🎬 Watch the video above for the full teardown—only takes 4 minutes at 2x speed. ⏩
My journey with Bolt started as many no-code tool adventures do: with excitement, frustration, and eventually a credit card. While trying to build a simple gift tracker app, I was amazed at how easy it was and also annoyed at the roadblocks.
This also echoes the sentiment of early adopters to AI tools: OMFG this is amazing when you first start but then you hit the hiccups. “You can’t vibe-code your way to a new payment processor”.
At Irrational Labs, we obsess over what drives human behavior. This post summarizes a few specific design principles that AI products should include to drive user delight.
TL;DR: Get me involved in creation, but make it easy for me to succeed (me = non-engineers).
3 lessons for AI product design
1. Error handling is part of the product
After an hour of failing to get my app working, I discovered the problem was a database connection error buried in a collapsible "show problems" section with confusing labels.
For engineers, this might be obvious. For the rest of us (me!), it's maddening.
Engineers building no-code tools are at risk of expertise bias. They don’t realize that stuff that is easy for them is tough for newbies.
But good news. For Bolt, the solution is behavioral, not technical. And it’s a small one. It comes down to error handling that considers user psychology. Auto-expanding critical errors, using clear language, and offering actionable next steps would have saved my sanity.
With AI tools, we all should expect more zigs and zags from the product user experience. Why? Because products use gen AI, by definition the actual output will always be variable. Error handling has moved from an afterthought done before launching to the actual product spec.
The lesson? Bolt should hire a behavioral or UX designer, even if the app can code itself. The small hiccups matter, but good design and thoughtful error handling solves this.
2. Let users add their own egg
AI should help users create, not create for them. When I asked about UI improvements, Bolt replied "I'll help you enhance the UI/UX" and simply executed changes without involving me in decisions. It added a gradient and rounded corners.
WTF. I didn’t want rounded corners and certainly not a gradient. Good guess, but maybe next time ask me what I want?
The most engaging AI products will find the sweet spot between automation and user involvement. Ask users for input on key decisions. Let them contribute to the creation process in meaningful ways that create ownership.
Counterintuitively, making some parts of your product slightly more difficult can increase user satisfaction and perceived value.
This is the Egg Effect.
Back in the 1950s, cake mix companies faced a surprising problem. Their mixes made baking too easy—just add water, mix, and bake. Sales were dismal until marketers had a counterintuitive insight: make it harder by requiring users to add an egg.
This simple change transformed the experience from "heating up food" to "baking." Users felt ownership over the final product because they contributed.
3. After users are invested, WTP is higher
My final insight is perhaps the most surprising: the frustration itself might be valuable. Especially when most AI features/tools aren’t yet foolproof.
As a non-coder using coding tools, I need to develop the same persistence engineers have cultivated for decades. And this investment pays off.
In fact, Bolt's pricing model brilliantly leverages this psychology. By the time I hit the rate limit, I had invested too much effort to walk away. The $20 monthly fee seemed reasonable compared to losing my work and starting over.
The balance of magic and effort
The best AI tools will balance automation with meaningful user input. They'll handle tedious tasks and speak without jargon, while preserving the parts of creation that generate pride and ownership.
As we design AI products, we should ask: Where's our egg? What small bit of effort can we preserve to maintain the user's sense of accomplishment?
While it’s nice to make users feel good, what you really want is to create products they truly value and will continue using.
🎬 WATCH THE VIDEO for the full teardown journey, including all the UI issues and my growing frustration. At 2x speed, it's only 4 minutes! ⏩
💬 Have you tried no-code AI tools? What's your "egg" — the part of creation you don't want automated away? Share your experiences in the comments.
Have a friend who would enjoy these teardowns? Click the button below to refer them (& earn some great rewards).👇
📧 Questions about product adoption? Shoot me an email: kristen@irrationallabs.com.
Want to increase conversion, retention, engagement? Reach out to Irrational Labs.
We design products that change behavior, using behavioral science. Check out our case studies to see it in action.
Share this post