Anthropic Automates 95% of Business Analytics Queries with Claude at ~95% Accuracy
bibryam · x · 2026-09-09
Anthropic details how Claude now automates 95% of its business analytics queries at 95% aggregate accuracy. Key lesson: adding thousands of SQL examples barely improved accuracy — in 80% of failures a relevant query was already retrievable. The bottleneck was structure, not access: wide denormalized tables breed inconsistent metric definitions, while ringfenced environments miss the long tail of questions. Letting agents loose on a warehouse creates a false sense of precision by cutting stakeholders off from documentation and expertise; the fix is baking curated datasets, definitions, and guardrails into the agent workflow for true self-service analytics.
More from coding & agent
- Proposal: Bringing MCP Apps to the Terminal with Dual HTML/JSON Views for CLI Agents — fka · 2026-09-09
- Engineers worry AI-generated code is killing code review — and team dynamics with it — charles_irl · 2026-09-09
- ExecCritic: test-verify-revise RL scaffold lifts Qwen3.5 on SWE-bench from 61.2% to 72.6% — SharonYixuanLi · 2026-09-09
- GPT-6 Astra one-shots a planetarium app; the killer feature was a light-pollution toggle — mhmazur · 2026-09-09
- Claude Code user burns $4,500 of API spend in 30 days on a $200/month plan — PratikKadam_ · 2026-09-09
- Oktobot: A Free Multi-Agent AI App That Runs on Android via Termux and Controls Your Phone — seko121 · 2026-09-09