For twenty years, SaaS ran on a beautifully simple deal. A human logs in, the software helps that human work, and the vendor charges per human, per month.

Now software is starting to do the work itself. AI agents answer tickets, draft contracts, reconcile accounts and write code, and none of them hold a login or occupy a seat. If your revenue depends on counting humans, that fact deserves your full attention.

So is the SaaS business model breaking? My honest answer: it is bending in specific places, and the bends are getting sharper. The companies that understand exactly which assumptions are cracking will adapt. The ones that debate “SaaS is dead” versus “SaaS is fine” will miss the real story.

I have spent more than twenty years building and leading technology teams, and I have watched several platform shifts rewrite what customers expect and what vendors can charge. This one feels different in one respect: it touches the unit of value itself, not just the delivery method.

This workplace post separates what AI is genuinely breaking from what it leaves intact, and it ends with a practical playbook for leaders.

Let’s start by naming the assumptions SaaS was built on.

The Five Assumptions Behind the Classic SaaS Model

Every business model rests on a few beliefs that nobody questions until they fail. Classic SaaS rests on five:

Notice that none of these assumptions fails overnight. Each one erodes at its own pace, which is why the change feels slow and why it is easy to underestimate.

Let’s look at the first crack, because it shows up on the invoice.

Crack One: Seats Stop Measuring Value

When an AI agent resolves a customer ticket, it does not need a license, yet it creates value. When a customer uses agents to do the work of five people, they may cut five seats, and the vendor’s revenue falls while the customer’s results improve. Analysts call this effect seat compression.

The market has noticed. Industry analysts, project that a significant share of enterprise SaaS spend will move toward usage-, agent- or outcome-based pricing by 2030. Several vendors already charge per resolved conversation or per completed task, and customer-support platforms led the way with per-resolution pricing for AI agents.

Yet the picture is messier than the headlines. A widely cited April 2026 survey of SaaS CEOs, found that most planned to retire seat-based pricing within two years, while nearly as many said seat-based pricing still matched their product’s value today. That contradiction explains where pricing is actually landing: hybrid models, which combine a base subscription with credits, usage or outcome-based charges.

For many SaaS providers, an abrupt jump to usage- or outcome-based pricing carries more risk than staying with current models, and most companies lack the operational capacity to run those models well. Their advice, in essence, is to build the capability to experiment before the market forces your hand.

Pricing is only the first crack. The second one hides in your margins.

Crack Two: Software Is No Longer “Free” to Serve

Traditional SaaS earned its famous margins because the cost of serving another customer was tiny. You wrote the software once and sold it many times.

AI features break that pattern. Every model call, every agent action and every generated document consumes compute, and that compute costs real money each time. A power user who runs heavy AI workflows can cost far more to serve than a casual user paying the same flat fee.

This reality explains why usage-based pricing fits AI products so naturally. Large majority of AI-native SaaS companies already offer usage-based pricing, largely because their underlying costs, such as token consumption, map directly to consumption-based billing.

For incumbent vendors, the lesson is uncomfortable but clear. If you add AI to a flat-fee product without rethinking pricing, your best users may become your least profitable ones. Gross margin, the metric investors prize most, becomes something you must manage actively.

Margins and pricing are visible problems. The third crack is quieter and may matter most.

Crack Three: The Moat Moves

For years, a SaaS product’s code and feature set offered a degree of protection. Building a comparable product took teams, time and money, and customers hesitated to rebuild what they could rent.

AI-assisted development narrows that gap. A small team, or even a single skilled person, can now build a credible version of many software categories far faster than before. Commentators gave the market jitters of early 2026 a dramatic nickname, the “SaaSpocalypse,” and while the panic has cooled, the underlying question remains valid: if features are cheap to copy, what keeps customers?

The answer is that defensibility shifts away from features and toward assets that code alone cannot replicate:

• Proprietary data and the context built on it. Years of customer data make your product smarter and harder to leave.
• Workflow depth and integrations. A product woven into dozens of daily processes costs real effort to replace.
• Trust, security and compliance. Regulated buyers pay for proven reliability, audit trails and certifications.
• Distribution and brand. Customers who already know and trust you are easier to keep and expand.
• Outcomes you can guarantee. Vendors who deliver measurable results compete on proof, not features.

Features still matter, but they no longer form a wall. Relationships, data and results form the wall now.

Before you panic, though, let’s be fair about what AI has not broken.

What Is Not Breaking

The loudest predictions often overlook real strengths that SaaS still holds:

• Systems of record remain sticky. Customer data, financial records and employee information live somewhere, and moving them is painful and risky.
• Enterprises buy reliability. Large organizations value uptime, security reviews, support and accountability, and agents need trustworthy systems to act on.
• Switching costs are real. Training, integrations and habits keep customers in place even when alternatives look cheaper.
• AI needs a home. Agents need data, permissions and workflows to operate within, and well-run SaaS platforms provide exactly that.
• The model has already evolved before. SaaS survived the shift from licenses to subscriptions, and many vendors have added usage components already.

So the evidence points to transformation, not extinction currently. The vendors most at risk are not “SaaS companies” in general. They are the ones with thin features, weak data, seat-only pricing and no AI strategy.

That leads to the question every leader should ask next: what do I do about it?

A Playbook for SaaS Leaders

Use these steps to adapt without overreacting:

1. Audit your seat dependence. Calculate how much revenue depends on seat counts, and which customers could cut seats if they adopt agents.
2. Map your AI cost structure. Know the cost of each AI feature per customer, and identify where flat pricing exposes your margins.
3. Pilot hybrid pricing. Combine a platform fee with credits or usage, and test it with a small group of customers before you roll it out widely.
4. Build the billing foundation. Usage and outcome pricing demand accurate metering, clear invoices and strong reconciliation, so invest in that architecture early.
5. Define outcomes carefully. If you move toward outcome-based pricing, agree with customers on what counts as a success and how you measure it.
6. Deepen your moat. Invest in data, integrations, security and workflow depth rather than racing to add more features.
7. Make your product agent-friendly. Offer strong APIs, permissions and audit logs so customers’ agents can use your platform safely.
8. Talk to customers openly. Explain pricing changes in terms of shared value, and avoid surprise bills.

Start with steps one and two this quarter. Those two audits cost little and reveal exactly how exposed you are.

My Advice: The traditional SaaS business model is not collapsing, but it is slowly breaking in the places where it assumed humans do the work, software costs almost nothing to run and code is hard to copy. Those assumptions are weakening, and the pricing, margin and moat questions they raise are real.

The best response is neither denial nor panic. Understand which assumptions your business depends on, test new pricing in small steps, strengthen the assets that AI cannot copy and build the operational muscle for a more variable, outcome-focused world.

SaaS began as a bet that customers would rather rent software than own it. The next chapter bets that customers will pay for results. Leaders who prepare for that shift now will shape it, and those who wait will adapt to it later, on someone else’s terms.

By Dushyant Gadewal #AskDushyant
Technology Leader | Builder | Emerging-Tech Explorer | 20+ Years in Technology

Note: The names and information mentioned are based on my personal experience and publicly available data; however, they do not represent any formal statement.

Tags: #SaaS, #FutureOfSaaS, #AI, #AgenticAI, #SaaSPricing, #TechLeadership, #ProductStrategy, #BusinessModel, #Startups, #EnterpriseSoftware, #PricingStrategy, #AskDushyant #TechConcept #TechAdvice