Is Generative BI failing?
summary
Argues GenBI stalls because of an unrealistic “ask anything, get magic” promise and users expecting one-shot answers; pilots mostly die. Builds a RAG + text-to-SQL prototype against Helium Mobile data, sees only a few users try it a couple times, then pivots to a semantic editor starting from chart that shows the derived table and provenance and lets users iterate and save. Notes dozen+ billion-dollar coding copilots, almost no GenBI unicorns. Integrates with Superset and MCPUX, open-sources semanticgrid.ai, and reports much better uptake.
AI-written summary · about
chapters (5)
- 0:00Introduction and background
- 3:18First attempt at GenBI
- 6:52Why GenBI fails users
- 9:44Redesigning the GenBI flow
- 12:58Future integration plans
related talks
16mGen AI Killed BI. Claude Says It's Fake News!
45mFrom hand-down maps to AI/BI: a journey toward Databricks
17mGenerative AI: Keeping it real
15mPractical Evolution of Data Science with GenAI
20mUsing Generative AI to Increase the Impact of Your Data Science Work
24mScalable GenAI in Production
45mGenerative AI in medicine: Improving patient & provider experiences
17mOn Bringing AI to Business Intelligence