By Mahroof K., Entrepreneur-turned Program & Product Leader with 12+ years of experience building and scaling AI, SaaS, web, and mobile products. Former Founder & CEO of Cedex Technologies (acquired).
There is a question circulating in every tech boardroom, investor call, and founder group right now. It goes something like this: is software as a service finally over?
When discussing the future of SaaS, stock charts seem to say yes. Sentiment seems to say yes. The sheer weight of the news, AI agents doing what used to take entire teams, enterprises slashing licence counts, point solutions losing relevance almost overnight, all of it seems to be screaming yes.
But that framing is wrong. Or rather, it is asking the wrong question entirely.
SaaS is not dying. It is being rewritten from the inside out. And the difference between those two things, dying versus rewriting, determines everything: which companies survive, which ones get acquired at a steep discount, and which ones quietly become the infrastructure layer of the next era of enterprise software.
Let us break it down properly.
The Traditional SaaS Pricing Model That AI Broke
To understand what is happening, you have to understand what made SaaS so extraordinarily durable for two decades.
The model was elegant in its simplicity. More employees = more licences = more revenue. Every time a company hired, expanded into a new market, or onboarded a new department, the SaaS vendor got paid. It was recurring, predictable, and almost frictionless to scale. Investors loved it. It is why SaaS companies traded at 18–19x revenue multiples during peak years. The math just worked.
Then AI walked in and snapped that equation in half.
When AI agents can perform the cognitive work of multiple employees, drafting, analysing, routing, deciding, following up, the logic of paying per human seat no longer holds. Fewer people doing more work means fewer licences. And fewer licences means the entire revenue architecture of the SaaS industry is under structural threat. Not from a better competitor. From a fundamental shift in how work gets done.
The market has been pricing this in aggressively. A Morgan Stanley basket of SaaS stocks fell 15% in mid-January 2026 alone, following an 11% drop across all of 2025. Between January 15 and February 14, 2026, the global software market lost nearly $2 trillion in market capitalisation in just 30 days. Look at the damage across major players:
- HubSpot: Down over 50% from its peak
- ServiceNow: Down close to 37% year-to-date, despite beating earnings for nine consecutive quarters
- Atlassian: Dropped 30%
- Salesforce: Lost 26% of its market cap
- Adobe: Fell roughly 19%
This is not a market overreaction driven by panic. It is a market trying, imperfectly, to price in a structural shift it does not yet fully understand.
The Paradox of the Future of SaaS: Crashing Stocks, Growing Revenues
Here is where things get genuinely interesting, and why the "SaaS is dead" narrative is too simple to be useful.
The global SaaS market is projected to grow from $266 billion in 2024 to $375 billion by 2026, on a trajectory toward over a trillion dollars by 2032. Organisations now spend an average of $55.7 million on SaaS annually, up 8% year-over-year. Enterprise software budgets are not shrinking in absolute terms. They are shifting.
How do you reconcile crashing stocks with revenues that are still growing? The answer is that markets are not pricing today's revenue. They are pricing the future business model, and that future is genuinely uncertain.
Consider Salesforce. Mark Benioff is publicly calm. "People think we have our backs against the wall when the opportunity has never been greater," he said recently. The market pushed back hard, sending the stock down nearly 28% this year. So who is right?
Probably both of them, and that tension is exactly what makes this moment so interesting. The growth is real. The threat is real. And they are happening at the same time, which is why this moment is so disorienting for everyone inside it. The SaaS industry is being asked to run two operating systems simultaneously: the one that pays the bills today, and the one it needs to build to survive tomorrow.
From Human Interfaces to AI Agents: What Actually Changed
For 25 years, enterprise software was built around one core assumption. A human would sit down, log in, navigate a UI, and do something. Every feature, every dashboard, every workflow was built to serve that human operator.
That assumption is now obsolete.
Bain and Company put it plainly in their 2025 Technology Report: in three years, any routine rules-based digital task could move from "human plus app" to "AI agent plus API." The interface becomes optional. The human steps back. The agent steps in.
Salesforce made this concrete at TDX 2026 with the launch of Headless 360. The message was striking in its clarity: you do not need to log into Salesforce anymore. Agents do. Every workflow, every dataset, every process is now accessible via API, without a UI. Salesforce essentially announced that the dashboard is no longer the product. The system underneath it is.
That is not a feature update. It is an architectural admission, and it is one of the most honest signals any major SaaS company has sent about where this is all heading.
This is not a small design change. It is an architectural inversion that the entire industry is being forced to confront. Software was built to serve human attention. Now it needs to serve machine execution. Those are fundamentally different products, and most SaaS companies were not built for the second one.
The numbers reflect how fast this is moving:
- By 2026, IDC expects AI copilots to be embedded in nearly 80% of enterprise workplace applications
- The AI agent market is growing at a projected 46% annually, expanding from $7.84 billion in 2025 to over $52 billion by 2030
- Gartner predicts 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% just a few years ago
When agents are the primary users, you do not need a beautiful UI. You need clean APIs, reliable data models, and agentic orchestration layers. The companies that have those things are in a completely different position from those that have spent the last decade perfecting dashboards.
Which Enterprise Software Survives? AI Agents vs. Point Solutions
This is the part most "SaaS apocalypse" coverage misses entirely. The disruption is not uniform, and treating it as uniform leads to genuinely bad decisions, both for investors and for the people building and buying software.
Here is a useful framework for thinking about who is actually at risk.
Point solutions are in real trouble.
Tools that do one specific thing, like automating a particular report, scheduling a specific kind of meeting, or managing a particular support queue, are the most exposed. Gartner expects 35% of point-product SaaS tools to be replaced by AI agents or absorbed within larger ecosystems by 2030. When a general-purpose AI agent can perform the same task without a dedicated subscription, the standalone business case collapses. Publicis Sapient has reportedly reduced its traditional SaaS licences by approximately 50% by substituting them with AI tools. Thoughtworks eliminated three narrow SaaS platforms in 2025 alone, replacing them with bespoke AI workflows.
Horizontal platforms with deep data moats are a different story.
As Mary Meeker's Bond Capital report highlighted, the era of the SaaS point solution is ending, but horizontal platforms will dominate, assuming they successfully become AI-native rather than just AI-featured. These platforms are harder to displace because their value is not the software itself. It is the years of proprietary workflow data, integrations, and institutional context built on top of it.
This is actually where Benioff has a point, even if the market is not giving him credit for it yet. You can vibe-code a front end. You can prompt your way to a prototype. But you cannot easily replicate 20-plus years of enterprise compliance layers, security frameworks, and deep workflow integrations that Fortune 500 legal teams have already signed off on. That moat is real. The question is how long it holds as AI continues to mature.
Systems of record are safer than they have ever been.
ERPs, compliance platforms, financial ledgers, healthcare records management. When enterprises deploy AI agents today, they are not replacing these systems. They are building orchestration layers on top of them. A financial services CIO put it plainly: an LLM that gets the right answer six out of ten times is useless for underwriting. Mission-critical processes demand 100% consistency. AI agents are excellent at reasoning and pattern recognition. They are not yet a substitute for deterministic systems where errors carry legal and financial consequences.
This is the survival filter that matters most right now. Companies that own the deterministic core of enterprise operations are not going anywhere. Companies living in the probabilistic, task-automation layer are the ones facing an existential question.
The Shift Toward Usage-Based and Agent-Based Pricing
Even the companies that survive the product disruption face a serious pricing problem.
The per-seat model worked when humans were the users. In an agent-driven world, there are no seats. There are actions, API calls, tokens, and outcomes. The entire commercial architecture of the SaaS industry needs rebuilding, and that is genuinely hard to do while simultaneously running a public company and managing investor expectations.
Salesforce is already feeling this tension directly. The company still generates the vast majority of its revenue from per-seat pricing, yet it is actively shifting toward consumption-based pricing for AgentForce, charging per action, per agent, per outcome. That gap between today's business model and tomorrow's is precisely what the market is pricing in.
Gartner projects that by 2030, at least 40% of enterprise SaaS spend will shift toward usage-based, agent-based, or outcome-based pricing. Customers will pay for what agents actually do, per action, per task completed, per outcome delivered, rather than per licence held.
Early movers are already showing the way:
- Zendesk has incorporated outcome-based pricing into its AI agents, tying costs directly to measurable results rather than seat counts
- Salesforce is shifting toward consumption-based pricing for AgentForce, with 23,000 customers already using it
- By 2022, 61% of SaaS companies had already adopted some form of usage-based model, and that proportion is accelerating fast under competitive pressure
The transition is messier than it sounds. Outcome-based pricing requires defining what an outcome is, measuring it reliably, attributing it correctly, and billing for it in a way customers actually trust. That is a product, operational, and legal challenge all at once, with no established playbook. Deloitte is honest about this: it could take years for standard practices to emerge, if they ever do.
In the meantime, many vendors are taking the path of least resistance by layering AI tiers onto existing subscriptions and charging premiums that inflate costs without adding proportional value. Zylo's 2026 SaaS Management Index found that even well-governed enterprise portfolios are absorbing unplanned mid-contract cost increases driven by AI pricing changes. That creates friction, which creates churn risk, and that is exactly what SaaS companies cannot afford right now.
The AI Infrastructure Split: Who the Market Is Actually Rewarding
While traditional SaaS is being repriced downward, a different set of software companies is having a very different 2026:
- Palantir: Up 142% over the past year
- Cloudflare: Up 80%
- MongoDB: Grew 70%
- Snowflake: Re-accelerated to 29–32% year-over-year growth
- CrowdStrike: Up over 50%
These are not outliers. They are proof of concept for what the market actually rewards right now: AI-native infrastructure, data platforms positioned for agentic workloads, and security tools that become more essential as AI expands the attack surface.
The lesson is stark. Do not build SaaS and add AI. Build AI and add SaaS economics. The companies winning are the ones that already had that proof before the market started demanding it.
Goldman Sachs CEO David Solomon called the broader software selloff "too broad." Bank of America's senior analyst Vivek Arya argued the market is simultaneously pricing in two mutually exclusive scenarios: that AI capital expenditure will deteriorate due to weak ROI, while also being so powerful it renders all software obsolete. Those two things cannot both be true. Quality platforms being repriced alongside obsolete point solutions may represent a genuine long-term opportunity for those with the patience to wait for the rewrite to complete.
Winning the AI Era: From Software-as-Interface to Software-as-Infrastructure
The companies coming out of this ahead are not the ones adding AI features to existing products. They are rethinking what the product fundamentally is.
The shift is from software-as-interface to software-as-infrastructure. From "log in and do something" to "connect and let agents run." From selling access to a UI to owning the data layer, the workflow layer, and the agent orchestration layer that everything else depends on.
Salesforce's move at TDX 2026 is a clear illustration of this playbook in action. They are now operating across three distinct layers:
- Infrastructure: Headless 360, APIs, MCP tools, CLI access. Agents can run the entire system without a human ever opening a browser tab
- Intelligence: Investments in Anthropic, OpenAI integration, and multi-model support across Claude, GPT, and others
- Application: AgentForce, Slack as the primary interface, Agent Exchange marketplace
What Salesforce is doing, quietly and without making too much noise about it, is dismantling its own product from the inside. They are betting that in five years, nobody will be logging into a CRM. Agents will be running it on behalf of humans. The product that survives is not the one with the best UI. It is the one with the deepest data, the most trusted workflows, and the most reliable system underneath.
ServiceNow's $2.85 billion acquisition of Moveworks tells the same story. Enterprise SaaS M&A activity in Q4 2025 hit $83.7 billion across 245 deals, much of it driven by platform companies acquiring AI capabilities before the market forces their hand at worse valuations. The incumbents who understand what is happening are not defending their current product. They are rebuilding something fundamentally different underneath it.
The Real Question for the Future of Enterprise Software
Everyone keeps asking: is SaaS dead?
That question misses the point entirely.
The real question is who owns the enterprise AI layer once SaaS as we knew it disappears. That is the prize worth competing for. Not whether subscription software survives, but who ends up owning the data, controlling the workflows, and powering the agents that run the modern enterprise. Because whoever controls that layer controls everything downstream.
Salesforce built the CRM category. Now it is trying to rebuild it for a world where humans are no longer the primary users. Benioff is not in denial. He is just not saying it out loud yet. The Headless 360 launch, the AgentForce rollout, the multi-model AI investments, these are not product updates. They are a quiet acknowledgement that the SaaS era is ending and something else is beginning.
The SaaS companies that survive this shift will not be the ones that fought hardest to preserve the per-seat model. They will be the ones that own data, control workflows, and power the agents that replaced the humans who used to click through their interfaces.
The rewrite is not coming. It has already started. The question is whether you are reading the new draft, or still trying to protect the old one.
If you are working through any of these questions in your own organisation, whether to consolidate your SaaS stack, evaluate AI-native alternatives, or think through the shift to agentic workflows, let's talk.