---
id: "concept-codifying-judgment"
type: "concept"
source_timestamps: ["¶6", "¶7", "§ Where to Start"]
tags: ["knowledge-transfer", "tacit-knowledge", "context-engineering"]
related: ["concept-judgment-infrastructure", "framework-scenario-based-extraction", "contrarian-experts-cannot-document"]
definition: "Translating tacit decision-making principles — like risk tolerance and exception handling — into structured, explicit guidance that AI agents can execute."
sources: ["agentic"]
sourceVaultSlug: "hbr-seg-agentic"
originDay: 6
articleStem: "hbr-new-27-teach-ai-your-decisions"
sourceUrl: "https://hbr.org/2026/06/teach-your-ai-how-you-make-decisions"
sourceTitle: "Teach Your AI How You Make Decisions"
---
# Codifying Judgment

Codifying judgment is the act of translating the implicit, internalized decision-making rules of experienced employees into explicit, structured formats that AI agents can reference. Historically, organizational expertise was absorbed through mentorship, observation, and experience — a model that works for humans but fails for AI. AI agents cannot infer context from organizational culture or absorb norms through osmosis (see [[claim-agents-cannot-infer-context]]); they operate strictly on what is made explicit.

A common failure mode occurs when companies deploy customer-facing agents without codifying how top reps handle pricing exceptions, frustrated long-term customers, or out-of-policy requests. The agent eventually drifts off track because the necessary inferred context was never written down.

Crucially, the authors note that asking experts to simply write down their judgment rarely works — experts possess far more tacit knowledge than they can articulate in the abstract (see [[contrarian-experts-cannot-document]]). Effective codification requires creating conditions where judgment surfaces naturally, such as debating edge cases in a panel setting and using the resulting transcripts as the context layer for agentic deployments. The tactical method for this is [[framework-scenario-based-extraction|scenario-based judgment extraction]], which feeds directly into [[concept-judgment-infrastructure]].

**Enrichment note:** The mechanism is consistent with tacit-knowledge research (Nonaka & Takeuchi) and with cognitive task analysis / critical-decision-method elicitation, though empirical evaluation of transcript-as-agent-context specifically is still limited. Traditional documentation retains value alongside it — see [[cp-sops-still-valuable]].


## Related across articles
- [[action-convert-to-markdown]]
- [[concept-brand-code]]
- [[concept-professional-discretion]]