If you follow tech news, you’ve probably heard that software is finished. In early 2026, a wave of selling hit software stocks, and a Jefferies analyst coined a name for it that stuck: the SaaSpocalypse. The story was tidy and scary. AI agents can do the work of whole teams, companies will cancel their subscriptions, and the businesses that sold them will wither. It makes for a good headline. It also skips a lot of the story.
What Actually Happened
The selloff was real. Depending on whose count you trust, somewhere between several hundred billion and about two trillion dollars of market value left software stocks in a matter of weeks. One widely cited trigger was a run of AI agent product launches, including Anthropic’s Claude Cowork and its plugins, which showed agents handling tasks that used to need dedicated software and staff. The software ETF IGV fell more than 20 percent in the first part of the year, and one broad SaaS index was down about 26 percent by the end of March. Investors weren’t imagining a threat. They were repricing one.
The Fear Behind It: Per-Seat Pricing
Here’s the part worth understanding. A lot of software is sold per seat. More employees means more licences means more revenue. If an AI agent lets one person do the work of several, companies may buy fewer seats, and that squeezes revenue. This isn’t a fantasy. Pricing data shows pure per-seat models fell from about 21 percent of SaaS companies to about 15 percent in a year, while hybrid pricing jumped from 27 to 41 percent. So the pressure is genuine. But notice what that data actually describes: a business model adapting, not a product category vanishing.
Why “Dead” Is the Wrong Word
A few things keep the doom story from holding together. First, AI agents don’t work in a vacuum. They need context and data, and that usually lives in the systems of record companies already run: financial ledgers, customer databases, enterprise platforms. An agent that can’t reach reliable, permissioned data isn’t much use, which makes those systems more valuable, not less. Some analysts argued the sell-off ignored strong double-digit growth in contracted future revenue at large software firms, a sign that customers weren’t actually leaving.
Second, replacing software with a homemade AI build sounds easy until you think about security, compliance, uptime, integrations and support. Regulated industries in particular tend to want deterministic, auditable processes, and vendors are already weaving AI agents into those processes rather than being pushed aside. Forrester has noted exactly that pivot is underway.
Software Is Embracing the Agents
The vendors aren’t standing still either. Salesforce reported delivering 2.4 billion “Agentic Work Units” in a recent quarter, tasks completed autonomously by agents inside its own platform. Gartner projects around 40 percent of enterprise applications will include AI agents by the end of 2026, up from under 5 percent in 2025. IDC expects 70 percent of software vendors to move away from pure per-seat models by 2028. That’s a reshaping of how software is built and charged for, not a funeral.
Where the Pain Is Real
Being honest means admitting some software is more exposed than others. Tools that exist mainly to log or automate simple tasks, where an agent can do the same job directly, face the most pressure. Products built around thin workflow layers or easily copied features have the least protection. Usage-based infrastructure businesses tend to look sturdier, since agents can increase consumption rather than reduce it. So the outlook is mixed: fragile software gets squeezed, durable platforms often get stronger.
What It Means for Buyers and Builders
If you buy software, the practical advice is to renegotiate. Contracts built around seats are likely to shift toward usage or outcomes, and some vendors are already offering flexible credits to use across people and agents. If you build software, the lesson is to own something agents depend on: proprietary data, trusted workflows, integrations, compliance. Features alone won’t protect you for long.
The Bottom Line
The SaaSpocalypse is a dramatic name for a real but narrower shift. Per-seat pricing is under pressure, weaker software is being exposed, and markets overreacted in places. Commercial software isn’t dying. It’s being forced to prove what it’s actually worth, and the companies that adapt will probably come out sharper than before.
