Clark Kampfe

This is my prediction: non-developer end-users are not going to use AI to generate source code to build their own software products. Not enough to matter economically.

Most people don't do their own plumbing, electrical, work on their own cars, design their own houses, etc. Some do, of course, but not enough to displace plumbers, electricians, mechanics, or architects. The creation of software demands a similar level of systems knowledge in order to reach a level of quality that end-users find acceptable, and end-users aren't about to start designing, generating and deploying software artifacts. Motivated amateurs and entrepreneurs who previously taught themselves to use Ruby on Rails to start their own business will continue to do so, but will do so with the aid of AI. These people are not a large cohort, and I don't see AI meaningfully contributing to the growth of this cohort.

End-users will use AI heavily, though. Chat is not going away. We will see increasingly many products collapse into chat only interfaces, and many users will be quite happy to use many consolidated products through a single chat interface rather than learn a large number of bespoke visual interfaces. Software UI development will wither as a field. One can imagine a great many people not just using but preferring to use their bank's software like, "show me me the current balance of my checking account and the last ten transactions" or "send an email to Jill Smith telling her thank you for the great coffee conversation the other day".

AI will also take the place of end user scripting in professional software like Photoshop or CAD or video editing, which will not go away but will see their UIs decrease in power and complexity. Scripting never had a large user base but was important to power users. Using free-form chat to manipulate APIs that previously required the user to write Python or Javascript or Lua will become more popular.

But non-developer users are not going to generate and manage their own source code and software artifacts. The limiting factor here is not the generation of source code but rather the knowledge required to create whole software systems. Users will do the easiest thing, which will be to solve their tasks directly with existing software and AI products produced by professionals.