Skip to content

nanda-tech-buy-vs-build.md

Latest commit

 

History

History
35 lines (24 loc) · 2.42 KB

nanda-tech-buy-vs-build.md

File metadata and controls

35 lines (24 loc) · 2.42 KB
Raw
title NANDA Buy vs. Build for GenAI
source-id mit-nanda-genai-divide-2025
wiki-page (02_References/enterprise-ai/wiki/skills/process/buy-not-build-for-genai.md)
last-synced 2026-06-07

NANDA Buy vs. Build for GenAI

What this reference contains

MIT NANDA 2025's empirical finding on the build vs. buy decision for GenAI: firms that buy external AI capability are 2× more likely to capture value than firms that build internally. The foundation model commoditizes. The orchestration layer and the workflow design are where the moat lives.

Key claims (with citations)

  • 2:1 buy advantage. Firms purchasing GenAI capability (via APIs, SaaS, vendor platforms) vs. building in-house models are 2× more likely to achieve measurable P&L impact. Source: MIT NANDA 2025.
  • The model commoditizes. Foundation model capability is vendor-trained; the strategic question is orchestration, integration, and workflow design — not which model to train. Source: MIT NANDA 2025.
  • Orchestration layer is the moat. The AI core — orchestration, model lifecycle, self-service consumption — is what compounds. Source: MIT NANDA 2025; reinforced by Accenture (5 months vs. 18 months with shared AI core).
  • Three vendor archetypes. Different sourcing paths serve different needs: full-stack platforms, specialist APIs, and systems integrators. Choosing wrong vendor archetype is as costly as choosing wrong build/buy direction. Source: MIT NANDA 2025.
  • Low setup burden, fast time-to-value outperforms heavy enterprise builds. Tools with minimal integration overhead generate faster returns for most use cases. Source: MIT NANDA 2025.

Decision framework

  1. Can an existing vendor capability solve ≥80% of the use case? → BUY
  2. Does the use case require genuinely proprietary data or process logic? → BUILD that specific layer only
  3. Is the value in the model, or in the integration and workflow? → Almost always the latter → orient the build investment there

When to consult

  • For skill tech-buy-vs-build: this is the primary empirical anchor for the BUY recommendation and vendor-archetype selection.
  • For skill tech-stack-diagnostic: the 2:1 finding reframes "build vs. buy" as a stack diagnostic question — where in the stack does the value actually live?

Source

MIT NANDA GenAI Divide Study 2025. Buy-vs-build analysis across enterprise GenAI deployments; vendor-archetype taxonomy; sourcing discipline findings.