---
id: "framework-ai-accountability"
type: "framework"
source_timestamps: ["§ What Leaders Should Do"]
tags: ["governance", "risk-management", "organizational-design"]
related: ["concept-deliberate-inefficiency", "action-extend-provenance", "action-mandatory-sign-off", "action-pair-senior-junior", "action-escalation-rules"]
steps: ["\\\"Extend software provenance to capture AI use and human accountability (e.g.", "using the SLSA framework).\\\"", "\\\"Make named human sign-off mandatory for AI-generated production code", "pairing senior and junior engineers to preserve the apprenticeship pipeline.\\\"", "\\\"Build escalation rules that link AI-related failures to accountability and cost (e.g.", "triggering higher review requirements and structured postmortems for repeated incidents).\\\""]
sources: ["futures"]
sourceVaultSlug: "hbr-seg-futures"
originDay: 2
articleStem: "hbr-cl-84-big-tech-capability-crisis"
sourceUrl: "https://hbr.org/2026/06/big-techs-looming-capability-crisis"
sourceTitle: "Big Tech’s Looming Capability Crisis"
---
# AI Accountability and Capability Mitigation Framework

## Framework: AI Accountability & Capability Mitigation

A three-step managerial framework to reintroduce [[concept-deliberate-inefficiency|deliberate inefficiency]] into software-engineering workflows — preventing the accrual of [[concept-capability-debt-d2|capability debt]] and [[concept-judgment-debt|judgment debt]] while managing the [[claim-latent-ai-errors|latent risks]] of AI-generated code.

### Step 1 — Extend software provenance
Capture AI use and human accountability, e.g. via the [[entity-slsa-framework|SLSA]] framework. Every shipped module carries metadata: which AI tools touched the code, who reviewed it, who signed off. → [[action-extend-provenance]]

### Step 2 — Mandate named human sign-off
Require a specific, named engineer to sign off on AI-generated production code, and **pair senior + junior** so the sign-off doubles as a teaching moment that preserves the apprenticeship pipeline. → [[action-mandatory-sign-off]], [[action-pair-senior-junior]]

### Step 3 — Build escalation rules
Link AI-related failures to accountability and cost: repeated incidents trigger higher review requirements, senior sign-offs, and structured postmortems — making negligence expensive enough that hiring and training humans stays viable. → [[action-escalation-rules]]

All three share one feature: they put deliberate inefficiency back into a system AI is racing to eliminate (see [[quote-deliberate-inefficiency]]).