Vibe Coding Is Impressive. But I Think We're Asking the Wrong Question.

Over the last few months, I've noticed a change in almost every customer conversation.

A year ago, AI discussions were mostly about copilots, agents and chatbots. Today they're about Lovable, Replit, Claude Code, Codex, Antigravity and the growing number of AI coding platforms.

The excitement is understandable. These tools are genuinely changing how software is built.

In fact, during a recent meeting, a customer said something that would have sounded unrealistic just a few years ago.

"Maybe we shouldn't buy another CRM. Maybe we should just build one ourselves."

What surprised me wasn't the comment. It was that nobody around the table thought it sounded crazy. And honestly, I didn't either.

Building software is becoming dramatically easier. Screens, forms, APIs, databases and business objects that once required weeks of work can now be generated in hours. Every month these tools become more capable, and I have no doubt they'll continue to improve.

I don't see that as a threat. I think it's one of the most important shifts our industry has seen in decades. But after thinking about it for a while, I realized that the conversation often stops one step too early.

Building an Application Is Different from Building Enterprise Software

We're asking whether AI can build an application. I'm more interested in whether AI can build something an entire business can rely on. Those sound similar, but they are very different problems.

Imagine that two companies need exactly the same supplier approval process.

The first team opens an AI coding platform and starts with a blank page.

The second team opens an enterprise platform where users, permissions, security, connected data, integrations, audit history and business rules already exist.

Both teams describe the same business process and both use AI. Yet they are solving completely different problems. The first AI is generating an application. The second AI is extending an existing business. That distinction feels increasingly important.

Enterprise AI Needs Business Context, Not Just Code

Most organizations aren't starting from zero. They already have employees, customers, products, permissions, approval chains, reporting structures and years of operational knowledge. They don't need AI to recreate all of that. They need AI to understand it.

The more I think about it, the more it reminds me of professional kitchens.

Modern chefs use incredible tools. They automate repetitive work, prepare ingredients faster and achieve a level of consistency that wasn't possible years ago. Nobody concludes that the value of a restaurant comes from the equipment. The value still comes from understanding food.

I believe software is moving in the same direction. Writing code is becoming easier. Understanding businesses is becoming more valuable. That's why I think the next chapter of AI won't be defined by who generates applications the fastest.

It will be defined by who understands organizations well enough to generate the right business capabilities on top of everything that already exists. Perhaps that's the real opportunity.

Not asking AI to start from scratch every time, but giving it a foundation where security, governance, permissions, connected data, workflows and business rules already exist, so it can focus on what actually changes: the business process itself.

The Future of Enterprise Software Starts with Understanding the Business

I don't think the biggest impact of AI will come from helping us build software faster. I think it will come from changing where software begins. For years, software started with requirements, screens and code.

We're getting closer to a world where it starts with understanding how a business actually works. If that happens, vibe coding won't replace enterprise software. It will change how enterprise software is created.