As the generative AI field continues to evolve, teams face a critical task in deciding whether to construct or procure their AI observability infrastructure. How should they navigate this decision as new research, foundation models, orchestration frameworks, and methods constantly upend established techniques? Informed by work with dozens of enterprises and companies that have both traditional ML models and LLM apps live in production, this paper suggests some approaches for making build-versus-buy decisions in today’s world.
This guide includes:
- Tradeoffs between building and buying technical infrastructure
- A menu of capabilities for LLM and ML observability to aid in estimating the man-hours and timetables required to build, and associated opportunity costs
- An approach for calculating ROI of AI observability, and results of a study on ROI from over 50 teams and 500 models
- Perspectives from top teams in the industry on how they approach build-versus-buy decisions
Read the Guide
About the author
