DataFramed #329 Building Trust in AI Agents with Shane Murray, Senior Vice President of Digital Platform Analytics at Versant Media
55 snips
Nov 3, 2025 Shane Murray, a seasoned data executive and SVP at Versant Media, discusses the critical interplay between data quality and AI reliability. He shares cautionary tales of AI failures impacting reputations, highlighting essential data quality controls for successful implementations. Shane explains the concept of context engineering and the roles vital for developing trustworthy AI systems. He emphasizes that many AI hallucinations stem from poor data rather than just model flaws, advocating for rigorous data monitoring and clear accountability to build user trust.
AI Snips
Chapters
Books
Transcript
Episode notes
Invest In Context And Quality Roles
- Do invest in context engineering to provide trusted inputs to agents via RAG, prompts, and APIs.
- Do assign data scientists to define and evaluate AI quality using human feedback and LLM-as-judge techniques.
Build Modular Platforms When Needed
- Do build a platform only after several successful use cases to standardize tooling and avoid forced adoption.
- Do design for modularity so you can swap components as models and tools evolve rapidly.
Certify Data Before Agent Access
- Do certify core datasets before exposing them to conversational BI or agents, including synonyms and metadata.
- Do plan monitoring and incident response for both structured and unstructured data used by many downstream teams.
