(Credit: Zain bin Awais/PCMag/Getty Images)
AI can already do a lot for you. AI agents can handle tasks, AI chatbots can hold conversations, and AI search engines and web browsers can help you navigate the web faster than ever. But AI can also build things for you. With the right guidance, these tools can help create real apps—from a better Zillow-style browsing tool to something niche like a Warframe build planner. The catch is that working with AI this way requires a different approach than simply asking a question or generating a quick answer. After spending time building apps with AI myself, I've picked up a number of lessons about how to make the process smoother, avoid common mistakes, and get better results. These are the seven that made the biggest difference.
1. Don't Bite Off More Than You Can Chew
AI is powerful, but it’s not magic. You aren’t going to build an entire app or even add major features in a single step. For example, if you want to make an app to help you study Spanish, don’t say that. Instead, ask AI to make an interface for digital flashcards. Then, ask it to add flashcards section by section for specific cultural knowledge, grammar concepts, vocabulary topics, and whatever else. Doing too much at once can lead to problems such as bugs, crashes, and incomplete features.
(Credit: Anthropic/PCMag)2. Pay Close Attention to What the AI Does
When you set AI loose to add a feature or make a change to your app, keep an eye on the process. These kinds of requests often take a while and result in big changes to the underlying code. If you notice something off early on, it’s better to stop the AI and make adjustments. Otherwise, you risk having issues down the line and simply wasting your time.
(Credit: Anthropic/PCMag)For example, during the process of creating a Warframe builder planner app, I asked Claude to pull stats from the official Warframe wiki and use that as the basis for entries in a database I wanted to create. However, I noticed that Claude couldn’t actually access the wiki and decided to search for an alternative. Rather than let Claude make that decision, I stopped the process and picked a different source.
3. Check, Double-Check, and Check Again
Every time an AI completes a request, whether it’s adding a major feature or a small interface refinement, make sure to confirm that the app works correctly. Making a minor change in one area might end up breaking a big feature in another, especially when an AI is in charge.
I experienced this first-hand when I asked Claude to make some improvements to the mobile interface of the Warframe app I was building. Upon checking the mobile and desktop interface afterward, I discovered that Claude erroneously applied some changes to both.
4. Make Backups Early and Often
Sometimes things just go wrong, even if you’re careful. Say, for example, that an AI accidentally inserts data from the wrong third-party source into your app. Fixing such problems and other similar ones is a pain, so you always want to have recent backups of your app to restore.
Backups are important in pretty much every aspect of digital life, but they're even more essential when you work with AI because you aren't the one primarily in control.
5. Be Wary of Usage Caps
Making an app requires you to spend a lot of time with reasoning models, such as Claude’s Opus or ChatGPT’s Thinking models, so you might need to sign up for a paid plan. The latter is one example of a model that isn't available for free. Even if you pay for the premium version of a service, you might still have to be careful with how many requests you make during the app creation process, lest you run into a usage cap before you finish what you’re doing.
6. Remember the Limitations of Making Apps With AI
You can make almost any type of app with AI, so long as the complexity and scope are reasonable. It's not a problem to make a simple game, for example, but you absolutely can't prompt your way to generating the next Skyrim. Even the most advanced AI models aren’t a replacement for a large team of artists, musicians, producers, programmers, writers, and more. As a rule of thumb, focus on creating apps that do a few concrete and specific things, especially if you don’t have much (or any) coding experience.
(Credit: Anthropic/PCMag)7. Don't Make Apps That Require Meaningful Security
You should avoid building apps that handle sensitive data, such as password managers. When AI makes an app, it doesn’t automatically make it secure. Even if you ask an AI to tighten up the security of your app, that’s not a guarantee it will actually comply. For now, you should always ask an experienced human programmer to review your code and properly secure it.
Bonus Tip: Go Slowly and Iterate Endlessly
When you attempt to make an app with AI, I advise you to work slowly toward your goal, implementing features and refinements one at a time, testing them, and moving on to the next thing until you’re happy with the final product. Just like with actual software development, don’t rush the process. Doing so will likely just lead to headaches later on.