Enterprise software has been built around a simple idea for years: more users, more licenses, more revenue. Each department has its tools, each tool its interface, and each interface its way of capturing time, data, and budget. CRM, ERP, human resources, support, analytics, documentation, sales, purchasing, and finance have all developed much of their value around screens, bespoke flows, and additional modules.
Agentic AI is beginning to challenge that logic. According to Gartner, up to $234 billion in enterprise application spending could be exposed to the impact of AI agents by 2030. This figure represents about 20% of SaaS enterprise software expenditure, says the consultancy. It doesn’t mean companies will stop using applications, but they may change how they interact with them.
The shift can be summarized with a concept Gartner calls agentic arbitrage. It occurs when an AI agent completes tasks across multiple systems, reducing the need for the user to open each application and shifting value from the interface to the outcome. Instead of navigating five tools, filling out fields, copying information, and reviewing dashboards, the user commands an action, and the agent coordinates the process.
SaaS isn’t dying, but it’s losing its prominence in the interface
For years, the interface has been a central battleground among SaaS providers. A clearer, faster, or more convenient application could justify a migration. Vendors added dashboards, automation, alerts, assistants, and new views to keep users engaged within their products.
With agentic AI, this relationship could evolve. If an agent operates over CRM, ERP, document management, and email without the user entering each system directly, the application remains necessary but becomes less visible. The system retains data, rules, permissions, and traceability, but daily experience moves to a higher layer.
Gartner frames this as a metamorphosis of SaaS, not its disappearance. The term “Saaspocalypse” may sound exaggerated, but it describes the pressure many business models—based on seats, active users, and interface adoption—will feel. Software isn’t destroyed; it’s reorganized.
| Traditional SaaS | SaaS with agentic AI |
|---|---|
| The user works within each application | The agent works across applications |
| Value is centered on screens and features | Value is measured by completed tasks |
| Revenue grows with users and licenses | Revenue can be linked to usage, outcomes, or flows |
| Each provider protects its interface | The agentic layer coordinates multiple systems |
| The buyer evaluates modules and dashboards | The buyer demands return and efficiency |
The difference may seem subtle but affects the core of the business. If a company no longer needs all users to access a tool for work to progress, charging per user loses some of its strength. If an agent executes, summarizes, updates, compares, and prepares decisions, the interface ceases to be the main source of perceived value.
From software as a tool to software as infrastructure
Gartner’s thesis reflects a broader transition. Enterprise software is shifting from a place where employees work to an infrastructure on which agents act. That infrastructure will remain important: it must expose APIs, permissions, events, histories, business rules, and reliable data. But the end user may relate less and less directly to the original application.
A simple example is sales. Today, a salesperson can review CRM, check email, open a meeting tool, search documents, update an opportunity, and prepare a proposal. An agent could handle much of this: reading call transcriptions, detecting next steps, updating fields, drafting an email, searching for marketing material, and leaving a draft proposal for review.
In support, an agent can analyze tickets, consult knowledge bases, review customer history, draft responses, and escalate only the cases needing human intervention. In finance, it can cross-check invoices, purchase orders, and contractual conditions. In HR, it can handle internal queries, prepare documentation, and coordinate approvals.
In all these cases, the applications still exist. What changes is who manages them. The user stops jumping between interfaces, and the agent becomes the operational layer.
SaaS providers will need to sell outcomes
Gartner’s forecast points directly at software vendors. Adding a co-pilot in a corner of the screen may fall short if the product doesn’t enable full process execution. Agentic AI isn’t limited to answering questions about app data; it must act where work happens.
This will compel providers to move from interface-based value to results-based value. Customers won’t want to pay more just because an application has “AI.” They will want to know how many hours it saves, errors it reduces, processes it accelerates, costs it avoids, or revenues it helps generate.
Institutional memory will become a crucial part of this shift. A useful agent doesn’t just need data access; it must understand how an organization works, common exceptions, the tone used with clients, internal rules, necessary approvals, and prior decisions. That contextual layer will be far harder to replicate than isolated functionalities.
For large providers, the challenge is twofold. They must protect their position in the value chain and simultaneously open their systems for agents to work with them. If they close off too much, clients might seek more interoperable alternatives. If they open without a strategy, another agentic layer could dominate the user relationship.
New entrants targeting legacy systems
The change not only threatens incumbents but also creates space for new players. A startup doesn’t need to build a full ERP to compete in certain processes. It can create an agentic layer operating over existing systems, solving a specific task more efficiently.
Such a company can deliver results without replacing the entire tech stack. It can connect to already deployed software, automate a cross-system flow, and capture budgets previously allocated to additional licenses, manual integrations, or professional services.
Integrators and consultancies also have an opportunity. Many organizations won’t be able to simply activate agents and expect results. They’ll need to redesign processes, review permissions, clean data, define metrics, implement human controls, and adapt architecture. Agentic AI doesn’t work well on messy processes; it only accelerates disorder.
That’s why Gartner highlights an opportunity for service providers and platforms capable of developing cross-domain agentic flows. The value will be not just in technology but in understanding the full process and making it measurable.
The battle will be over the execution layer
The tech market is entering a phase where controlling the execution layer might become more important than managing a specific application. Who manages user requests, decides which systems to consult, executes actions, and presents the final result will hold a strategic advantage.
This layer could reside within a traditional SaaS provider, a horizontal agent platform, an integrator, a productivity suite, or an AI-native tool. Competition will be fierce because the prize is high: becoming the central point orchestrating enterprise work.
For clients, this shift offers benefits and risks. The advantages are fewer context switches, less manual work, more automation, and a more direct link between software and outcomes. Risks include security issues, traceability gaps, dependency on a new layer, execution errors, over-permissions, and loss of control if the agent acts unsupervised.
Companies will need to demand operational explainability: what the agent has done, what data it accessed, what changes were made, with what permissions, and which parts require human approval. Without this traceability, agentic AI could become a black box over critical systems.
Implications for CIOs and procurement
For CIOs, the key message isn’t to abandon current SaaS but to rethink how it will be consumed in the coming years. The question shifts from which tool has the best features to which integrates best into agentic flows, exposes robust APIs, respects granular permissions, and allows automation without losing control.
Negotiation models will also change. As value shifts from users to outcomes, pricing will need to adapt. Expect more discussions around consumption, tasks performed, savings achieved, automation scope, process volume, and measurable impact. Per-seat pricing will persist but coexist with more usage- and value-based models.
Business units will demand fewer dashboards and more operational responses. It won’t suffice to simply “see” an incident, opportunity, or budget deviation—the system will need to propose, execute, or prepare the next action.
Gartner’s estimate serves as a warning for the industry: the $234 billion at risk doesn’t predict immediate destruction but signals a reallocation of value. Money will continue flowing into enterprise software but might shift toward those able to enable AI to perform real work across systems.
SaaS will still be there. The question is whether it remains the primary interface for work or becomes the silent foundation upon which intelligent agents operate.
Frequently Asked Questions
What is agentic AI in enterprise software?
It’s the use of AI agents capable of planning and executing tasks across multiple enterprise systems, not just answering questions within an application.
What does “agentic arbitrage” mean?
It’s the value shift that occurs when an agent completes tasks across several applications, reducing the need to interact with their traditional interfaces.
Will SaaS disappear?
No. Gartner refers to a market transformation, not disappearance. Applications will remain necessary but may become less visible to users.
Why is the per-user model at risk?
Because if agents do tasks previously performed by multiple employees across different tools, the active user count no longer captures all the value being created.
What should SaaS providers do?
Integrate agentic capabilities at the execution point, capture specific customer context, open up workflows between systems, and demonstrate measurable results.
via: Gartner
