Don’t Make Me Talk to My Light Switch

Tech futurists say chats and agents will replace graphical interfaces. I disagree. It’s like saying:

“Because a hotel concierge is so helpful they should replace vending machines.”

A concierge is universal; they can call you a taxi, book a tour, and change the room if it has bed bugs.

But if I want a Snickers? I don't want to greet them, explain the request, and find out that my idea is brilliant. I want to press the bloody button and eat my Snickers. I want zero cognitive load.

Same with interfaces. There are two core metrics of interface quality:

Effectiveness: can the user accomplish the task?
Efficiency: how quickly and effortlessly can they do it?

Conversational UIs are effective. You can initiate almost any task through a universal interface: pull any data from the database, present it in any way you want, operate with it in any way you need.

But conversational UIs fail at efficiency, speed, and reducing cognitive load.

They rely on recall (you have to remember what is possible and how to ask for it). They suck at recognition (seeing the option in front of you).

Graphical interfaces (GUIs) are efficient. They are optimized for speed, muscle memory, and predictable, structured output.

GUIs are also way better at providing status. I don’t want to ask: "What’s the current PnL of my Apple position?" I don’t need reasoning and interpretation.I don't want to wait for a free-form text summary. I want to see a simple optimized UI that I can process in milliseconds. And if the position goes from profit to loss, I can spot the color change with peripheral vision.

The future isn’t chat/voice UIs replacing graphical UIs. They should coexist, each used where it’s strongest.

Don’t make me talk to my light switch.