By Daniel Lambert

​​​​​​

Artificial Intelligence is fundamentally changing the economics of enterprise software. As AI agents become increasingly capable of executing business processes, automating workflows, analyzing information, and supporting decisions, organizations face a strategic question: Should they continue paying recurring subscriptions for every SaaS application, or can some of these business capabilities be delivered more efficiently through AI? Enterprise Architecture provides the framework for answering this question systematically, balancing business value, cost optimization, flexibility, security, and long-term architectural sustainability.

​

For the past two decades, Software-as-a-Service (SaaS) has transformed enterprise technology. Organizations have increasingly adopted cloud-based applications to support customer relationship management, finance, procurement, human resources, marketing, project management, and numerous other business functions.

SaaS has delivered significant benefits, including faster deployment, reduced infrastructure responsibilities, continuous software updates, and access to sophisticated business functionality without substantial upfront development investments. However, this convenience has come at a cost.

​

Many organizations now operate hundreds, sometimes thousands, of applications. Their portfolios often contain overlapping functionalities, underutilized licenses, fragmented data, complex integrations, and increasing vendor dependencies. Subscription costs continue to accumulate, even when the business value delivered by individual applications remains difficult to quantify.

​

Artificial Intelligence introduces a new alternative. AI agents can increasingly perform activities previously requiring dedicated software applications, specialized interfaces, or extensive human intervention. AI-assisted software engineering is also reducing some of the historical costs and complexity associated with developing customized solutions. This development challenges a long-standing assumption: Organizations must purchase or subscribe to a software application whenever they need a new business capability. In her LinkedIn article, “AI Is Changing the Economics of Enterprise Software – and the Business Case May No Longer Add Up,” enterprise architecture strategist Gwen Murphy raises an important question: Why should organizations continue renting capabilities they can increasingly build, own, extend, and economically optimize themselves? [i]

​

However, replacing SaaS applications with AI is not simply a software development or cost-reduction exercise. It requires understanding how business capabilities are delivered, how applications interact, where enterprise data resides, which operational controls must be preserved, and whether alternative architectures generate sustainable business value. The objective should not be to replace SaaS with AI indiscriminately. It should be to determine which business capabilities can be delivered more effectively, economically, and securely through SaaS, AI, or a combination of both. This is where Enterprise Architecture (EA) becomes indispensable.

​

1. The Rise of AI Agents: Is the Traditional SaaS Model Under Threat?

​

Traditional SaaS applications are built around predefined functionality, standardized business processes, configurable workflows, and user interfaces. Organizations subscribe to these applications and adapt their business operations to the capabilities and constraints of the selected platforms. AI agents introduce a fundamentally different approach. Instead of requiring users to navigate multiple application interfaces, AI agents can interpret business objectives, retrieve information, interact with enterprise systems, execute workflows, and coordinate activities across applications.

​

• Customer Relationship Management: AI agents can qualify leads, prepare personalized communications, update customer records, and recommend sales activities.

• Human Resources: AI agents can coordinate employee onboarding, answer policy questions, process routine requests, and support employee administration.

• Procurement: AI agents can research suppliers, compare quotations, prepare purchase requisitions, and monitor procurement activities.

• Finance: AI agents can assist with invoice processing, expense classification, reconciliations, financial analysis, and management reporting.

• Project Management: AI agents can track deliverables, identify risks, prepare status reports, and coordinate follow-up activities.

• Enterprise Architecture: AI agents can support application-to-capability mapping, application rationalization, architecture documentation, impact analysis, and capability-based roadmap development.

Many of these activities currently require specialized SaaS functionality. The emergence of AI raises the possibility that organizations could deliver selected capabilities through AI agents integrated with enterprise data and existing systems rather than purchasing additional software products.

​

From application-centric to capability-centric technology

​

Historically, enterprise technology investments have often followed an application-centric approach. When a business function requires new functionality, the organization evaluates available software products, selects a vendor, implements the application, integrates it with existing systems, and pays recurring licensing fees. AI enables an alternative approach. Instead of beginning with software products, organizations can begin with the business capability itself. For example, rather than asking which procurement application should be purchased, an organization could first determine which procurement activities require automation, which decisions can be supported by AI, and which transactions must remain within existing enterprise systems. 

​

An AI agent might research suppliers, analyze quotations, and recommend purchasing decisions while an ERP platform continues to manage purchase orders, financial commitments, and transaction records. The result is not necessarily the elimination of the ERP application. It is a more intelligent distribution of responsibilities between AI agents and traditional enterprise systems. AI may replace the functionality of certain SaaS applications without replacing the systems of record, data structures, or transactional controls on which the enterprise depends.

​

The SaaS business model is evolving

AI is also challenging the economics of per-user SaaS licensing. Traditionally, software spending increases as organizations add employees, licenses, modules, and applications. However, AI agents can perform substantial volumes of work without requiring a conventional human user interface. [ii] This creates an opportunity to reconsider whether software pricing should be based primarily on user access, transactions, consumption, or measurable business outcomes. It also creates competitive pressure on SaaS vendors to incorporate AI capabilities, revise pricing models, and demonstrate greater business value.

​

The likely outcome is not the disappearance of SaaS. Rather, the enterprise software market may evolve toward a combination of traditional applications, AI-native solutions, intelligent agents, and consumption-based services. For enterprise architects, application portfolio strategies must evolve to accommodate new ways of delivering business capabilities. [iii]

​

2. SaaS vs. AI: Comparing Costs, Flexibility, Capabilities, and Business Value

​

One of the strongest arguments for reconsidering SaaS investments is economic. Organizations often continue renewing SaaS subscriptions because the applications are already implemented, integrated, and embedded in daily operations. However, historical investments should not automatically justify future expenditures. As AI changes the cost of developing, integrating, and operating software functionality, organizations should revisit the traditional build-versus-buy decision.

​

From renting software to owning capabilities

Gwen Murphy’s article highlights a potentially significant shift: organizations may increasingly be able to own the data models, business logic, knowledge, and workflows that support their operations rather than depending entirely on proprietary SaaS platforms.

​

This is particularly relevant when the organization’s competitive advantage resides in its unique business processes, institutional knowledge, and enterprise data. A SaaS vendor may provide a sophisticated application interface, but the underlying business knowledge and relationships often belong to the enterprise.

​

AI-powered solutions built around enterprise-owned data, open standards, reusable services, and knowledge graphs may provide greater architectural flexibility and control. However, ownership does not automatically mean lower costs. Organizations that develop their own AI-enabled solutions also assume responsibility for engineering, maintenance, cybersecurity, reliability, monitoring, governance, and ongoing technological evolution.

​

The hidden cost of SaaS plus AI

​

A particularly important concern is the tendency to add AI functionality to existing SaaS environments without eliminating corresponding costs. An organization might continue paying its SaaS subscriptions while introducing AI agents, orchestration platforms, model inference services, data pipelines, and additional integration infrastructure.

Instead of reducing technology expenditures, AI becomes another layer of recurring costs. Every new agent can generate additional inference costs, API transactions, data retrieval operations, monitoring requirements, and maintenance responsibilities. Organizations must evaluate both the initial implementation cost and the ongoing operational economics. ​Figure 1 above compares traditional SaaS and AI-powered alternatives across financial, functional, operational, and governance criteria.

​

Neither approach is universally superior. A mature SaaS application supporting complex financial transactions may remain more economical and reliable than developing an AI-native replacement. Conversely, an expensive specialized SaaS application supporting simple administrative workflows may be an attractive candidate for replacement. ​An illustrative financial comparison Figure 2 illustrates the potential annual cost difference between the two architecture approaches discussed below. ​Gwen Murphy presents an illustrative Enterprise Architecture scenario involving approximately 1,000 applications. Her analysis compares an annual run rate approaching $915,000 for a traditional SaaS architecture supplemented with agentic capabilities against approximately $570,000 for an alternative architecture using an enterprise-owned graph, an open-standards data model, and an AI platform. [iv]​