AI Exposes SaaS's Flawed Growth Model. Investors Take Note.

The public markets have delivered a brutal verdict on enterprise software: the business model that created fortunes over two decades may have been fundamentally flawed even before AI agents emerged to accelerate its demise. This isn't merely a market correction—it's an existential reckoning that exposes how much of SaaS growth relied on contractual tricks rather than genuine value creation.

What should terrify software executives isn't just AI replacing their products, but what the panic reveals about how growth actually worked. A significant chunk of that celebrated 20% annual expansion came from automatic price escalators hidden in contracts, not from solving harder problems or winning new business.

This contractual sleight of hand allowed companies to report steady growth without delivering proportional new value. Software firms banked on those escalators whilst hunting for additional revenue through seat expansion, new modules, or fresh customers. When those engines fired simultaneously, results looked spectacular.

Enjoying this article?

Get stories like this in your inbox every week.

Join the Community

The problem? Several of those engines had already started sputtering well before ChatGPT became ubiquitous.

The Cracks Were Showing Before the AI Earthquake

SaaS growth rates dropped from 20% to 12% annually, and much of that decline predated the current AI frenzy. According to analysis from enterprise software investors, Fortune 500 headcount has grown by less than 1% over recent years—a statistic that matters enormously when your pricing model depends on adding seats.

The seat-based licensing model that built the SaaS empire assumed companies would keep hiring, and each new employee would need their own login to Slack, Salesforce, or any of the dozens of tools considered essential to modern work. Peak penetration meant the easy growth was over.

You couldn't sell more seats to a sales team that already had the software on every laptop.

Expanding into adjacent products or new buyer personas worked for elite operators, but plenty of mid-tier software firms found themselves running out of road. Enterprise software still represents only 1-2% of revenue at Fortune 500 companies, which suggests there's theoretically room to grow.

Yet that number has remained stubbornly flat even as software became 'essential' to every business function. The optimistic reading is that there's headroom for expansion; the pessimistic one is that software has already captured all the value enterprises are willing to allocate to it.

AI Threatens the Unit of Value, Not Just the Product

Generative AI doesn't just compete with existing software products—it fundamentally questions what software should charge for. If an AI agent can handle customer service enquiries that previously required three support staff using a help desk platform, does that company pay for three seats, one seat, or some entirely different pricing scheme based on tickets resolved?

The companies showing early success with AI products are discovering this tension firsthand. Datadog reports that 12% of its revenue now comes from AI-native customers, growing 17% quarter-on-quarter from 6% eighteen months ago. ServiceNow's AI product, Now Assist, has reached $600m in annual recurring revenue.

Those numbers sound impressive until you realise they represent a fraction of these companies' total revenue, and enterprise budgets haven't expanded to accommodate both legacy SaaS and new AI spending simultaneously. CFOs are doing the maths.

Why commit to a three-year contract with annual escalators for software that might be obsolete when AI agents can do the same work without per-seat licensing?

That hesitation shows up in delayed renewals, reduced seat expansion, and consolidation of 'nice-to-have' tools. The immediate threat isn't that customers are ripping out their existing software stack—replacing entrenched enterprise tools remains nightmarishly complex—but that the next dollar of spending is going elsewhere.

Where AI Will Strike Hardest

What's interesting here is where AI will likely do the most damage. The real disruption may not come in categories where SaaS already dominates, but in areas where software never fully solved the problem.

Coding is the clearest example: AI assistants are already changing how developers write and review code, filling a need that integrated development environments only partially addressed. Customer service represents another gap, where AI agents can now handle interactions that required both software and humans.

These aren't cases of AI replacing good software; they're instances of AI finally solving problems that SaaS merely papered over.

The Mainframe Analogy Cuts Both Ways

History offers both comfort and caution. The transition from mainframe computing to SaaS took 25 years and continues today—IBM's mainframe business still generates billions annually. That glacial pace suggests incumbent software companies have time to adapt, provided they move decisively rather than drift.

Technical debt accumulates during periods of hesitation. Enterprises currently deferring upgrades and delaying migrations aren't making those problems disappear; they're compounding them. Eventually, those chickens come home to roost, potentially creating a wave of forced modernisation that savvy software companies can capture.

The winners will be those that figure out how to tie their products to measurable value in a world where headcount no longer correlates with software spend. They'll need to hunt aggressively for AI-native customers, because the Fortune 500 roster could look dramatically different in five years.

They'll need to answer honestly whether the problem they solve will still exist when work becomes more autonomous. The SaaS model isn't dead, but its foundations have shifted.

The contractual escalators and seat-based expansion that built a trillion-dollar industry won't carry companies through the next decade. Some will adapt, finding new units of value to monetise and new ways to demonstrate ROI in an AI-augmented world. Others will discover they were selling a vitamin when customers now demand surgery.

The 70% valuation haircut suggests investors are still figuring out which is which. As AI becomes a direct substitute for software rather than just a productivity layer, agentic AI's ability to automate tasks and replicate workflows will force SaaS leaders to fundamentally rethink their approach, and predictions of SaaS's collapse may prove prescient for those who fail to adapt.

More articles by Ross Williams