The software industry stands at a major turning point. For years, SaaS, or Software as a Service, followed a simple model. A company sold access to an application, and people used that application through a screen. Employees opened CRM systems, checked dashboards, created tickets, updated records, sent emails and moved data from one tool to another.
AI agents now challenge that model.
An AI agent can understand a goal, choose the required steps, use several software tools and complete a task with limited human help. Instead of a person moving between five applications, an agent can work across those systems and produce the final result.
This change raises a much bigger question than whether AI will add another feature to SaaS. The real question asks whether software will become less visible as agents take over more work.
The answer does not look simple. AI agents may destroy value for some SaaS products while creating far more value for others. The biggest change may not come from the end of SaaS. It may come from the end of the traditional software interface.
$234 Billion of SaaS Spend Faces Risk
Gartner has placed a large number on the potential impact. Its July 2026 analysis estimates that up to $234 billion of enterprise application software spend could face “agentic arbitrage” through 2030. That figure represents about 20% of enterprise SaaS spend.
The idea behind agentic arbitrage is simple. Many business applications depend on people who sit in front of screens and perform routine tasks. An AI agent can handle many of those tasks without the same level of human interaction.
A sales employee may no longer need to open a CRM, search for old leads, check customer activity, prepare a message and create a follow-up task. An agent could handle the full process after a simple request.
That creates a serious problem for software companies whose main value comes from the interface itself.
A business may still need the database, rules, security and workflow system. It may not need every employee to spend hours inside the application.
Salesforce Shows the Other Side of the Story
The current results from Salesforce provide an important counterpoint to the “SaaS is dead” argument.
Salesforce reported Agentforce ARR above $1.5 billion in its Q2 FY27 results. Agentforce ARR rose by more than 240% year over year. Agentforce and Data 360 together reached almost $3.9 billion in ARR.
The scale of agent activity also stands out. Salesforce reported 7 billion Agentic Work Units delivered cumulatively, with 3.2 billion AWUs in Q2 alone. That quarterly figure rose 97% from the previous quarter.
Salesforce also raised its FY27 revenue guidance to $46.1 billion to $46.4 billion, while current remaining performance obligations grew 14% year over year.
These numbers show a different path for SaaS. Salesforce does not need to fight AI agents from outside the software platform. It can place agents inside the platform and sell them as a new layer of value.
The strategy turns the threat into a product.
ServiceNow Shows Similar Strength
ServiceNow offers another strong example.
In Q2 2026, ServiceNow reported $3.877 billion in subscription revenue, with subscription growth of 24.5% year over year. Current remaining performance obligations reached $13.2 billion, up 21% year over year.
ServiceNow also reported that ServiceNow AI crossed $1 billion in annual contract value. Agentic deployments rose 9 times within nine months.
These figures matter for the wider SaaS market. They show that enterprise customers have not simply stopped buying major software platforms after the arrival of AI agents.
Instead, large customers may place more value on platforms that combine data, workflows, permissions, integrations, security and AI execution.
That creates a useful distinction. AI may threaten a weak application while it strengthens a powerful software platform.
The Interface May Lose Its Power
The biggest change may happen at the interface level.
The old software model looks like this: a person opens an application, finds the correct menu, enters information and completes a process.
The new model can look very different. A person gives an agent a goal. The agent finds the required data, uses the correct systems, follows company rules and completes the task.
A sales manager could ask an agent to find enterprise deals without meaningful customer contact for 14 days, identify the reason for the delay, prepare suitable follow-up messages and arrange meetings where appropriate.
The CRM still matters in that example. Customer records still matter. Company rules still matter. Security still matters. The agent simply becomes the main path between the employee and the software.
This could make software less visible without making it less important.
Microsoft Moves Toward a New Software Layer
Microsoft has also made a major move toward this model.
In June 2026, Microsoft announced Work IQ APIs. The system aims to give agents access to work context across areas such as email, calendars, meetings, chats, files, people and business systems.
The larger idea matters more than any single feature. An AI agent needs context before it can act well. It needs to know who matters, what happened before, what information exists, which rules apply and what actions it can take.
Traditional applications often keep that context inside separate systems. A new agent layer can connect those pieces.
That creates a new software structure. AI models provide intelligence. Agents provide action. Context provides knowledge. APIs connect tools. Enterprise systems provide records, rules and permissions.
The result looks less like a collection of separate applications and more like a connected digital workforce.
OpenAI and Anthropic Move Higher Up the Stack
OpenAI and Anthropic have also pushed further into enterprise agent systems.
OpenAI launched Frontier in February 2026 as part of a broader effort around enterprise agents. Anthropic has also expanded its focus on agents and business workflows.
This creates a new strategic contest.
Traditional SaaS companies own applications. AI companies want to own the agent that decides how those applications should work together.
That difference could become critical.
If an employee asks an agent to complete a task, the agent may not care which application handles each step. It may choose Salesforce for one action, ServiceNow for another and an internal company system for a third action.
The agent becomes the decision layer.
That could reduce the power of individual applications that lack a strong data, workflow or infrastructure advantage.
Seat-Based Pricing Faces a Test
Traditional SaaS often relies on seat-based pricing.
A company may pay for 100 employees even if each employee uses only part of the available software. The model works well when people perform most tasks themselves.
AI agents can change that calculation.
If one agent can perform work that once required many employees to use an application, the customer may question the need for the same number of seats.
A new pricing model may then emerge.
Software companies could charge for tasks, transactions, agent use, workflow volume, compute use or successful outcomes. A hybrid model could combine subscriptions with usage fees.
Deloitte has highlighted this shift and cited a Gartner forecast that at least 40% of enterprise SaaS spending could shift toward usage, agent or outcome-based pricing by 2030.
That would represent a major change in software economics.
The customer would no longer pay mainly for access to software. The customer would pay more directly for work completed through software.
Point SaaS Faces the Greatest Risk
Not every software company faces the same level of danger.
A simple application with a narrow workflow may face strong pressure. An AI agent can often perform the same task through an API or a connected tool.
Basic dashboards may also face pressure if agents can retrieve the required data and provide a direct answer.
Simple project tools, basic CRM products, routine customer support platforms and other narrow applications may face similar challenges.
The risk becomes lower when a company owns critical data, complex workflows, identity, security, compliance or a major system of record.
A company that only offers a screen faces one type of competition.
A company that controls the data and business process underneath that screen faces a very different situation.
Data May Become the Strongest Moat
The next software advantage may come less from interface design and more from proprietary context.
An agent needs answers to many questions before it can safely perform a business task.
Who owns this account? What contract applies? What action has approval? Which employee has authority? What happened last month? What policy applies? What counts as a successful result?
Enterprise software often holds these answers.
That makes systems of record highly valuable in an agent-first world.
A CRM with deep customer history may gain value when an agent needs customer context. An ERP platform may gain value when an agent needs financial records. An identity platform may gain value when an agent needs permission to act.
The agent may reduce human use of the application while increasing machine use of the underlying platform.
That creates one of the biggest paradoxes in the new software market.
AI Can Also Make Internal Software Cheaper
Another threat comes from AI coding tools.
An enterprise once had to buy software for almost every specialised process. A company may now ask an AI coding system to create a custom internal tool for a narrow need.
This could reduce demand for some point products.
The new choice may look like this: pay a software vendor every year or create a small internal application with AI assistance and connect it to existing systems.
If AI reduces software development costs, more companies may choose custom solutions.
This does not mean every company will build its own enterprise software. Large systems still require security, reliability, compliance and long-term support. Yet simple workflows may no longer justify a separate SaaS subscription.
Trust Remains a Major Barrier
Full software autonomy still faces a major limit: trust.
An agent that writes a draft email creates little risk. An agent that changes a contract, approves a payment or alters a production system creates far greater risk.
Enterprise agents need clear identity controls, permissions, audit records, monitoring, policy checks and human approval for sensitive actions.
They also need a way to reverse mistakes.
Deloitte has highlighted transparency, reversibility, auditability and diagnostics as important requirements for autonomous SaaS agents.
This means autonomy will not arrive at the same speed across every business process.
Simple and repeatable work should move first. Sensitive financial, legal and operational tasks will need stronger controls.
The Rise of the Agent Control Layer
A future enterprise may have hundreds of agents from different vendors. Some may handle sales, others may handle support, finance, HR or software development.
That creates another software category: the agent control layer.
Such a system could decide which agent gets access to which data. It could control permissions, track actions, measure costs and record results.
The control layer could become as important as the application layer.
ServiceNow already places strong emphasis on this type of enterprise AI control. Its platform approach aims to help companies deploy, govern and scale AI across business processes.
This could give large enterprise software companies another path to long-term relevance.
Software May Become Outcome-Based
The long-term shift could change how companies describe software itself.
The old model sells access.
The newer model sells execution.
A customer may eventually care less about buying a support application for 500 employees and more about resolving 10,000 customer cases each month.
The difference looks small at first, but the economics change dramatically.
Software becomes closer to a digital employee or service provider. An agent receives a goal, performs a workflow and produces an outcome.
That model could reshape budgets across sales, customer support, finance, HR, procurement and operations.
What Happens to SaaS?
SaaS probably will not disappear.
Instead, SaaS will split into different layers.
Some products may lose value as agents replace much of their human interaction. Other platforms may gain value as agents rely on their data and workflows.
The strongest companies will likely combine reliable enterprise data, deep workflow control, strong APIs, identity, security and agent capability.
The weakest products may struggle when an agent can perform their main function without the need for a dedicated interface.
The market has already started to show both outcomes.
Salesforce has turned Agentforce into a major new revenue stream. ServiceNow has pushed ServiceNow AI above $1 billion in annual contract value. At the same time, Gartner estimates that $234 billion of enterprise application software spend could face agentic arbitrage by 2030.
These numbers do not describe a simple software collapse.
They describe a major redistribution of software value.
The Autonomous Software Era
The next stage of software may not remove applications. It may place agents between people and applications.
People will state goals. Agents will decide the steps. Enterprise systems will provide data and permissions. Software platforms will handle the underlying work.
That creates a new definition of software.
Software once helped people perform tasks.
Autonomous software will increasingly perform the tasks itself.
The winners of this shift will not simply have the best AI model or the prettiest interface. Strong positions will come from control of data, context, workflows, identity, security and enterprise relationships.
The biggest threat will hit software that only offers a convenient screen around a simple process.
The biggest opportunity will sit with platforms that can become the trusted operating layer for AI agents.
The real story, therefore, is not AI agents versus SaaS.
It is AI agents versus the old way of using SaaS.
Software may survive. Applications may survive. SaaS may survive.
What may disappear is the assumption that every software task needs a human sitting behind a screen.
Also Read – Startup Fundraising Deck: 12 Slides Investors Expect