The SaaSpocalypse Survival Guide: Why Domain-Specific Learning Is Your Only Moat

On February 3, 2026, Anthropic released Claude Cowork with a legal plugin.

Within 48 hours, $285 billion in market cap evaporated. Thomson Reuters lost 15.83% in a single day. LegalZoom fell 19.68%. A Jefferies trader coined the term that morning: “SaaSpocalypse.”

Over the next 30 days, $2 trillion in software market cap disappeared. ServiceNow down 41% year-to-date. Intuit down 50% from peak. Workday’s price target slashed from $325 to $150. Atlassian reported its first-ever enterprise seat decline.

These companies built good software and shipped reliable products for decades.

Then Anthropic replicated their core functionality as a plugin.

This isn’t a correction. It’s a repricing.

a16z published an explicit thesis: “AI Will Eat Application Software.”

Jason Lemkin did the math: if 10 AI agents do the work of 100 reps, you need 10 Salesforce seats, not 100.

That’s not a pricing problem. That’s demand destruction.

And value destruction is worse. Intuit spent billions on AI integration. Called themselves “an AI-driven expert platform” since 2023. Their stock dropped 50% anyway.

Calling yourself AI-native isn’t a moat. Features aren’t a moat. So what is?

Every moat you had just collapsed

Features are plugins now. Anthropic replicated Thomson Reuters’ core legal functionality overnight. Your compliance engine, your workflow builder, your reporting dashboard — all features that can be shipped as an AI plugin.

Switching costs are collapsing. Data conversions, integrations, and compliance — the things that made leaving expensive — are being eaten by AI. Multi-year contracts are the last wall, and customers are negotiating shorter terms.

Per-seat pricing is dying. Fewer humans means fewer seats. That’s the math that repriced the entire sector.

You never needed to understand your users. Most enterprise SaaS companies have terrible product analytics. Why would you care who uses what when you’ve locked up multi-year deals? That was tolerable at 95% retention. Not anymore.

The only moat: domain-specific learning

Here’s the question that determines whether you survive:

“What does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day?” — Jack Dorsey

Companies where the answer is deep will use AI to reveal what they are.

Companies where the answer is nothing will use AI to cut costs until margin reaches zero.

They are the walking dead.

The real moat is domain-specific learning that compounds. Your understanding of your customers and your business — the patterns of how 5,000+ companies use your product differently, the causal structures behind why they buy, the WHY behind the workflows.

Anthropic and OpenAI can’t go map your territories like you can.

But are you building the infrastructure to compound that learning?

How to build a learning moat

The learning moat has two halves:

Customer world model — how the company understands who the customer is and why they do what they do. Per-user self-models that predict behavior, connecting marketing to product engagement to payment. The model gets smarter every interaction.

We built this for The Relationship Psychics — a session-based app spending $30K/mo on Meta ads that couldn’t see inside their product. By connecting their customer world model (marketing → product → revenue), we found $50K+ in annual revenue leaks and grew revenue 60%. Same traffic.

Company world model — how the company understands its own capabilities and coordinates execution. An organizational self-model that replaces hierarchy’s information routing with compounding learning.

Every function sees different terrain. Design learns the user journey. Product learns the market opportunity. Engineering learns the constraints. Sales learns what makes someone sign. Each builds their own map. Nobody has the full picture.

The old model: fragmented. Conway’s Law plays out — the product mirrors that fragmentation.

The new model: a continuously evolving map that everyone contributes to simultaneously. Same substrate, different perspectives. The map gets richer every sprint.

Both run on the same substrate: a model of self.

The Walking Dead Test

Ask yourself these three questions:

1. What does your company understand that’s genuinely hard to understand? If the answer is “our features” — you’re walking dead. Features are plugins now.

2. Is that understanding getting deeper every day? If your learning lives in Confluence pages last edited in 2023 and in people’s heads (and when they leave, the learning goes with them) — your learning is decaying, not compounding.

3. Can you connect that understanding to revenue? If you can’t trace from customer behavior to product engagement to payment — you have data, not a moat.

If you answered poorly to any of these, the gap between you and the frontier labs is closing every day. The good news: there’s still time to build the substrate.

Start with Sprint Zero

Sprint Zero is a 4-week, $15K diagnostic that answers the Walking Dead Test for your company. We map your domain-specific learning gaps — what you know about your customers and business, what you’re missing, and where the learning is decaying.

If the moat is there, we show you how to compound it. If it’s not, at least you know.

Find out if you’re walking dead →

---

This post is the companion piece to The Walking Dead of Enterprise Software on Self-Aligned.