Problem

The paper addresses the "SaaSocalypse" narrative, which posits that agentic AI will enable firms to build software in-house at a fraction of the cost, potentially rendering Software-as-a-Service (SaaS) obsolete. This widespread belief, however, often overlooks established information systems research indicating that operational costs typically constitute the majority of software's total cost of ownership, not just initial development. The central challenge for the paper is to systematically re-examine the traditional make-or-buy decision frameworks—Transaction Cost Economics (TCE) and the Resource-Based View (RBV)—to understand how the rapid advancements in generative AI and agentic coding systems fundamentally alter these long-standing economic considerations for enterprise applications. The core research question revolves around how these AI advancements modify the factors influencing the make-or-buy decision and what revised IT strategy logic should follow for organizations.

Method

The study employs a conceptual research approach, combining deductive analysis grounded in established information systems theories—specifically Transaction Cost Economics (TCE) and the Resource-Based View (RBV)—with a comprehensive assessment of the current capabilities and inherent limitations of emergent AI technologies, particularly agentic coding systems. This methodology facilitates a systematic re-evaluation of seven canonical factors that traditionally determine the make-or-buy decision. The analysis also develops a novel typology of enterprise applications, categorized by their complexity, domain specificity, and compliance exposure. This typology is then integrated into a revised decision framework that accounts for AI-induced shifts. This conceptual method is deemed suitable given the nascent stage of AI-augmented development at an enterprise scale, which currently restricts the feasibility of large-scale empirical investigations.

Results

The paper delivers three primary contributions. Firstly, it offers a detailed factor-level analysis, illustrating how AI reshapes seven traditional make-or-buy determinants: cost structure, strategic differentiation potential, asset specificity, vendor dependency and lock-in risk, time-to-market, quality/reliability/compliance requirements, and organizational capability. AI significantly reduces the development costs and time-to-market for less complex applications, but simultaneously introduces new cost categories such as AI infrastructure expenditure and governance overhead. It also shifts the primary source of strategic differentiation from the application itself to the firm's AI orchestration capability. While AI-augmented "make" offers protection against application-level lock-in, it introduces a new dependency on AI infrastructure providers. Critically, quality, reliability, and compliance requirements remain strong arguments for "buy," especially in regulated sectors, due to the challenges of liability, certification, and oversight for AI-generated code. Secondly, the paper develops a typology classifying enterprise applications into four categories: commodity utilities, differentiating custom applications, regulated standard applications, and mission-critical systems of record. It finds that AI-induced shifts favoring "make" are most pronounced for commodity utilities and differentiating custom applications, whereas regulated and mission-critical systems largely remain in the "buy" domain. Thirdly, the paper posits that AI fundamentally alters the governance properties of the "make" option itself. It transitions from Williamson's pure hierarchy to a hybrid governance form, combining internal code ownership and control over application logic with a reliance on external AI infrastructure providers. This transformation introduces distinct characteristics, including variable-cost AI production inputs, information asymmetry with AI providers, and non-deterministic production processes, fundamentally redefining in-house development.

Implications

The framework developed in this paper suggests three critical strategic implications for Information Systems (IS) management. Firstly, firms should undertake a comprehensive portfolio-level reassessment of their application landscape using the proposed typology to systematically identify suitable candidates for in-house "make" options, particularly within commodity utility and differentiating custom application categories where AI-assisted development offers the most compelling economics. Secondly, the analysis points to an "unbundling dynamic," where AI enables the selective replacement of individual modules or components rather than wholesale platform substitutions. This allows firms to combine vendor-provided foundational elements (e.g., data models, core processes) with custom-developed modules that deliver high differentiation. Thirdly, investment in developing organizational AI capability—encompassing prompt engineering, agent orchestration, and AI governance—is a crucial prerequisite for effectively realizing the potential of the "make" option in the AI era. This capability investment is a strategic decision with long-term competitive implications, as it can become a source of sustained competitive advantage. Practical challenges for adopting "make" in the AI era include navigating complex regulatory landscapes, ensuring data privacy and sovereignty when using third-party AI models, managing the unique quality and security risks of AI-generated code, mitigating the potential for increased technical debt, and overcoming significant organizational change management hurdles.