Every App Will Be AI-Native (And Most Companies Aren't Ready)

The Claude Code model is coming for every software category. Here's what that actually means for incumbents and startups.

I’ve been using Claude Code for a few months now. It’s changed how I think about software.

Not because it writes code for me. Because it showed me what AI-native actually means.

And it’s coming for every software category.

What “AI-Native” Actually Means

Let me be clear about what AI-native isn’t:

  • Adding a chatbot sidebar to your app
  • Auto-generating description fields
  • Suggesting next actions with a sparkle emoji

AI-native is:

  • Conversation as the primary interface — not a feature, THE interface
  • Agent execution, not just suggestions — it does things, not just recommends
  • Context that persists and compounds — it remembers and builds on previous work

The difference between “AI-enhanced” and “AI-native” is like the difference between a horse with a motor strapped on and a car. Fundamentally different architectures.

The Claude Code Model

Here’s what using Claude Code actually looks like:

  1. I describe what I want in plain English
  2. It reads my codebase to understand context
  3. It makes a plan
  4. It executes — writes files, runs commands
  5. It validates — runs tests, checks for errors
  6. It iterates until the job is done

I don’t tell it how to do things. I tell it what I want. It figures out the rest.

This workflow is natural. It’s how we describe tasks to humans. And it works shockingly well.

This Model Applies to Everything

The Claude Code pattern isn’t just for coding. It works for:

  • Code — Claude Code, Cursor, Windsurf (we’re here now)
  • Design — “Landing page, dark theme, SaaS style” → done
  • Data analysis — “Analyze my expenses, show trends” → charts appear
  • Writing — “Blog post about X in my voice” → draft ready
  • CAD — “Phone stand, 45 degrees, thick base” → printable model (I’m building this)
  • CRM — “Draft follow-ups for stale opportunities” → emails written
  • Spreadsheets — “Project next quarter based on these assumptions” → forecast done

Every software category. Same pattern.

Why Incumbents Will Struggle

Adobe, Autodesk, Microsoft, Salesforce — they’re all adding “AI features.”

But they can’t go truly AI-native. Their architecture won’t allow it.

The problem:

  • Built for mouse-and-menu interaction over decades
  • Data locked in proprietary binary formats
  • Workflows assume human step-by-step execution
  • Business model requires complexity (training programs, consulting, certifications)

You can’t make a horse into a car by adding more horsepower. You have to start over.

And incumbents can’t start over. They have millions of customers depending on the existing paradigm. They have revenue tied to the existing complexity. Their incentives are fundamentally misaligned.

The Interface Shift

Every piece of software you’ve ever used made you learn its language. Where to click, which menus to open, what keyboard shortcuts to memorize. You spent hours watching tutorials before you could do anything useful. That learning curve was the product’s moat — and the reason companies could charge thousands per seat.

AI-native flips this completely. The tool learns your language. You describe what you want in plain English, and it figures out the rest. Someone who’s never touched CAD software can get a printable 3D model in minutes. Someone who’s never written a formula can get a full financial analysis.

That expertise barrier everyone spent years building? It just became a liability. And the complexity that justified premium pricing? It’s the exact thing users are trying to escape.

The Skill Shift

For decades, being “good with software” meant knowing where things were. The right menu, the right shortcut, the right sequence of clicks. People built entire careers on mastering specific tools — Photoshop experts, Excel wizards, Salesforce admins.

When the tool does the execution for you, that expertise stops mattering. What matters instead is knowing what to build and whether the output is actually good. A designer who understands visual hierarchy and user psychology will run circles around someone who just knows Figma’s keyboard shortcuts. A sales leader who understands deal dynamics will get more from an AI CRM than someone who memorized every Salesforce workflow.

Domain expertise wins. Taste wins. Judgment wins. The people who understand the problem will thrive. The people who only understood the tool will need to adapt fast.

The Opportunity

Every software category will be rebuilt AI-native.

Not by the incumbents. By startups that:

  • Start with conversation as the interface
  • Build for agent execution from day one
  • Design for context persistence
  • Charge for outcomes, not seats

$100B+ of enterprise software is waiting to be disrupted.

The playbook is clear:

  1. Pick a category with painful tools
  2. Build the AI-native version
  3. Undercut on price (lower complexity = lower support costs)
  4. Win on experience (instant productivity vs. months of training)

What I’m Building

I’m starting with CAD because:

  1. The incumbents are especially slow and resistant
  2. The market is huge (3D printing, manufacturing, prototyping)
  3. The learning curve is brutal (50-100+ hours minimum)
  4. LLMs are already good at generating parametric code

But the model applies everywhere.

Describe what you want → Get what you want.

That’s the future of software. And it’s closer than most people think.

This is part of my building in public series. I’m working on AI-native CAD tooling — follow along for updates.