Open Up or Get Replaced

What this article answers (plain language): Software platforms that don't let AI agents access, read, and operate them will get replaced by AI-native tools that are built transparent from day one. This is already happening in design. Every software category is next.

Google launched Stitch this week — an AI-native design tool where you describe what you want in plain language and get production-ready UI. No wireframes. No design skills. Just intent in, interface out.

Figma dropped 8% in a single day.

The reaction is understandable but misses the real story. This isn't about Google vs. Figma. It's about what happens to any software platform that was built for humans to click through — when the primary user is increasingly an AI agent.

The Tool That Requires a Designer Loses to the Tool That Is the Designer

Figma is a brilliant product. But it was designed for a specific workflow: a trained designer manipulates layers, components, and frames through a visual interface. The value lives in the designer's skill with the tool.

Stitch skips the tool entirely. You tell it what you need. It builds it. It can even generate working React code from the output.

This isn't Figma with an AI sidebar bolted on. It's a fundamentally different architecture — one where the AI isn't assisting a human operator. It is the operator.

Figma is responding. They've partnered with Anthropic on "Code to Canvas," trying to position themselves as the bridge between AI-generated code and human-centered design. But the structural challenge remains: Figma is a tool that needs a human in the loop. Stitch is a tool that doesn't.

The Protocol Question

Here's where this gets bigger than design.

Every software platform built in the last 20 years was designed for human users navigating graphical interfaces. Menus, buttons, dashboards, drag-and-drop. The entire UX paradigm assumes a person sitting at a screen.

AI agents don't sit at screens. They need structured access — APIs, readable state, documented data models. They need to see into the system, understand its current state, and act on it programmatically.

The platforms that will survive the agent era are the ones that open up. Not just adding an "AI feature" to the existing product, but making the entire system transparent and operable by agents. This means embracing protocols like MCP (treating data as readable resources), A2A (declaring capabilities via Agent Cards), and open APIs that expose the full state of the system — not just a thin layer of approved actions.

The platforms that don't open up will be bypassed. Not slowly. Not over a decade. In months. Because an AI-native alternative that's transparent from day one doesn't need to "add AI" — it already is AI.

Adding AI Features Is Not the Same as Being AI-Native

This is the distinction most incumbents are getting wrong. Adding Copilot to Office. Adding AI generation to Figma. Adding smart suggestions to Salesforce. These are features bolted onto architectures designed for human operators.

An AI-native tool is built differently from the foundation. The primary interface is language, not GUI. The data layer is readable by default, not locked behind visual abstractions. The system exposes its state to any agent that asks — it doesn't hide it behind dashboards only humans can parse.

The gap between "AI-enhanced incumbent" and "AI-native alternative" will widen, not shrink. Because every AI feature an incumbent adds still operates within the constraints of an architecture built for clicks. The AI-native tool has no such constraint.

The Pattern

Stitch vs. Figma is just the most visible example this week. The same pattern is playing out — or about to play out — across every software category. CRM, project management, analytics, content creation, accounting. Each one has incumbents built for human operators and AI-native challengers built for agent operators.

The question for every platform: can an AI agent see your data, understand your state, and take action? If not, something else will replace you — and it will happen faster than you expect.

That speed is what Part 2 of this series is about.

This is Part 1 of "The Displacement Pattern." Part 2: How Fast Can Software Die? explores why SaaS disruption happens at a pace that's hard to comprehend — and where the wave goes after software.