JP Beeghly & Josh Maynard on AI Quality: Train Agents Like Junior Employees, Not Magic Tools

About this episode

The episode explores why enterprise AI requires robust business context. Both guests stress that “dimensional modeling isn’t obsolete in the age of LLMs — it’s more important than ever,” particularly when systems generate SQL queries. A central theme emerges: organizations should onboard AI “like a junior employee — given access, trained on terminology, supervised.” Key topics include context limitations across multiple data tables, data fragmentation risks, quality control gaps, non-deterministic outputs in business intelligence, and why brand strategy remains distinctly human-driven.