Would you let AI build something your client actually uses?
I gave five AI models the exact same spreadsheet and asked them to build an interactive dashboard.
The results exposed a huge gap between looking impressive and actually being useful.
If you're building AI products or relying on AI-generated software, I think you'll appreciate
Trending on Indie Hackers
I come from a creative/non-programmer background. I wanted to use AI almost as a translator between me and the computer. I made all my own assets, Figma boards, and copy, and let Claude basically assemble everything.
My current dilemma is that I have is that I have a working app (works on Expo) and almost everything is done, yet I don't feel comfortable letting people use it before having the code audited by an actual dev. V1 is local-only, so hopefully that would minimize any security/privacy issues.
These tools are very useful and open the door for a lot of people, but not perfect by any means. I feel more comfortable having a human review my app/work before subjecting anyone to something that might not work once it leaves my phone/computer 🙃
The interesting gap here isn’t “which model built the best dashboard,” it’s the difference between outputs that look correct and outputs that survive real client usage. In spreadsheets especially, usability breaks at the edges—filters, assumptions, and how someone actually interacts with the data under pressure. That’s usually where AI-generated systems quietly fail.
I couldn't agree more.
What I've found is that AI is really good at building something that demos well. The real test is what happens after you hand it to a client.
Can they figure it out without calling you? Do the filters behave the way they expect? What happens when they paste in messy data or ask for one more feature? That's where the difference between an impressive demo and a genuinely useful tool becomes obvious.
That's actually something I spend a lot of time teaching on my YouTube channel. I don't just build dashboards—I explain why they're structured the way they are, how to make them resilient, and how to design spreadsheets that people can actually use day after day.
AI can absolutely accelerate the build. But if you don't understand the fundamentals of good spreadsheet design, you're just generating technical debt a whole lot faster.
Interesting.
Reading your reply made me think less about AI-generated dashboards and more about what happens once usability becomes the thing people are actually paying for instead of generation speed.
I don't think I can explain that line of reasoning properly in a thread without oversimplifying it.
If you're interested, what's the best email to reach you on?