Hey everyone,
Why does everything have to become a conversation?
This is part 1 of The Interface Is the Instrument, a three-part series. Over the next three Tuesdays I want to look at a quiet assumption moving through the software industry: that the future of sophisticated applications is a chat window.
There's a peculiar trend taking hold in the technology industry. It seems that every software company has decided that the future of its product is a chat window.
Enterprise software companies are racing to make their applications accessible through conversational AI. Salesforce and Anthropic have expanded their partnership to bring Salesforce capabilities into Claude—through an announced plugin, Salesforce in Claude, that lets sellers work with live Salesforce context and take governed sales actions from a conversation, not through a wholesale replacement of the Salesforce interface. Music technology companies like Suno and Soundverse are offering MCP integrations that let people create music through conversations with AI assistants; Suno's MCP, announced in late October 2026, is currently in beta. Project management, design, analytics, productivity, and just about every other category of software seem to be heading in the same direction.
In announcing Claudeforce on August 26, 2026—the latest chapter of the Salesforce and Anthropic partnership—chair and CEO Marc Benioff put that assumption unusually clearly. “For decades,” he said, “enterprise software required users to manually navigate static UI to get work done.” In the same announcement he added, “Here, the UI is the AI.” That is the stated direction, not my inference about anyone’s motives. Salesforce and Anthropic Announce Claudeforce https://www.salesforce.com/news/press-releases/2026/08/26/salesforce-and-anthropic-announce-claudeforce/
The proposition is straightforward: instead of learning an application, navigating its interface, and figuring out how to accomplish something, just tell an AI what you want.
It sounds convenient. It sounds accessible. It sounds like progress.
But I have a question.
Why the fuck would we want to do everything in a chat window?
I don't mean that conversational interfaces are inherently bad. They're useful, sometimes extraordinarily so. I use them. I build with them. I believe deeply in the potential of AI to change how we interact with technology.
What I object to is the increasingly pervasive assumption that conversation is the destination.
It's as though we've collectively decided that the ultimate expression of decades of technological progress is to take everything we've learned about designing software interfaces and replace it with a text box.
I think that's a mistake.
And I think we're in danger of throwing away something extraordinarily valuable.
The interface isn't the inconvenience
Consider a digital audio workstation.
I've spent a considerable amount of time working in Ableton Live. Live first shipped in 2001, and its interface is the product of more than two decades of thinking about how musicians interact with sound—along with a much older tradition of consoles, meters, and direct manipulation.
The mixer isn't simply a collection of controls. It's a visual representation of a complex system.
Faders communicate relative levels. Meters communicate signal behavior. Sends and returns establish relationships between tracks. Groups create hierarchies. The arrangement view communicates time, structure, and relationships between musical ideas. Instruments expose parameters through knobs, envelopes, modulation controls, and visual feedback.
These aren't arbitrary design decisions. They're carefully developed ways of making complicated things understandable and manipulable.
More importantly, they allow a musician to interact with many dimensions of a project simultaneously.
I can look at a mixer and immediately understand a surprising amount about what's happening. I can see that a track is too loud, notice that a send is unusually high, recognize that something is clipping, and make several adjustments without interrupting the larger activity.
The interface is participating in my thinking.
Now imagine replacing that experience with a conversational interface.
"Can you show me the levels?"
"Turn down the bass."
"Increase the reverb on the second guitar."
"What's clipping?"
"Can you compare the levels of the vocal tracks?"
Sure, an AI agent might be able to accomplish all of those things. It might even be able to do them accurately and efficiently.
But why would I want to conduct an entire creative session through a series of instructions and responses when I already have an environment designed specifically for the work?
A fader isn't a primitive way of expressing a request to change volume. It's a purpose-built instrument for controlling volume.
The interface isn't an obstacle standing between me and the functionality.
The interface is part of the functionality.
We've spent decades learning how to make complexity understandable
There's an enormous amount of intellectual and creative work embedded in the interfaces we use every day.
Graphic designers, interaction designers, industrial designers, software engineers, and countless others have spent decades developing visual languages for communicating complex systems.
We've developed conventions, metaphors, hierarchies, spatial relationships, gestures, affordances, and feedback mechanisms.
We've learned how to expose complexity without overwhelming people. We've learned how to make relationships visible, how to communicate state, and how to give people direct control over the things they're manipulating.
A good interface doesn't simply expose functionality.
It makes functionality understandable.
Think about a spreadsheet. A spreadsheet isn't merely a place to store numbers. It's a visual environment for exploring relationships, testing assumptions, discovering patterns, and understanding how changes propagate through a system.
You can look at hundreds of values simultaneously. You can follow a formula. You can change an assumption and immediately see the consequences. You can discover something you weren't even looking for.
The visual representation is part of the thinking process.
The same is true of a design application, a music production environment, a CAD system, or a sophisticated CRM.
These interfaces aren't just collections of buttons that we haven't figured out how to replace yet. They're carefully constructed representations of the domains in which people work.
And I find it remarkable that we're suddenly treating all of that accumulated knowledge as an inconvenience.
Creativity isn't a sequence of instructions
There's another reason this matters, particularly in creative work.
Sometimes you don't know what you want until you see what you don't want.
Sometimes you discover an idea by moving a control, changing a parameter, rearranging something, or experimenting with an unexpected combination.
You might be adjusting an equalizer and discover something interesting in the relationship between two frequencies. You might be moving clips around in an arrangement and stumble across a musical transition you hadn't considered.
You might open a design file intending to change one thing and discover an entirely different direction because something catches your eye.
These aren't necessarily activities that begin with a clearly articulated intention.
They're exploratory.
They're often nonlinear.
And they're frequently dependent on the ability to interact directly with a system and observe what happens.
Conversational interfaces tend to encourage a different pattern: articulate an intention, receive a response, evaluate the response, and articulate another intention.
That's useful for many things. But it's not a universal model for how people think or work.
In fact, I worry that an overreliance on conversational interaction could subtly change the nature of the work itself.
If every action begins with a request, we risk designing environments that encourage people to become increasingly proficient at describing what they want rather than exploring what might be possible.
Those are not the same skill.
This isn't an argument against AI
I want to be clear about something.
I don't think AI is the problem here. Quite the opposite.
I think we're making a category error by confusing the capabilities of AI with the limitations of the interface through which we've become accustomed to accessing it.
AI can understand context, interpret visual information, reason about complex relationships, manipulate objects, and coordinate sophisticated workflows.
Those capabilities should give us an extraordinary opportunity to rethink how we interact with software.
Instead, we're frequently taking those capabilities and putting them behind the same conversational interface we've been using since the first widely accessible AI assistants appeared.
It's a little like inventing a new kind of engine and deciding that the most innovative thing we can do with it is put it in the same vehicle we've always driven.
I'm not suggesting that existing interfaces should never change. They absolutely should. AI will make entirely new kinds of interfaces possible, and many of our existing conventions will eventually become obsolete.
But we shouldn't confuse evolution with elimination.
We should be asking what becomes possible when intelligent systems are integrated into the environments in which people already think, create, explore, and work.
We should be asking what new visual metaphors, controls, interactions, and experiences we can invent.
We should be asking how to make interfaces more expressive, not how to make them disappear.
Because the interface is where the capabilities of a system meet the intentions of the person using it.
And that relationship deserves more imagination than a chat window.
Next Tuesday I'll publish part 2, The Imagination Constraint.
Talk soon,
Todd