Greg says SaaS is dead, and the first comment rebuts him with a Stripe bill

Greg Isenberg posted a diagram with a very direct caption: "SaaS is dead." He added that he consults on SaaS for the world’s biggest companies and would explain in 11 minutes where it’s headed.

The diagram isn’t complicated. On the left is old SaaS, selling tools per seat, crossed out in red. On the right is so-called SaaS 2.0: tools are free for individuals, thickened by data and network effects, and the money-making spots move to two boxes—one called "intelligence provided by agents," the other "service provided by humans."

This framing spreads well because it turns many companies’ anxiety into a three-panel cartoon: first give software away for free, then sell AI agents, and finally sell agency services. Software becomes the entry point; execution and backstopping are where the money is.

The comments aren’t so tidy. The first reply, from Amy, reads: "dead but my stripe says otherwise." In other words, you say it’s dead, but my Stripe bill disagrees. That reply is more vivid than the original post.

Others dig in seriously. Keith Oliver Rull says it isn’t dead, just evolving dramatically. He sees plenty of people building their own personal software, companies building their own tools, and some people not using the SaaS-provided interface at all—just using an API key or MCP connection into Claude or an internal system. MCP can be understood as an interface between models and external tools; the key point is that users are starting to bypass ready-made interfaces and buy capabilities directly.

Danni Chen asks something more basic: are OpenAI and Anthropic SaaS companies? Monthly subscriptions, access to cloud software—by definition, they’re at least adjacent. Beth’s reply is sharper: SaaS isn’t dead; it just finally has no excuse to keep being lazy about pricing.

The really interesting part of Greg’s diagram is that the charging point moves downstream. In the past you sold tools and charged per seat: the bigger the company and the more people, the bigger the bill. Now tools can be free because customer acquisition and data entry points are more valuable; what you actually charge for is agents executing tasks for you, and humans handling the parts agents can’t. An agent is an AI program that can break down tasks, call tools, and run workflows on its own. It’s not just a chat box.

But this path isn’t easy. In the replies, Luis Junior jokes that it loops all the way back to selling people’s time, just under a new name, then asks whether the gross margins on agency services can support SaaS valuations. RAY is more direct, calling the diagram a lock-in plan with a friendly face. Nils Liesebach also asks: if software becomes a free sample and services become the product again, how do VCs price it at Series A?

These doubts are closer to reality than "SaaS is dead." Free tools aren’t charity; they’re the new entry point. Charging for agents tests whether they can reliably deliver results. Charging for human services tests margins and scalability. B Thomas Collins II puts it plainly: people don’t really pay for software; they pay so they don’t have to deal with it themselves. Mees Tierolff proposes charging for outcomes—no result, no payment. It sounds satisfying, but to do it you first have to define what counts as "solved."

So whether SaaS is dead depends on which SaaS you mean. The per-seat sale of generic tools is indeed loosening. Companies treating software as a free sample and agents plus human services as the product are testing a new ledger. Greg’s original post is here: https://x.com/gregisenberg/status/2116212861466755509 , and the full video is here: https://youtu.be/AqqjH_uv-_w?si=m14FjqHv_PZU5VEy .