Part 1 — The AI's Blank Box Trap
"Multi-agent" is eating the engineering world. Teams shipping product in 2026 often have:
- an AI agent for copywriting
- SEO
- security
- GTM
- code review
A whole staff of AI specialists.
But consumers opening ChatGPT / Gemini still see… a blank box.
For consumers, the breakthrough isn't multi-agent coordination
Consumers won't care about "multi-agent architecture." They just want:
- dinner solved
- admin cleared
- money managed
- time back
They need pre-made specialists (Cook, Admin, Money, Fitness…) that:
1. don't start cold every time, and
2. make them think: "Oh — I can use AI for that."
That's the "rich-person effect." Wealthy people don't have one assistant. They have a team: EA, accountant, trainer, travel planner, stylist, researcher, on call, with full context and endless opportunities.
AI can democratize that — but a blank box won't.
The "Search Bar" legacy trap
The search bar is powerful when users already know what they're looking for. But AI isn't a search engine.
Search: fetch information vs. Service: do work
In a hotel, nobody wants a blank room and "How can I help?" They want signage: Spa, Gym, Bar, Concierge.
Copying the search bar removed the signage.
The imagination tax
AI companies are outsourcing discovery to users:
> "We built a god. You figure out what to do with it."
So users must contextualise the use case, phrase it right, then iterate. "Prompt engineering" is often just guessing the magic words.
That's why adoption can look shallow: trivia, drafts, images — while the real value is delegation.
Consumers shouldn't have to imagine that AI can deeply help with taxes, refunds, meal planning, travel, negotiation. The product should make the next best action obvious. And GPTs and Gems should only be a transition, not the end play.
The great disintermediation
As AI becomes a service engine, it becomes the front door to work.
Users stop thinking "which app?" and start thinking "which outcome?"
Apps don't die — they get pushed down into infrastructure: capabilities, integrations, endpoints. The AI becomes the interface, router, and broker.
This isn't just UX. It's distribution and business model warfare.
In Part 2, we'll look at a possible fix: contextual signage (pre-made specialists + reviewable artifacts) — and what it means for apps when AI becomes the front door (capabilities, embedded modules, or marketplace inventory).
← Back to all posts