For two decades, SaaS had a simple promise: take a messy human workflow, wrap it in a clean UI, put it in the cloud, become a System of Record, and charge per seat. It worked spectacularly—and then it got bloated. 

Feature sets ballooned to justify renewals. Platforms went horizontal to chase TAM. Systems of Record became systems for storing evidence of work, not systems for doing it. 

AI companies are playing a different game entirely. They’re not building better tools for humans—they’re abstracting the human out of the loop. The CRM, the project tracker, the analytics dashboard—all of it becomes commodity infrastructure that an agent draws from. The human doesn’t operate the software anymore, nor do they need to learn it. They set the intent. Agents execute. 

We’ve gone from User to Orchestrator and are moving from a world of Systems of Record to a world of Systems of Agency. That shift breaks many assumptions the SaaS playbook was built on. 

The Playbook That Built a Trillion-Dollar SaaS Industry—And Why It’s Breaking 

SaaS has been the undisputed king of software for a generation. It disrupted the legacy licensed model by offering recurring cash flows, high gross margins, and a single code base. Its “moats” were not necessarily technological, but behavioural: once a system of record was integrated into a company’s workflow, it became a nightmare to rip out and have humans change their way of working. Human habits are hard to change, and “edge-case” engineering created a formidable barrier to entry. 

Until the unit of value stopped being “the human”. 

Three things are breaking simultaneously. 

The first is pricing. When a single agent does the work of dozens of employees, seat-based pricing stops making sense as a value metric. Usage-based models feel like the natural answer for builders—but buyers are pushing back. They don’t want to pay for activity. They want to pay for results. Outcome-based pricing is the logical destination, but nobody has cleanly cracked what that contract looks like at scale. 

The second is onboarding. Traditional SaaS sold you a product and then spent six months 

helping you configure it—custom workflows, lengthy onboardings, dedicated implementation teams. That was the cost of entry, and customers accepted it. Today, leaner teams armed with coding agents, FDEs and AI-native tooling can stand up the AI ecosystem in days. More importantly, buyers now expect agents to be autonomous—they want business-context-aware, personalised, ready out of the box AI systems. The old onboarding playbook is a liability. The third is the growth model itself. SaaS scaled horizontally—more seats, more departments, more geographies. AI scales vertically, with one agent owning an entire workflow end-to-end. Land and expand still applies, but expansion no longer means more users. It means more outcomes, more workflows, more of your business handing over execution to the agent. NRR will be driven by depth of integration, not breadth of adoption—and that changes everything about how you structure a sales motion, a CS team, and a renewal conversation. 

That changes the GTM motion fundamentally. SaaS sold to department heads who needed tools for their teams. AI sells to founders and operators who want to scale output without scaling headcount. The champion isn’t the VP with 200 reps—it’s the person trying to do the work of 200 without hiring them. 

What Might the New Playbook Look Like? 

The founders who win the next decade will be the ones who have the courage to throw out the SaaS playbook and build a whole new one. But what does that actually look like in practice? 

If outcomes replace seats as the unit of value, how do you price a contract when the outcome is hard to define upfront? If time-to-value and friction-to-switch collapse, what’s your moat—and how long before a better-contexted competitor displaces you just as quickly as you displaced the incumbent? 

The distribution question is equally unsettling. If you’re no longer selling to large teams, how do you build a sales motion around a buyer who measures success in output, not adoption? 

The companies that figure this out won’t just build great products. They’ll invent a new language for what a great business looks like. 

(The author is Partner, Stellaris Venture Partners. Views are personal.)