AI Generates the Code, Not the Experience

Enterprise software requires expertise, not just a good AI prompt.

In my previous post, I explored how easy it’s become to build your own tools with AI. While building a personal productivity tool to make your own life easier is a great option that agentic AI offers, building an enterprise app is a completely different story. I met a few people who were vibe coding to build alternatives to popular and commonly used enterprise tools. I am sure these are fun and satisfying to build, but as a sellable product, I’m not so sure.

To have a competitive product in the market and to take on tools that have thousands of man-years of experience, you need experienced architects and domain experts to drive these developments. Else the tool will end up as proof-of-concept build, not production-grade, well-thought-out tool, like the one I witnessed. .

It makes sense to build a tool with a focused use case, like a translation engine or a tool to manage expenses across different credit cards and bank accounts, using vibe coding. But trying to build an ERP, CRM, or ticket management system to beat an incumbent provider using nothing but Claude or Codex is a little too ambitious. It might be possible if you are an expert in that domain.

Unfortunately, many creators of these types of tools do not realize the subtle difference between building a product with a vision and generated code that feels like a product. I suppose the newfound freedom of agentic development is driving many of these efforts, and they will eventually slow down or pivot once the creators realize there isn’t a paying market for these tools.

It brings me back to the question of the best use of AI. Right now, it is being used for a variety of purposes that we can live without. Creating tools and capabilities that already exist, creating images and videos that are funny, and so on. Driving the planet to create massive, power-hungry data centers to support individual flights of fancy seems like a waste of technology. Some solopreneurs will definitely succeed with innovative and disruptive ideas, and many innovations that could not be brought to fruition in the past because of the sheer cost of resources will now have a chance to succeed. A majority of individual initiatives will die down. I am guilty of using AI for trivial jobs as well. Over the last few months, I have built tools to manage my own finances, do any-to-any language translations, a video generator and so on. Most of these are indeed proof of concepts and work very well in a non-GPU, air-gapped environment for my personal use. They also helped me explore and better understand how to use AI. I am now building an educational tool to help underprivileged kids who cannot easily access educational resources.

The real power of AI in everyday applications lies in making connections that were hitherto impossible or difficult. Is there a connection between philosophy and math? Can we use art to better understand genetics? The real ROI for AI investments is in solving real-world problems like climate change, disease, and food scarcity.