The Golden Age Is Over
For ten years, the SaaS playbook was unbeatable: sell software as a monthly subscription, charge per user, hide features behind paywalls and optimise for net revenue retention. Investors loved the predictability. Founders loved the margins. Customers had no alternative.
Then AI arrived – and with it the question every SaaS company must now answer: what happens when an LLM does what customers have been paying £49 per seat per month for?
Why Seat-Based Pricing Is Dying
The seat-based model rests on one assumption: more users = more value. But AI agents don't need a seat. A single employee with a well-configured AI assistant now does the work that previously required a team of five.
The consequence: companies need fewer seats. Fewer seats mean less revenue – with the same or even greater customer value. That's a fundamental contradiction in the SaaS model.
We're already seeing this in practice. Companies are radically consolidating their tool landscape. Instead of seven specialised SaaS tools, they use a single AI assistant that orchestrates between APIs.
Entire Product Categories Are Disappearing
The first wave hits the most obvious candidates: text generators, simple analytics tools, template builders, formatting helpers. Products whose core function a prompt can replace.
The second wave is more dangerous. It hits SaaS products marketed as "workflow tools" but that actually just move data from A to B. Automation platforms, reporting dashboards, simple CRMs – anything an agent with API access can handle on its own.
The third wave will hit platforms that derive their value from network effects. When AI agents can independently aggregate information, platforms built on user-generated content lose their moat.
What Remains: Data and Infrastructure
Not every SaaS company is under threat. The survivors share one thing: they own something an LLM cannot replicate.
Proprietary Data – Companies that have built unique datasets unknown to any language model from training have a real advantage. This applies to industry data, real-time market data or specialised knowledge bases.
Infrastructure – Tools deeply integrated into technical infrastructure (monitoring, security, DevOps) can't easily be replaced by a chat interface. They need persistent agents, integrations and reliability.
Regulated Industries – Compliance, auditability and certifications are requirements an LLM alone cannot fulfil. SaaS products in regulated areas like healthcare, finance or law have more breathing room.
The Shift to Outcome-Based Pricing
The smart SaaS companies are already pivoting: away from "per seat, per month" towards "per outcome". Instead of paying for access to the software, customers pay for what the software achieves.
A recruiting tool no longer charges per recruiter seat but per successfully filled position. A marketing tool charges not per user but per generated lead. That's fair – and it's the only model robust against AI disruption.
What This Means for Founders
Anyone founding a SaaS company today must answer the AI question from day one: what can my product do that an AI agent with API access cannot? If the answer is "nothing", the product is a feature, not a company.
The winners of the next decade won't be those who bolt AI onto their existing SaaS product as a feature. They'll be those who rethink their entire value creation around AI – and in doing so find pricing models that reflect actual customer value.
Conclusion
The AI threat to SaaS is real, but it's not evenly distributed. Products that rely on convenience and interface are most at risk. Products built on proprietary data, deep integration and regulatory complexity have a future. The key is to honestly evaluate which category your product falls into – and then act quickly.