O'Reilly Builds 4-Step AI Intelligence System, Grounding Decisions in 67% Toil Analysis
Updated
Updated · O'Reilly Media · Aug 5
O'Reilly Builds 4-Step AI Intelligence System, Grounding Decisions in 67% Toil Analysis
1 articles · Updated · O'Reilly Media · Aug 5
Summary
O’Reilly said it built an internal organizational intelligence system that links company data to AI through MCP, adds an expert review layer, and uses a GitHub-based tool called Superanswers for collaborative analysis.
A test case shifted a hiring request into a structural diagnosis: after adding O’Reilly’s Expert MCP, the system flagged about 67% operational toil—above a roughly 50% SRE threshold—and recommended a toil audit over adding headcount.
The 4-step setup maps an organization’s information hierarchy, connects systems such as Jira, GitHub, Cortex and Datadog, applies a reasoning “skill” file, and keeps humans in the loop to review assumptions and gaps.
Superanswers stores AI-generated reports as Markdown in GitHub, renders them through GitHub Pages, and captures discussion in GitHub Discussions so the AI can revise documents using versioned comments and emerging consensus.
O’Reilly argues the key gain is not more facts but framework-grounded recommendations with named citations, while stressing the system still requires human judgment and does not eliminate hallucination risk.