The Buy-or-Build Decision, Revisited: How Agentic AI Changes the Economics of Enterprise Software

Abstract: Advances in generative artificial intelligence, particularly agentic coding systems capable of autonomous software development, are disrupting the economics of the make-or-buy decision for enterprise applications. The "SaaSocalypse" narrative predicts that AI will render large segments of the Software-as-a-Service market obsolete by enabling firms to build software in-house at a fraction of historical cost. This paper adopts a conceptual research approach, combining transaction cost economics and the resource-based view with an assessment of current AI capabilities, to systematically re-evaluate the factors underlying the make-or-buy decision. It makes three contributions. First, it provides a factor-level analysis of how AI reshapes seven canonical decision determinants: cost, strategic differentiation, asset specificity, vendor lock-in, time-to-market, quality and compliance, and organizational capability. Second, it develops a typology of enterprise applications by their sensitivity to AI-induced shifts in make-or-buy economics. Third, it demonstrates that AI fundamentally transforms the governance properties of the Make option, shifting it from Williamson's pure hierarchy to a hybrid governance form that combines code ownership with external AI infrastructure dependency, with qualitatively different economics, capability requirements, and governance structures than pre-AI in-house development. The analysis finds that the SaaSocalypse thesis is overstated for most enterprise application categories; Make is most compelling for commodity utilities and differentiating custom applications in the AI era, while regulated and mission-critical systems remain predominantly in the buy domain.

• The paper’s main contribution is a revised framework that integrates agentic AI into the traditional make-or-buy decision, emphasizing shifts in cost, quality, and strategic control.
• It employs a blend of transaction cost economics and resource-based views to reassess application typologies and vendor lock-in risks in modern enterprise settings.
• The study highlights that developing internal AI orchestration capabilities can offer competitive advantages despite emerging compliance, operational, and technical debt challenges.

Agentic AI and the Make-or-Buy Decision in Enterprise Software

Introduction

"The Buy-or-Build Decision, Revisited: How Agentic AI Changes the Economics of Enterprise Software" (2604.26482) critically examines the implications of emergent agentic AI—specifically autonomous coding systems—on classic make-or-buy frameworks for enterprise software sourcing. The essay employs an analytic synthesis of transaction cost economics (TCE), resource-based view (RBV), and pragmatic IT portfolio strategies to derive revised logic for in-house software development versus external procurement in an era where AI obviates many historic constraints.

Analytical Foundations and Factor Framework

The paper grounds its reassessment in established IS theory, highlighting TCE's asset specificity, uncertainty, and transaction frequency as determinants of governance modes, and RBV's VRIN criteria as drivers for retaining strategic IT resources in-house. The analysis inspects seven canonical decision factors: lifecycle cost structure, strategic differentiation, asset specificity, vendor lock-in risk, time-to-market, quality/compliance, and organizational capability. These are systematically re-examined vis-à-vis AI’s current corpus-level coding capabilities, agentic workflows (SWE-agent, Claude Code, Codex, etc.), and the limitations seen around specification ambiguity, integration, and compliance.

Transformations of Canonical Make-or-Buy Factors

Cost Structure and Lifecycle Economics

Agentic AI compresses in-house build CAPEX, enabling rapid delivery of greenfield or commodity applications. OPEX is only partially reduced, as new AI-related governance and quality assurance overheads emerge. Vendors leverage the same productivity gains, often reinvesting realized cost savings into expanding platform integration and feature sets rather than passing savings to customers. The result: the historical TCO asymmetry favoring buy is diluted for commodity and custom-but-low-compliance apps, but persists for mission-critical and regulated systems.

Strategic Differentiation, Asset Specificity, and Rarity

AI reduces cost barriers for building differentiating custom applications, but widespread access to agentic coding tools creates a rarity paradox—competitive advantage shifts from the application itself to the organizational capability around AI orchestration, prompt engineering, and agent governance. For highly specific requirements that are well-articulated, AI sharply lowers the threshold where in-house development is rationalized, but tacit, ambiguous, or contradictory requirements remain resistant to automation.

Vendor Lock-in and Infrastructure Dependencies

In-house AI-driven development trades application lock-in for infrastructure lock-in. Dependence on model APIs and cloud compute is more fungible and less binding than traditional SaaS lock-in. Vendors counter by deepening switching costs via AI-native integrations and proprietary workflow personalization. The result is a moderate shift toward in-house make, provided infrastructure dependencies are actively managed.

Quality, Reliability, Compliance

While empirical benchmarks (e.g., SWE-bench Verified) demonstrate significant progress in agentic code reliability (80% resolution rate), unresolved issues concentrate in integration, compliance reporting, and security vulnerabilities. AI coding agents explicitly disclaim responsibility, transferring full liability for correctness and regulatory compliance onto the buyer or builder. Certified vendor offerings retain considerable advantage in regulated and mission-critical domains.

Organizational Capability

The minimum internal capability threshold for AI-augmented make is qualitatively distinct; organizations require prompt engineering, agentic validation, and governance skills. Firms below this threshold remain locked into buy; those investing in capability acquisition accrue path-dependent advantages. This dynamic alters the landscape of sustainable IT competitiveness.

Typology and Governance Transformation

The paper introduces a refined application typology based on complexity, domain specificity, and compliance exposure, yielding four principal categories: commodity utilities, differentiating custom applications, regulated standard applications, and mission-critical systems of record. AI-induced shifts are strongest for commodity and differentiating custom apps, weakest for regulated/critical systems. Furthermore, AI transforms make from pure hierarchy (internal control) to hybrid governance—code ownership is retained, but infrastructure dependency introduces market-like flexibility, information asymmetries, and non-determinism unique to agentic AI workflows.

Revised Decision Framework and Strategic Implications

A two-stage decision model emerges:

• Application classification by typology, which drives baseline make-or-buy heuristics.
• Factor weighting and capability assessment, with the organizational AI readiness threshold as a gating variable.

Implications are substantial: IS management should regularly re-evaluate portfolios using this typology, invest in AI orchestration capability as a strategic asset, and anticipate a competitive dynamic wherein both make and buy options co-evolve. The normalization of hybrid governance in make is projected to reshape standard development practices over a period of transition.

Empirical, Practical, and Competitive Considerations

The analysis points to several operational constraints: regulatory and data sovereignty requirements, quality/security risks associated with AI-generated code, management of technical debt, and challenges in reskilling/organizational change. Empirical validation is limited by the nascent state of real-world adoption, and the rapid advance of AI capabilities necessitates periodic framework reassessment. The competitive outcome will likely be a segmentation where vendors retain regulated and complex domains, while make expands in custom and commodity applications.

Conclusion

This work recalibrates classic make-or-buy analyses: AI-driven make is not merely a cost-reduction vector, but entails qualitatively transformed governance and capability requirements. The typology presented offers actionable heuristics for categorizing application sensitivity to AI. Theoretical contributions include the formalization of hybrid governance in AI-era make, and explicit identification of sustainable competitive advantage residing in organizational AI orchestration capability. Operational complexity, compliance exposure, and technical debt risks persist, underscoring the enduring relevance of lifecycle cost frameworks.

Empirical studies exploring the evolution of these dynamics, formal competitive interaction models, and updated TCO frameworks incorporating AI infrastructure costs are needed for further refinement. The essay ultimately positions AI as a potent force for selective insourcing and capability transformation, rather than universal vendor disintermediation.