---
id: "claim-standard-rai-too-slow"
type: "claim"
source_timestamps: ["§ Flaw #1: It's too slow to be effective."]
source_url: "https://hbr.org/2026/05/what-are-your-companys-ai-nightmares"
source_title: "What Are Your Company's AI Nightmares?"
tags: ["speed", "policy-failure"]
related: ["concept-standard-rai-approach", "concept-agentic-ai-governance-gap"]
confidence: "high"
testable: true
speakers: ["Reid Blackman"]
sources: ["governance"]
sourceVaultSlug: "hbr-seg-governance"
originDay: 7
articleStem: "hbr-cl-82-ai-nightmares"
sourceUrl: "https://hbr.org/2026/05/what-are-your-companys-ai-nightmares"
sourceTitle: "What Are Your Company’s AI Nightmares?"
---
# Standard RAI policies are too slow for modern AI

**Claim:** Getting a standard Responsible AI policy from kickoff to board approval takes a **minimum of one year** for large organizations. Because AI technology changes on a **monthly** basis — Blackman cites OpenAI's ([[entity-openai-d7]]) introduction of agentic AI rendering a newly approved policy out of date in **five months** — centralized policy creation is structurally incapable of keeping pace with AI development.

This is the quantified core of [[concept-agentic-ai-governance-gap]] and the first of the three flaws that break [[concept-standard-rai-approach]].

**Confidence: high. Testable: yes** (timelines and obsolescence events are observable).

**Enrichment calibration:** *Directionally and qualitatively supported* — traditional AI policy programs are widely reported to take many months to a year or more (education, gap analysis, strategy, implementation), and policy cycles measured in quarters clash with monthly model releases. However, the specific **"minimum one year"** figure and the **five-month OpenAI anecdote** are *experiential / consulting data* (the anecdote is not publicly documented with client attribution or verifiable timeline), not universal empirical constants. Present them as illustrative practitioner data, not proven law. Some firms are experimenting with *agile* policy development (living documents, sprint-integrated updates), which pushes back on the "inherently year-long" characterization.


## Related across articles
- [[claim-consensus-fatal-post-ai]]
- [[claim-regulators-poorly-positioned]]