---
id: "contrarian-humans-teach-implicit-rules"
type: "contrarian-insight"
source_timestamps: ["§ The Hidden Substitution"]
tags: ["human-in-the-loop", "knowledge-management"]
related: ["entity-ramp"]
challenges: "The view of human reviewers as mere quality-assurance backstops rather than active codifiers of tacit institutional memory."
speakers: ["K. Sudhir"]
sources: ["agentic"]
sourceVaultSlug: "hbr-seg-agentic"
originDay: 6
articleStem: "hbr-new-26-agentic-systems-implicit-rules"
sourceUrl: "https://hbr.org/2026/06/how-to-design-agentic-systems-around-the-implicit-rules-that-govern-your-company"
sourceTitle: "How to Design Agentic Systems Around the Implicit Rules that Govern Your Company"
---
# Humans in the loop aren't gatekeepers; they are teachers of implicit rules

**Contrarian insight — challenges:** the view of human reviewers as mere quality-assurance backstops rather than active codifiers of tacit institutional memory.

Typically, human reviewers in AI systems are seen as **quality-control gatekeepers** meant to catch errors. Using the [[entity-ramp-d26|Ramp]] example, the author reframes their role: humans handle the toughest 10–15% of edge cases *specifically to surface the tacit rules the written policy didn't anticipate*, thereby **teaching the system what the [[concept-implicit-organization]] knows.** Fragile institutional memory becomes durable infrastructure.

**Enrichment note:** Well-aligned with contemporary human-centered AI and MLOps practice — algorithmic auditing / red-teaming to discover blind spots, active-learning loops where human experts refine rules over time, and overseers positioned as *co-designers* rather than backstops. **Caveat:** codifying implicit rules can also *ossify* them — including dysfunctional or biased ones — so teams must judge which tacit rules are worth preserving vs. redesigning.