AI Can Write Code, But It Will Not Replace The Enterprise Tech Stack

The rise of generative AI and agentic AI, with AI coding tools capabilities, has created a powerful bear case against traditional software, SaaS, and technology services companies. The argument is seemingly straightforward: if AI can generate software quickly and cheaply, enterprises should be able to replace much of their existing technology stack, thereby reducing what they spend on software vendors, systems integrators, outsourcing providers, and managed services firms.

I believe that this thesis is substantially overstated. AI will change the economics of technology, but wholesale replacement of the enterprise tech stack is likely to take much longer than many investors expect. There are several reasons why I have come to this conclusion.

Reason 1: AI Can Rewrite Code, But It is Not the Same as Replacing a Tech Stack

There is no question that AI can generate code quickly and effectively, and increasingly so. We already see examples where AI tools are being used to refactor applications or port code from unsupported environments into newer technology environments. Those cases can work well. The cost of generating the new code can also be quite reasonable. However, that does not mean enterprises will suddenly start rewriting their entire technology estates.

The reason for this is that the difficult part is not generating the code. The difficult part is understanding everything the existing application does, every system it touches, every dependency it contains, and every business process that relies on it. Even if code generation became effectively free, testing and integration would not be free.

For a large enterprise, the business risk of missing one critical dependency can be substantial. The cost of proving that a rewritten system behaves correctly across thousands of processes and interfaces can also be substantial, which changes the economics of replacement considerably.