AI Reveals Why BI Still Matters
Summary
Simon Späti argues that BI was never really about dashboards — it was about the primitives underneath: metrics, semantics, governance, and trusted definitions. He traces the recurring 'BI is dead' narrative through industry voices like Benn Stancil, Hex, and Rill's Mike Driscoll, showing that each wave rediscovers the same truth. While AI agents can generate dashboards and chat-based analytics, they actually make BI's foundational layers more important, not less, because agents hallucinate without governed semantic layers. The post highlights maintenance as the elephant in the room: generating dashboards is easy, but maintaining them at scale is where real value lies. Späti proposes BI-as-Code as a path toward maintainable, versioned, agent-friendly BI infrastructure. The conclusion is that BI primitives are not casualties of the AI era but essential infrastructure for it.
Key Insight
AI agents don't eliminate the need for BI — they expose that the real value was always in the governed metrics, semantic layers, and maintained infrastructure underneath the dashboards.
Spicy Quotes (click to share)
- 6
Ask a BI engineer what they actually spend their time on: it's not building dashboards.
- 6
Agents are blind — they can't see dashboards. But they do need access to the primitives behind them.
- 7
BI was never about dashboards.
- 7
A pivot table is a REPL for BI; that's not possible with chats.
- 8
Everyone wants to build. Nobody wants to maintain.
- 6
Maintenance is a luxury these days.
- 7
An AI agent can run 50x faster, but if the tools it depends on were designed for human speed — slow query APIs, brittle CLIs, unversioned metrics — the overall gain collapses to 2-3x.
- 7
The hard part was never the visualization. It was always the semantics beneath.
Tone
opinionated, practical, reflective
