing them exactly what I want and reviewing it at the end to make sure they did it right.
Each of those takes different levels of my effort and involvement. But in every case, I chose the painting to reference and I’m the one who decides if it’s right.
But there’s always a fifth option. It just depends on how much trust I’m willing to put into the artist.
Level 5: I commission an artist to paint a piece about families at the park, but I don’t give them the Seurat to copy. They create their own based on how they interpret my instruction paired with their own style and expertise. They get my request and I trust them to create as they see fit.
What I just described for art is the same concept for the spectrum of AI usage. Most of us are at Level 2, maybe Level 3. Level 4 is technically complex (or so it seems) and Level 5 is either too complicated or we just don’t trust anyone enough, especially not true Agentic AI.
You’ll see a lot about AI Agents, “Vibe Working”, and AI Architecture soon. It will be full of technical jargon and concepts. I don’t think you need to learn the lingo to understand the concept. Just think of art. How do you want the painting created? How much are you willing to do yourself and how much are you willing to let go?
It’s a really important question. But for me to really understand how it worked, I needed to apply it to a real life task that I was truly struggling with. The art example was fun. What I needed was something where I felt real pain.
Which took me to, of all places, the Happiest Place on Earth:
Disney World.
I’m not an artist. My kid is probably a better painter than me, and he’s 5. So while the image of that masterpiece helped me frame the Agentic AI spectrum, it’s not something that would ever really happen to me. But planning a trip to Disney? Now that’s something I can relate to.
Planning a typical family trip can be rough. It’s stressful and expensive and time consuming. Disney World is that on steroids. Now I need to deal with Lightning Lanes and Genie passes and reservation windows that disappear in 12 seconds flat. The information is out there, but it’s in a mountain of blogs and guides and who knows what else. And if you screw it up? You’ve got a kid crying on the teacups because there wasn’t enough time for Tron (IYKYK). It’s a lot.
So to solve this problem, I turned to AI for a solution. Unfortunately, there were a lot of options based on how much I wanted AI to do. So I built this Disney-Inspired AI Involvement Scale to better describe what my role truly was at each level. Not because it’s the best title for an agentic AI framework. But because I could remember it more easily, and honestly the associated quotes for each level just make me laugh.
Level 1: “You want something done, you gotta do it yourself.”
This is where I go deep into the Disney travel universe. I’m reading Facebook posts and insider blogs. I bought “The Official Unofficial Guide To The Magic Kingdom”. I’ve read reviews and whatsits galore.
Anyway, you get the picture. You’re exhausted just thinking about it. I know I am. It’s a lot to manage and coordinate. And it’s important. I want to have a great time. I just don’t want all my energy going to figuring it all out. But if I don’t have a better option, this is the best I can do.
Level 2: “On a scale of one to ten, how would you rate your pain?”
I find a free Disney researcher: AI. I call this version “glorified Google”. I don’t mean that in a negative way. The things I’m looking up and searching for show up a lot faster and a lot better. I can have it sift through all those reviews and blogs, and give me something I can actually digest and decide on. Yeah it makes mistakes and asks me a lot of questions. I have to remind it that while crib availability is nice, my 5 year old doesn’t need one anymore. I have to look up why the hotel is cheaper than normal (AI failed to mention the pool is drained for ongoing maintenance). I still have to find flights that work and read the tutorial on Genie+. But it’s easier and faster than going alone. I’ve got a fighting chance. I’m still doing a lot of work, but the effort to even build a semblance of a plan has reduced quite a bit.
Level 3: “Poof! What do you need?”
Level 2 gives me one assistant that I manage for every topic. Level 3 is where I get smarter about breaking the work apart. I open one AI chat and zero in on hotel options. It researches the pros and cons, and I make sure it knows my priorities. I’m getting better at prompting and being specific, and I don’t have to worry about it getting off track. I open another to get serious about how I can master the lines and wait times and various passes to optimize our time. I also see where I can fit in that cool Star Wars bar so I can sneak off for a much deserved drink. I skip over to one of the big booking sites (Expedia, Kayak, etc.) and research flights, finding the best time for us that won’t break my wallet before the trip even begins.
Each conversation is better than Level 2. Fewer corrections, more focused results. But it’s my job to put it all together. The hotel expert doesn’t know what the flight expert found. The excursion plan doesn’t account for the hotel location. I’m the one carrying the information between them, making sure it all fits. I’m not doing the detailed research anymore, but I’ve become a full-time project manager.
Level 4: “Avengers Assemble!”
This is where I start to really get creative and put AI to work. I take what I learned in Level 3 and apply it in a new way. I use a workflow automation tool (Zapier, Make, etc.) and actually define what I want each AI to do and how they need to “work” together, every time.
In Level 3, I told the “Hotel AI” what I look for in a hotel. It doesn’t really change. So I capture that in a document and I never have to re-explain. I give it my preferred sites to search and specific rules to follow (“lean on reviews in the last 12 months”, “put extra weight on comments regarding cleanliness and noise”, etc.). I do the same thing for “Flight AI” and “Excursion AI”. Things like “I prefer not to fly with my knees in my chest” and “My Kindergartener is NOT 5 feet tall. Recommend rides accordingly”.
Then I connect my agents to get to work. I submit the trip details once and the whole thing runs. Hotel researches, recommends, and notifies me. I approve and Excursion picks it up, already knowing where we’re staying. Then Flights. It takes minutes instead of days. I review, revise, and teach when needed. I’m still involved, but I’m not carrying the canvas between workshops anymore.
Level 5: “Surprise me.”
(You’re welcome for not having you click the much more obvious Frozen link)
This is the level that’s coming with a full head of steam. The level where I’ve read breathless proclamations from people really invested in the tech. This is truly “Agentic AI”, the dream they’ve pitched for so long. And while it’s real, it’s far from perfect. That’s a different post. But the biggest thing I realized as I learned more about it is that it presents a very different, very new choice: Do I trust AI to make the decisions for me? Levels 1 through 4 are some level of me deciding how much work to delegate somewhere else. Level 5 is me delegating choice and decision-making. That’s a whole new ballgame.
This level is the fully autonomous Disney travel agent. I tell it I want to take this trip. It asks me some relevant questions and then plans the whole thing. I don’t give input along the way. I don’t guide its actions. I just get a final itinerary handed to me. I’m out of the decision-making loop throughout the process. I only see the end result.
If I hired a human travel agent and they did all this work for me, there would be a clear cost: money. With AI, the cost isn’t money. It’s either my time or my critical thinking skills. Time is when I have to go back and redo the work. I’ve spent a lot of energy setting up these agents, and now I need to spend more time fixing what they got wrong. Time I probably don’t have. The other cost is harder to see. If I just take the outputs at face value and don’t bother to check them, I’ve outsourced a piece of my brain. I expect the AI to be great. But if I don’t verify, I slowly erode my ability to see things from my own perspective.
Every decision comes with a cost.
When using AI, those costs can grow before I even realize it. At Level 1, I’m spending my time. At Level 4, I’m saving it. That tradeoff makes sense and I can weigh the value. But at Level 5, the thing I’m giving up isn’t time. It’s judgment. And if I keep letting AI make more of my decisions without checking, I save a lot of time. But I lose something I can’t get back.
That’s the real question behind all of this. Not “what can AI do?” We already know the answer is “a lot.” The question is: what do I want to keep doing myself? Not because AI can’t. Because I shouldn’t stop.
