In the same way that home DIY is different from hiring a professional tradesperson, D-AI-Y is different from professional AI development.
Both have value, and both have their place.
D-AI-Y is great for vibe-coding a disposable app, a personal solution, or a bespoke tool that doesn't require scalability or follow-the-sun customer support or maintenance or 24/7 reliability...
In the same way that some DIY work is professional-grade-or-better (looking at you, Dad!), a good AI-collaborator can do professional AI development from a home office (or coffee shop)...but they can't (yet) support that work at enterprise scale and quality.
Enterprises still need reliable, accountable vendors with solid, tested, supported, maintained products, and partners for implementation and training.
Can all of those functions be done by AI? Arguably yes.
When will we see that? I think we're several years away from practical solutions.
In the meantime, current enterprise leaders are not going to throw out deeply-embedded software to re-create it internally. Aside from needing to rebuild all of the roles provided by traditional vendors, installers, and maintainers (each of which comes with a pricetag and which is NOT part of the core value proposition of the company), no CIO would risk a project that important without devoting some of their BEST internal folks...thereby taking some of the highest contributors away from serving the core value prop of the company.
Finally, the ecosystem around large systems provides a network effect that could not be created by building internally. This cuts both ways -- you wouldn't have bloated features built for other customers...but you also wouldn't have new features built for other customers that you didn't know you needed.
An advantage to having vendors that serve other customers is they solve problems you haven't yet encountered, they build integrations you haven't yet needed, they handle all of the technology updates before your upgrades...all without needing internal resources.
So, will vibe-coding kill software? I say yes for some personal and productivity apps that benefit from bespoke customization and an understanding of your life, no for anything regulated (healthcare, finance, pharma) or ultrasecure (defense, government) or mission-critical that requires deep ongoing support. These latter categories will benefit increasingly from AI, and longer-term will succumb to professional AI development (for the stability, decision-making, efficiency, and reliability gains that will appear with better models), but we're talking far-distant AI future (5+ years).
I work with a lot of companies, and have yet to see a single successful initiative to replace a commercial tool with a vibe-coded one. I HAVE seen a few instances where internal tooling is done with agentic engineering, harnesses, and traditional software engineering practices adapted to the speed and limitations of Generative AI tools...but I would not consider that vibe coding. Agree? Disagree? Other thoughts to add?