Most enterprise software companies are the walking dead.
They just don’t know it yet.
That’s a strong claim. Let me show you why I believe it.
I’ve spent the last decade working in B2B enterprise software companies Workday, Dayforce, and now my own.
What’s your moat as an enterprise software company?
That question used to have an easy answer.
Multi-year contracts.
Data conversions.
Compliance baked in.
Implementation costs that made switching painful.
Your customers stayed because leaving was expensive.
On February 3, 2026, the answer changed.
$285 billion in 48 hours
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. Company record. LegalZoom fell 19.68%. RELX, parent of LexisNexis, had its steepest fall since 1988.
A Jefferies trader coined a term that morning: “SaaSpocalypse.”
You watched the rest play out over the next 30 days.
0 trillion
in software market cap disappeared in 30 days
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.
Dan Ives at Wedbush: “Software right now is under massive pressure because AI is eating their lunch.”
Jason Lemkin did the math: “If 10 AI agents can do the work of 100 reps, you need 10 Salesforce seats, not 100.”
a16z published an explicit thesis: “Good News: AI Will Eat Application Software.”
Fewer humans means fewer seats. That’s demand destruction. Value destruction is worse. Intuit spent billions on AI integration. Called themselves “an AI-driven expert platform” since 2023.
Their stock dropped 50% anyway.
I’ve watched this from the inside
I was lead product architect at Workday. We had 5,000+ customers and 20+ SKUs, and nobody could tell you which customers used what, or why.
I built a framework to answer that question, got a patent for it, and that understanding is what helped Workday grow from $2B to $4B.
The framework was valuable because it encoded how enterprise customers adopt HR and finance software.
The workflows in enterprise software used to be hard to replicate because software engineering and compliance were hard to replicate.
Not anymore.
But the patterns, the compounded learning of why 5,000+ companies use them differently, are harder to replicate.
The WHY behind the workflows, and the true causal structures that led to their decision making.
The true moat of any company, I’ve always believed, is in its domain specific learning. In a business context, there are causal structures that lead to someone buying and using your product or service.
If you’re in product, you’ll be familiar with the Jobs to be Done lens.
Causal structures that lead to why they buy and love your product.
That’s what I saw earlier in my career, and it is even truer now in the age of AI.
After Workday I was Chief Product Architect at Dayforce, a $1.8B company where I helped improve AI adoption across their engineering org from 15% to over 50% in two months.
Now I run Epistemic Me building Clarity.
Based on all of my experiences, I keep coming back to this prediction: the companies that will survive the SaaSpocalypse compound their learning.
They will solve for creating better domain specific maps of these causal structures, from their specific territory, than the frontier labs can.
Everything else is a feature that Anthropic or OpenAI can ship next week to ruin your life and tank your market cap.
How businesses actually make money
Let’s take a step back a bit.
How do businesses make money?
At the end of the day, businesses are just a bunch of people (and now agents) coming together with a goal: make money.
People driving execution of learning loops to capitalize on market demand by investing in and delivering on products and services to fulfill that demand.
How do they do that?
There’s a quote I love, “the map is not the territory”. It relates here.
People in companies go make contact with the market (potential customers), and make maps (persona cards, slide decks, landing pages, etc.) of the territory, to share with their teammates to execute on business goals:
Create and refine an asset or service that fulfills on the demand, make money, then make more money.
Now, let’s back up a little bit and align on the concept of compounded learning, and maximizing information flow to the people (and now agents) driving that execution.
2,000 years of information routing
Jack Dorsey published something on X recently that speaks to exactly this.
He calls it “From Hierarchy to Intelligence” and the core argument resonated from all of my experience working in B2B enterprise software.
His point: 2,000 years of organizational design was information routing built for human limitations.
The Roman Army invented span of control.
Prussia invented middle management.
McKinsey in the 1900s codified the matrixed organization.
Spotify tried squads.
Zappos tried holacracy.
Valve tried flat hierarchy.
Every one of those experiments reverted to hierarchy at scale because there wasn’t a tool that could replace what those layers did.
Until now.
Your information flow is your product quality
Let’s bring another concept into the conversation. Bear with me.
If you’re in tech, you probably have heard of Conway’s Law. Conway’s Law says organizations design systems that mirror their own communication structure.
Your information flow determines your product quality.
If your org fragments learning across Slack threads, Jira tickets, Confluence pages, and people’s heads, your product reflects that fragmentation.
You’ll ship the wrong thing faster.
Traditional organizational hierarchy has been a necessary tool for information flow. It hasn’t been perfect, to be clear. It’s been good enough for millennia.
And now things are changing.
Four things replace the hierarchy
Jack talks about four key concepts for this new world of business:
Capabilities. Atomic primitives. Payments, lending, banking, for their case. Hard to acquire, no UIs of their own. Building blocks.
World Model. This is the part that gets me. He splits it into two: the company world model (how the org understands itself, its operations, priorities, performance) and the customer world model (per-customer understanding built from proprietary data).
He writes: “Money is the most honest signal in the world.”
Very true. So how do you build a harness around that signal for YOUR business? That’s the real question.
More on that in Issue 2.
Intelligence Layer. Composes capabilities into solutions for specific customers at specific moments. His example: a restaurant’s cash flow tightens before a seasonal dip. The intelligence layer composes a short-term loan and surfaces it before the owner thinks to look. No product manager scoped that solution.
Interfaces. Delivery surfaces. Where solutions arrive. The model is where they’re born.
People move to the edge. ICs build. DRIs own cross-cutting problems. Player-coaches combine building with developing people. The system coordinates.
The question he poses is the key one to ask of your own company:
“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.
Continue reading
Get the full newsletter, free.
Join founders and builders who read Self Aligned every week.
Two world models, one substrate
I’ve been thinking about Dorsey’s “customer world model” and “company world model” through the lens of what we build at Clarity, and the distinction maps cleanly.
We make digital twins of people, customers, and organizations. We specialize in modeling subjective experience of an individual with a well-defined data structure and set of interfaces that are conducive to compounded learning.
Our primitive is the Self Model.
Customer world models: how the company understands and organizes its model of who the customer is and why they do what they do (and why they buy).
For this, we create per-user Self Models with Clarity that predict behavior, connecting marketing to product engagement to payment, and create a learning loop.
B2C: In my experience it’s more obvious to see this in any transactional B2C business where the value chain between user behavior and revenue is tight and fast-cycling has this shape. We got our first customer live at the beginning of this year and they went up 60% in monthly app revenue.
B2B: Less obvious in B2B enterprise software because the data patterns take shape across a longer timeline. But, from my Workday experience helping instrument product adoption data tied to customer success data, it is longer tail but still just as high leverage.
Company world models: how the company understands and organizes its own capabilities, to execute to fulfill their customers’ problems to capitalize on demand.
If you’re a VP of Engineering, you care about how 1,000+ engineers coordinate around AI adoption.
You want someone to own the evals roadmap for example to build better AI products than the competition.
You want a great feedback loop of execution from the territory to your maps, to your people to drive better execution.
But traditional hierarchy makes it difficult: telephone games, political games, land grabs, and more.
These are internal coordination failures.
Every enterprise has different ones.
Both types run on the same substrate: a model of self.
One for external (customers), and one for internal (company).
Whether you’re predicting a user’s session quality or mapping where an engineering org’s AI adoption and execution breaks down, the infrastructure underneath is the same to us with Clarity.
Learning at the edge
I was talking to a design and research team this week at an enterprise software company. If you’ve worked in enterprise, you know this world.
If you’re an IC reading this, you’ve been here. You’ve gone and done hard work to make contact with the real world of your customers.
Product, design, and research go learn from the territory.
They talk to users. They observe behavior.
They come back and build artifacts (maps): personas, journey maps, storyboards.
And those artifacts start decaying the moment they’re created.
There’s no system to hold it, or compound it and distribute at scale.
They live in a wiki that nobody visits. In a Confluence page last edited in 2023. In someone’s head, and when that person takes a new job, the learning goes with them.
For B2B enterprise software companies in particular, it was always obvious to me that the map is not the territory.
The deck (map) you made last year does not survive first contact with the market (territory) this year. Especially when the technology underneath us shifts every day at rapid pace.
Every function sees different terrain
Every function learns from the territory through a different lens.
Design learns the user journey.
Product learns the market opportunity.
Engineering learns the technical constraints.
Sales learns what makes someone sign.
Each function has a subjective view of the territory. Each one builds their own map.
The old model: those maps are fragmented. Siloed. Design has their personas. Product has their PRDs. Engineering has their architecture docs. Sales has their battlecards.
Nobody has the full picture.
Conway’s Law plays out: the product mirrors that fragmentation. If you work in B2B enterprise software, especially on the technology side, you feel this everyday. I know it because I’ve been there.
The new model: a continuously evolving map that everyone contributes to simultaneously, from their own subjective lens.
Same substrate, different perspectives. The map gets richer every sprint because every function adds what they learned from the territory.
Design adds user behavior observations.
Product adds market signal.
Engineering adds technical constraints and decisions.
Sales adds what closed and what didn’t, and why.
In this new world: every function can contribute AND distribute their learnings at scale for both the company and customer world models.
That’s what our Self-Model is:
A shared substrate where subjective learning from every function compounds into a unified, continuously updated understanding of the territory you operate in.
In essence, we help companies build their company and customer world models. AKA, we help you build and maintain your moat against the frontier labs eating away the world.
Why enterprise never had to care
Most B2B enterprise software companies I’m aware of have terrible product analytics.
Why would they care who is using what feature, when they’ve locked up multi-year deals that pay them regardless of the outcome?
That was tolerable when SaaS margins were high and retention was 95%.
Ship based on a point-in-time map.
That’s fine, because your customers’ switching costs were too high.
That is very unlike B2C, where the switching costs are generally minimal.
So there traditionally has been no incentive for these companies to keep the map up to date, let alone it being a hard problem to solve.
All because you locked up the revenue behind a multi-year deal with a moat around the gigantic implementation cost and effort to get them onto your platform in the first place.
But data conversions, integrations, and features are now being eaten up by AI. Switching costs are lower than ever, and will become infinitely lower as time goes on.
For anyone watching the enterprise software space, that is obvious.
The features?
The features are now something Anthropic can ship as a plugin overnight.
So what’s your moat as an enterprise software company?
The real moat
Domain-specific learning that compounds.
Your understanding of your customers and your business, encoded into self-models that persist and evolve.
The more customers use it, the smarter it gets.
The smarter it gets, the harder it is to compete with.
Anthropic and OpenAI can’t go map your territories like you can.
But are you empowering your people to build that map in the first place?
If not, then mark my words:
You are the walking dead.
References
- The SaaSpocalypse: $285B Wiped, AI Agents Rising — Taskade, 2026
- Thomson Reuters, RELX, and Wolters Stocks Crushed — Morningstar, 2026
- AI software scramble: Anthropic triggers stock market slide — Axios, 2026
- Intuit was an AI pioneer. Why its stock became a SaaSpocalypse casualty — Fortune, 2026
- The SaaSpocalypse Deepens: Jefferies Downgrades Workday and DocuSign — FinancialContent, 2026
- Software Sector Repriced on AI Agent Fears — 24/7 Wall St, 2026
- Good News: AI Will Eat Application Software — a16z, 2026
- Cursor’s crossroads — Fortune, 2026
- From Hierarchy to Intelligence — Jack Dorsey, X, 2026
- Harness Engineering for Coding Agents — Martin Fowler, 2026
- How Do Committees Invent? — Melvin Conway, 1968