AI as a Universal Interface: How Agents Translate English Into Software

Key Insights

• Every software has its own "grammar"—Figma uses frames and components, Salesforce uses accounts and contacts, CAD uses splines and constraints
• AI eliminates the learning curve by making English the universal interface, letting users describe outcomes while agents handle the software's underlying language
• A non-technical founder used AI to design a previously unmanufacturable dress by directing Codex to operate CAD software on her behalf
• Product design is inverting: instead of simplifying every screen, teams now build "harnesses" that let AI agents safely navigate complex systems
• Expertise doesn't disappear—it shifts from syntax and menus to systems architecture and understanding how pieces fit together

The Problem: Software's Hidden Grammar

Every powerful software demands fluency in its own language. Figma speaks in layers and components. Salesforce talks in accounts and opportunities. CAD operates through splines and constraints.

The relationship is unforgiving: the more powerful the tool, the more expertise users must acquire. A usability lab can close this gap one user at a time, but it's slow. Product managers intend simplicity, yet capability and required expertise always rise together.

This onboarding burden hits product-led growth companies hardest—businesses built to sell themselves without a sales team. Customers must climb a learning curve before they see value.

AI as Translation Layer: English Becomes the API

AI solves this differently. It makes English the universal interface.

Instead of learning CAD's grammar, a user describes her desired outcome in plain language. An AI agent translates that intent into the software's native syntax—splines, constraints, coordinates—and executes it. English becomes a babelfish between human intention and application logic.

Programmers experienced this first: AI already translates English into C++, Rust, and TypeScript. Now the same principle reaches the application layer. An agent can hold, retrieve, and reason over more of a system's schema at once than a person clicking through menus, making it possible for non-experts to operate tools that were previously gatekept by expertise.

The Dress That Couldn't Be Made—Until Now

Yana Welinder, inspired by Ruth Asawa's looped-wire sculptures, wanted to create a garment with sculptural, interlocking forms. The design was "previously unmanufacturable."

She couldn't operate CAD software herself. So she asked AI—specifically, Codex—to translate her vision into CAD commands and produce the design. The result is a dress that demonstrates the higher level of abstraction AI affords: a non-technical founder achieved manufacturing-grade complexity by speaking her intent aloud.

If AI can translate intent into CAD's grammar, it can do the same for Salesforce, Figma, or any software with documented logic.

Product Design Is Inverting

This flips how software is built. Traditional design prioritizes simplicity—never-read-the-manual, jump in and try. But software designed for AI agents must be the opposite: complex and well-documented.

The product manager's job shifts from designing every screen to building the "harness" that lets an agent act safely, completely, and without being taught where anything is. Documentation moves from optional reading to a first-class design artifact—as essential to applications as APIs are to infrastructure.

Meanwhile, the UX lab evolves. Instead of watching humans navigate interfaces, teams evaluate how agents operate software.

Expertise Still Matters—It Just Moves Upstream

Experts don't disappear. In software engineering, AI writes most of the code, yet hard problems still demand someone who understands the system underneath. Depth shifts from syntax to systems architecture.

The same pattern holds elsewhere. With AI handling the interface translation, expertise becomes knowing how the pieces fit together—the relationships between accounts and opportunities, the constraints of a manufacturing process, the logic of a design system.

This mirrors Dieter Rams' design philosophy: everything starts as noise. Earn the right to keep only what's essential. English lets a person do the same with software: say what matters, and let the agent absorb the rest.

Conclusion

AI fundamentally changes how people interact with powerful tools. By making English the interface, agents democratize access to software that once demanded months of training. The future of product design isn't simpler screens—it's better documentation, safer agent harnesses, and a shift in expertise from operation to architecture.

The dress Yana Welinder designed proves it's already here.

---

Original source: Previously Unmanufacturable

powered by osmu.app