The Future of Software: AI Generating Interfaces On Demand
· LinkedIn post
A LinkedIn post recently made a bold claim: "Traditional applications like PowerPoint, Photoshop, Word, and Acrobat will be obsolete.…
A LinkedIn post recently made a bold claim: "Traditional applications like PowerPoint, Photoshop, Word, and Acrobat will be obsolete. AI will generate UI on demand."
I think that's roughly right. But the post glosses over the messy middle between now and that future. I build AI agents and workflow automations daily for small and mid-sized businesses. So I see what works and what doesn't.
It's Already Happening in Pieces
You can already use GPT-4 to write a VBA script inside Excel. You can tell ChatGPT to generate a simple HTML page with a button that does a calculation. Tools like v0 by Vercel let you describe a UI and get React code. But that's still code. The real shift is when the AI itself serves the interface—no code export, no separate app install. Just you saying what you need and the system showing it.
That's where we're heading. Instead of buying a license for Word, you'll describe a document layout and the AI will render it. Need a quick image edit? No Photoshop. You'll just describe the change and see the result. This is not years away. I've built AI agents and automation workflows that already generate reports, transcribe calls, and update CRMs without a single manual click. The UI is the conversation, or a temporary dashboard the AI spins up.
The Limits No One Talks About
But here's the honest part. This future has two big problems: reliability and context. When I build a chatbot for a client, we have to spend hours on guardrails. The AI hallucinates. It forgets rules. It builds a beautiful UI that does the wrong thing. For critical business tasks, you need deterministic logic. That's why I still rely on n8n workflow automations for payment processing or data sync—no AI, just rules. The on-demand UI idea works for simple tasks. But for complex workflows with compliance needs, you want a tested path.
Cost is another factor. Generating a custom interface on every request chews through tokens. Until inference costs drop further, it's cheaper to use a pre-built app for frequent tasks. So I don't see Excel dying tomorrow. But for one-off analyses, AI-generated tools will eat into app usage.
What Small Businesses Should Do Now
Don't wait for the app-less future. Start using AI tools that augment existing apps. I've helped clients where the AI pulls data from QuickBooks, generates a chart, and emails it—no manual work. Those are real use cases that deliver ROI today. The pure on-demand interface will come later. For now, the practical move is to automate repetitive tasks with AI agents that work inside your current stack.
My take: in five years, the app store on your phone will look different. You won't download an app for every need; you'll rely on a general AI assistant that builds micro-tools on the fly. But that future will be built by those of us testing the limits today. It won't be smooth, but it's coming.