For the past year, one idea has dominated conversations about enterprise technology: AI agents will replace Software as a Service (SaaS). It’s a compelling case on paper. If AI can reason across tools, write code, execute workflows, and interact with systems through APIs, then traditional SaaS applications start to look like unnecessary middlemen.
Seats become less relevant. User interfaces matter less. Software becomes cheaper to build. Custom internal tools become easier to create.
Richard Farrell is CIO at Netcall. Suraj Kika is Founder of Jadu and Chief Growth Officer at Netcall.
In that reality, much of SaaS gets pushed down into infrastructure, while value shifts to models, agents, and orchestration layers. This reaction is being driven by several parallel shifts.
Software is becoming cheaper and faster to build, AI agents are becoming better at navigating tools and executing work across systems, and the economics of agentic execution are bringing new attention to cost and latency, or time to serve.
The market is not wrong to think this way. AI agents will change software economics.
They will push platforms to become API-first, deeply connected, and capable of supporting autonomous activity at scale. They will challenge seat-based pricing and accelerate the move towards consumption and outcome-based models. They will expose weak products and reward systems that are trusted, extensible, and embedded in real operations.
This is the next phase of SaaS, rather than the end of it.
From replacement to expansion
The assumption behind the “end of SaaS” narrative is that software exists primarily as an interface. If agents can bypass that interface, the application becomes redundant. That logic holds for some categories of software, but it does not apply uniformly. Enterprise software has never been just a presentation layer.
Enterprise platforms derive their value from managing structured data, enforcing permissions, executing workflows, and maintaining the audit trails organizations rely on. Agentic AI does not remove that requirement; in many cases, it sharpens it. What AI changes is not whether these capabilities are needed, but which types of software are most exposed as agents become more capable.
Products whose value is largely defined by navigation, basic interaction, or shallow workflows are more vulnerable when agents can reason directly over APIs and complete tasks without human mediation. More configurable, deeply integrated platforms behave differently.
Solutions that act as systems of record, coordinate workflows across services, apply policy, manage state, and provide security, auditability, and evidencing are not easily displaced. Their role is not simply to present functionality, but to ensure work is executed consistently, safely, and at scale.
In that context, AI agents are not replacing platforms so much as being embedded within them, extending how work is initiated, coordinated, and completed. That direction is already visible in the data.
Gartner forecasts that by 2030, 85% of enterprise agentic AI investments will be bundled into existing SaaS and cloud renewals rather than delivered through net-new contracts, up from 55% in 2025.
SaaS as the execution and control layer
As agents become more capable, they shift where value sits and how platforms compete.
For years, software differentiation has been driven by features and user experience. In an agent-driven world, those differences begin to matter less. Agents can move across systems, access functions directly, and orchestrate tasks programmatically.
What matters instead is which platforms can actually complete work. This means coordinating processes across systems, applying the right type of automation at each step, and ensuring outcomes are reliable, traceable, and compliant.
In this model, SaaS does not disappear. It becomes the execution and control layer for enterprise AI. As this shift plays out, value and margin move away from individual features and towards the platforms that control execution and data access. Agents themselves are unlikely to be a sustainable point of differentiation. As capabilities converge, the focus moves towards control.
That’s because enterprise processes still need to be predictable, auditable, and, in many cases, reversible. Regulatory frameworks are reinforcing this, with increasing expectations around explainability and oversight. This places new importance on the platforms that sit at the center of operations.
Organizations that can orchestrate work across systems, apply automation selectively, and produce a clear evidence trail will capture more value over time. Those that cannot risk becoming passive data stores, rather than active systems of execution.
A multiplier effect on demand
There is another misconception shaping the conversation that AI will reduce the amount of work organizations need to do and, with it, the reliance on software.
In practice, the opposite is happening. As the cost and effort required to execute processes falls, more work becomes viable. More cases are identified, more customer needs are addressed, and more processes are triggered automatically.
We are already seeing this in service environments, where AI-driven detection and automation increase the volume of actionable work entering the system.
In the context of public services, for example, within a case management environment, a missed bin, pothole, or housing repair issue can be identified automatically, raised as a case instantly, and routed into the right workflow without waiting for a citizen or staff member to log it manually. In that model, the value sits not in who created the case, but in how effectively the platform absorbs, processes and resolves it.
This creates a multiplier effect. AI does not just reduce effort; it expands what organizations can do. AI expands demand, and SaaS platforms are where that demand is fulfilled.
The question is no longer how many users a platform supports. It is how much work it can handle, complete, and evidence effectively.
What this means for SaaS platforms
This shift will not benefit every platform equally.
As software moves from presenting information to completing work, value becomes tied to what a system can actually deliver, not just how it is accessed. This will also shape commercial models, with more platforms likely to combine traditional subscriptions with pricing based on consumption, throughput, or completed outcomes.
Organizations still need oversight, approvals, and ways to manage exceptions, but the focus is shifting towards how effectively platforms can absorb, process, and complete work at scale.
That change is already exposing the gap between strong and weak software, placing greater scrutiny on systems that offer limited workflow depth or rely heavily on manual effort. In contrast, platforms that are deeply embedded in operations, with strong data, logic, and execution capabilities, will become more valuable.
As agents become consumers of APIs, software needs to be connected and capable of supporting autonomous activity at scale. Those that are not will struggle to keep up.
The next phase of SaaS
Ultimately, what we are seeing is not the end of SaaS, but an evolution. SaaS is being reshaped, not just as software for human interaction, but as software designed for both humans and machines operating across systems.
A shift from systems that present information to systems that complete work. From user-driven interaction to orchestrated execution, and feature competition to control over how processes run.
This is not theoretical. It is already happening. Low-code platforms and productivity tools are being augmented by AI and agentic workflows, changing how applications are built, adapted, and used day to day.
But this shift also raises the bar for how organizations adopt AI. Those that simply layer agents onto existing processes risk creating more complexity rather than less. Introducing automation without understanding how work really flows can make outcomes harder to predict and harder to evidence, particularly in regulated environments.
The organizations that succeed will start with a real operational problem and introduce AI selectively. Where existing workflows provide sufficient structure and control, value can be delivered quickly without redesign. Where AI exposes friction or inefficiency, that insight can then inform targeted process improvement.
AI agents amplify the platforms beneath them; they do not replace them.
The result is not less software, but different software, running on platforms built to execute work, absorb demand, and stay in control. More automation, more cases handled, more throughput, more outcomes achieved through systems that reliably complete more work at scale.
We’ve featured the best vibe coding.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit
