How Fast Can Software Die?

What this article answers (plain language): Software companies can be displaced faster than almost any other type of business because there's nothing physical protecting them. AI is accelerating this even further by collapsing switching costs. But the disruption wave slows down as it moves from pure software into industries with physical infrastructure.

In Part 1, we looked at why platforms that don't open up to AI agents will get replaced. Now the harder question: how fast?

Figma was a $20 billion acquisition target in 2022. Adobe's deal collapsed under regulatory pressure in 2023. Figma IPO'd and hit $122 per share. Today it's at $25 — down 80% from its high — and dropped another 8% in a single day when Google launched Stitch.

Eighteen months from peak to existential question mark. That's the speed of disruption in software.

Why Software Has No Moat Against AI

When AI disrupts a manufacturing company, the factory still exists. The supply chain still works. The physical infrastructure creates friction that slows displacement — you can't 3D print a semiconductor fab overnight.

Software has none of that. A SaaS product is code running on someone else's servers, accessed through a browser. The only things protecting the incumbent are switching costs and user habits. Both of those are precisely what AI is eroding.

Switching costs drop when an agent can read your data from one platform and write it to another. User habits become irrelevant when the user isn't a human navigating a GUI — it's an agent operating programmatically. The two moats that protected SaaS for a decade are both dissolving at the same time.

This is why the SaaSpocalypse narrative has teeth — not because AI replaces all software, but because AI makes it trivially easy to switch between software. And when switching costs approach zero, the incumbent advantage disappears.

The Disruption Gradient

But here's what most AI disruption takes miss: the speed isn't uniform. There's a gradient, and it follows physical friction.

Fastest (months): Pure SaaS — design tools, project management, CRM, content creation, analytics. No physical component. The product is the interface, and AI-native alternatives can replicate or surpass the interface overnight. Figma to Stitch is the template.

Fast (1-3 years): Hybrid digital — fintech, edtech, media. Digital core but with regulatory requirements, compliance layers, or content licensing that slow displacement. You can build an AI-native learning platform in weeks, but accreditation takes years.

Moderate (3-7 years): Digital-physical bridge — logistics, real estate, insurance. Software controls the workflow, but the value chain includes physical assets, human relationships, and regulatory frameworks. AI improves the software layer quickly, but the physical layer resists.

Slow (7-15+ years): Physical-first — manufacturing, healthcare, energy, construction. Software is one layer of a deeply physical operation. AI transforms the software layer fast, but the physical infrastructure, safety regulations, and capital requirements create genuine friction.

The mistake most people make is treating AI disruption as uniform — either everything changes overnight or nothing does. The reality is a wave moving at different speeds through different substrates. SaaS is the beachfront property that floods first.

What This Means for Builders and Investors

If you're building in a pure SaaS category, your competitive window is measured in months, not years. The AI-native alternative to your product might not exist today but could launch next Tuesday. Your moat isn't your feature set — features can be replicated by a model in hours. Your moat is accumulated context, user data, and relationships that a new entrant can't bootstrap overnight.

If you're investing in SaaS, the question isn't "does this company have good technology?" It's "can this company be replaced by an AI-native alternative in one product cycle?" If the answer is yes, the risk premium should be much higher than traditional SaaS multiples suggest.

And if you're an incumbent: opening up is no longer optional. The platforms that expose themselves to agents — via APIs, MCP, A2A — at least give themselves a chance to be part of the AI-native stack rather than replaced by it. The ones that stay closed, hoping their GUI moat will hold, are betting against the direction of every protocol, every standard, and every developer building agents today.

Figma dropping 8% in a day isn't an anomaly. It's a preview of what happens to every closed platform in the blast radius of an AI-native competitor.

The only variable is speed. And in software, speed is the one thing that should scare you.

This is Part 2 of "The Displacement Pattern." Part 1: Open Up or Get Replaced explains why platforms that block AI agents from accessing their systems will be bypassed entirely.