The build vs. buy dilemma at the heart of enterprise AI

2026-07-17 · Source: CIO · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

The traditional "build vs. buy" paradigm for enterprise software, dominant for three decades, is being reshaped by AI. AI introduces an architectural layer across data, processes, and decisions, often requiring data in vendor cloud environments, which conflicts with many large enterprises' on-premises or private cloud infrastructure. The article reframes the choice from a binary "build vs. buy" to three distinct approaches: "buy embedded" (vendor-native AI), "buy platform" (vendor AI infrastructure), and "compose" (connecting third-party models like Claude, GPT, Gemini, or open-weight models to existing landscapes). These options differ significantly in data location, governance, and architectural change. Data sovereignty, exemplified by Disney's \$100 million annual spend with Amazon yet building its own AI, and GDPR obligations, is a critical architectural constraint. Enterprises will likely use a mix, leveraging embedded AI for in-application productivity and compose for cross-application orchestration, where McKinsey research predicts significant near-term productivity gains within eighteen months. The key is deciding which decisions vendor AI should make versus those kept in-house.

Key takeaway

For AI Architects or IT Leaders evaluating enterprise AI solutions, recognize that the "build vs. buy" framing is a category error. Your primary decision should be architectural: which business decisions will your vendor's AI make, and which will remain under your control? Prioritize embedded AI for in-application productivity and compose approaches for cross-application workflows, especially where data sovereignty or custom orchestration is critical. Be sure to budget for the governance, audit, and accountability costs inherent in composed solutions.

Key insights

Enterprise AI strategy requires distinguishing between embedded, platform, and compose approaches, driven by data sovereignty and architectural control.

Principles

• AI strategy is architectural, not procurement.
• Data sovereignty is a core architectural constraint.
• Vendor AI assumes specific data location and governance.

In practice

• Use embedded AI for in-application productivity.
• Use compose for cross-application orchestration.
• Budget for governance and audit trails with compose.

Topics

• Enterprise AI Strategy
• Data Sovereignty
• Cloud Architecture
• AI Governance
• Build vs. Buy
• Foundation Models

Best for: CTO, VP of Engineering/Data, Executive, Director of AI/ML, AI Architect, Consultant

Related on AIssential

• The Experiment Is Over: Enterprise AI Enters Its Execution Era — The enterprise AI landscape has shifted from experimentation to execution, demanding measurable ROI and strategic adoption.
• Build vs Buy in the Age of AI: Cloud Cost Management Tools — The build vs. buy calculus for cloud cost management has shifted, favoring buy as complexity and feature expectations grow.
• Buy the Harness, Build the Differentiator — Build cheaply on disposable layers, invest deeply in durable assets, driven by specs and evals.
• AI Models as a Commodity and Why Data Foundations Decide Who Wins - with Guillermo B. Vazquez of SAP — AI models are a commodity; robust data and process foundations are paramount for successful enterprise AI adoption.
• The Next Category Is Not Agentic AI — The next enterprise AI category is Sovereign Intelligence Architecture, not agent-only propositions, focusing on integrated human-digital workforce governance.
• Enterprise Architects Have Stepped Out Of The Ivory Tower — Enterprise architecture's value surged by aligning with business outcomes and addressing digital transformation and AI complexities.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by CIO.