"Let the financial SaaSpocalypse begin." That was the line in our community chat the day we opened Operation: Build it Better. It is a good line. It is also a claim about the world, so this piece does what we try to do with every claim: say where the word comes from, lay out the evidence with sources, give the other side its strongest form, and only then say where we stand.
Where the word comes from
"SaaSpocalypse" is the name markets gave to the early-2026 repricing of software companies, once AI agents started doing the work those companies sell. The term is credited to Jeffrey Favuzza, a strategist at Jefferies (The Batch). The sequence, as reported:
- January 12, 2026. Anthropic introduces Claude Cowork, an agent built for professional work.
- January 30. It publishes 11 open-source Cowork plugins, each aimed at a white-collar job function — legal, finance, sales, research.
- By February 3. Roughly $285 billion of software and data market value is gone (Tech Startups).
- January 12 to February 23. The S&P Software & Services Index loses about 25% (The Batch).
The single-day moves were the part people remember: Thomson Reuters down 15.8%, its worst day on record; LegalZoom down about 20%; Intuit down 11%; Salesforce and ServiceNow down 7% each (CNBC, Tech Startups). Look at that list again. The hardest-hit names were legal, tax and information businesses — not generic productivity software.
The argument in one sentence
Simon Taylor of Fintech Brainfood put it as a question: "why pay for ten software licenses when one AI agent handles the workflow?" (Fintech Brainfood, February 9). Per-seat pricing ties a vendor's revenue to its customer's headcount. An agent that does the workflow cuts that tie. Taylor's supporting evidence: about 4% of public GitHub commits now written by Claude Code, up from almost none thirteen months earlier; AI budgets up about 100% while IT budgets rose about 8%; Norges Bank reporting roughly 213,000 hours saved; Goldman Sachs deploying agents for trade accounting, reconciliation and client onboarding.
He was not first. Andreessen Horowitz argued in January 2025 that the two-decade SaaS rule — streamline a human task into software, then charge per user — is "no longer valid", and IDC expects pure seat-based pricing to be obsolete by 2028 (Userpilot's survey of the pricing debate). There is a second-order exposure too: Bloomberg documented how heavily private equity leaned on software subscriptions as collateral for leveraged buyouts (Bloomberg), and one tally puts software value lost over twelve months near $2 trillion (Tech Insider).
Finance got its own version
Wealth management. In early February a fintech startup, Altruist, launched an AI tax-planning agent. Within hours Charles Schwab fell 7.4%, Raymond James 8.8% — its worst day since March 2020 — and LPL Financial 8.3%, erasing more than $100 billion in value (Lex Sokolin, WealthManagement.com). McKinsey's read: "Wealth management is now experiencing its version of SaaSpocalypse," with investors now treating AI as "a structural threat to traditional business models" — and the greatest risk sitting at the client interface, where a digital-first competitor can take the relationship.
Financial data and research. On February 24 Anthropic connected Claude to FactSet, MSCI, S&P Global and LSEG, and FactSet rose about 7%: being plugged into the agent looked like distribution. On May 5 it shipped ten finance agents — earnings analysis, market research, financial modeling, audit — and FactSet and S&P Global traded lower, because those are the jobs analysts use those terminals for (Anthropic, Sherwood News). Same vendors, same partner, opposite reactions, ten weeks apart. That is what a moat being re-measured looks like.
The conversation has since reached crypto as well (Yahoo Finance).
The strongest case against
We would be doing the thing we criticise if we stopped there.
- The stocks came back. Software is up nearly 40% from the low (CNBC, September 3), and Yahoo Finance wrote that the panic "as we once knew it is over" (Yahoo Finance) — though the argument is far from settled (CNBC, August 7).
- The incumbents are reporting growth. On August 26 Salesforce posted revenue of $11.35 billion, up 11%, and the stock rose more than 20% on an expanded Anthropic partnership. Marc Benioff: "This nonsense of this SaaSpocalypse, I think it's time for it to stop," and "Frontier models depend on CRM. They don't replace it." (The Daily Upside, Benzinga).
- Real moats hold. A March 2026 a16z analysis argues AI does not kill software, it splits it: systems of record, proprietary data and regulatory embedment get stronger; thin wrappers do not. The first version of a product is "~2% of the work" — permissions, security, integrations, support and compliance are the other 98% (a16z moat scorecard, summarised; see also SignalFire, Forbes, Thoughtworks). The Batch's summary is the fairest one-liner: "SaaS isn't dying, it's becoming AI-native."
So what is actually ending
Put the two halves together and the picture is specific. What is ending is not software. It is the ability to charge a monthly seat fee for a thin tool — a screener, a dashboard, an alert, a calculator — whose whole value is a presentation layer over data the customer could reach another way. Those tools were the 2%. They are now a day's work for an AI pipeline. What survives is whatever sits on proprietary data, carries regulatory weight, or holds the record of who owns what.
Finance is full of the first kind. It is why the February damage concentrated where it did, and it is why the word "financial" belongs in front of "SaaSpocalypse".
Where we sit
Build it Better is that argument turned into a service. Name a financial app you pay for and say what it should do better — or describe one that does not exist yet — and our AI pipeline rebuilds it on EventTrader, in the open, on a public board, against a 24-hour clock. You judge the result with a pass or a fail. More than a hundred financial apps have already been through it. Filing is open to any account holding at least 100 ET10 — nothing is charged or locked — and the board is open to everyone. AI agents can file requests as well as people: API, MCP and SDK.
We hold ourselves to the counter-case too. Automatic builds are tools that read: screeners, dashboards, alerts, analytics, research. Anything that moves money — deposits, withdrawals, orders, signatures — is reviewed by a person before it ships, because that is the 98%, and nobody should pretend a weekend build replaces it. Every rebuild says plainly what it does not do, and ships under our own name, never the original's.
The SaaSpocalypse did not begin with us; by the evidence above it began in February. What begins here is narrower and more useful: if a financial tool's only moat was the invoice, it can now be rebuilt, better, for whoever asks.
Figures are as reported by the linked sources and were not independently audited by us. Nothing here is investment advice.