Task-Level Organization

Evidence Boundaries

このページでは次を混同しない。 OBSERVED: - AIが知識労働の一部タスクを実行し始めている - 企業システムが職種タイトルより細かい skills / tasks 表現へ向かっている兆候がある - 多くの役割は自動化可能なタスクと非自動化タスクの混在である INFERRED: - 組織はタスク単位の権限と作業配分モデルを必要とし始める - タスク自動化が進むほど human capability risk の重要性が増す SPECULATIVE: - Task-Level Organization が正式な企業オペレーティングモデル／評価カテゴリになる - タスク単位の ALLOW / HOLD / DENY が標準インフラになる 市場規模: 不明（推定しない）。 特定ベンダーの買収や導入事例は、リポジトリ内に検証済みソースがない限り引用しない。

Why Now

Enterprise software vendors are increasingly moving from job-title-based workforce management toward skills and task mapping. AI adoption makes this shift operationally important because automation does not affect every task inside a role equally. As AI becomes embedded in real work, organizations need a more granular operating model for: - task allocation - authority - human review - fallback - capability maintenance （特定ベンダー事例は、検証済みソースが追加されるまで断定しない。）

Structural Shift

FROM: JOB → PERSON → RESPONSIBILITY TO: TASK → CAPABILITY → ACTOR → AUTHORITY → REVIEW → CONSEQUENCE Actor may be HUMAN / AI / HYBRID. Authority and consequence remain organizational responsibilities even when execution is automated.

Decision Boundary Connection — ALLOW / HOLD / DENY

Task-Level Organization は既存の ALLOW / HOLD / DENY（および関連する Approve / HOLD Protocol）に接続する。自動化率のスケールではない。 AI-ASSIGNABLE → 定義された権限内では ALLOW に寄せ得る HYBRID / HUMAN REVIEW → 必要な人間レビュー／承認が満たされるまで HOLD HUMAN-RETAINED → AIによる自律実行は DENY CAPABILITY-AT-RISK は決定状態ではない。 ALLOW / HOLD / DENY と並存し得る resilience / human-capability フラグである。 例: 定型的な診断解釈は技術的に ALLOW でも、人間が練習しなければ fallback が劣化するなら CAPABILITY-AT-RISK。 既存関連: Pacing the Frontier（HOLD Protocol）、Agent Decision Constitution（モデル上の判断境界）。

AI-Off Test

Canonical scenario: 「AI is unavailable for 4 hours.」 Questions: - Can the team continue operating? - Which tasks stop immediately? - Which tasks degrade but continue? - Can humans identify incorrect AI output? - Which skills have not been practiced recently? - Are fallback procedures documented? - Is there a trained human backup? - Are junior employees still learning the underlying task? - Which decisions have become fully AI-dependent? - Which customer-facing processes fail? - Which regulated activities become non-compliant? AI-Off Test は災害復旧だけではない。capability-maintenance test でもある。 Canonical line: The question is not only whether AI can perform the task. It is whether humans can still perform it when AI cannot.

Human Reserve — task-level, not headcount-only

Human Reserve は、単に残した従業員数ではない。 次の状態を満たす人間能力の集合である: - available - practiced - accountable - recoverable Recommended line: Human Reserve should be measured at the task level, not only at the headcount level. Conceptual formula（定量メトリクスではない）: Human Reserve ≈ Critical Human Tasks Still Performable Without AI 既存関連: Headcount Is Losing Its Meaning（能力と人数の切り離し）、Future Senior Audit（将来の熟練が育っているかの監査隣接）。

Expertise Pipeline / Apprenticeship Risk

AI can remove junior tasks that historically trained future senior experts. AI performs junior task ↓ junior employee receives less practice ↓ fewer tacit learning opportunities ↓ future senior capability declines Canonical line: A task can be efficient to automate today and still be strategically dangerous to remove from human practice. CAPTURE（専門知の構造化）→ ALLOCATE（本 Signal）→ PRESERVE（能力保全）として読む。 既存関連: Synthetic Apprenticeship Market、Future Senior Audit Market、The Company Without Headcount 内の Expertise Pipeline 観測。

Reversibility

For every task allocation, ask: - Can AI execution be reversed? - Can authority return to a human? - How quickly? - Is the human still competent enough to take over? - Is the process state understandable? - Is decision history available? - Can the handoff occur mid-process? Canonical question: Which tasks require a reversible handoff between AI and humans? Preferred flow: AI → HUMAN REVIEW → AI → HUMAN TAKEOVER not simply: HUMAN → AI → PERMANENT AUTOMATION

Conceptual Sequence Note

本ページの中心連鎖: Expertise capture（CAPTURE） → Task-Level Organization（ALLOCATE） → Human capability maintenance（PRESERVE） 「Expertise as Training Infrastructure」「Human Capability Maintenance」という同名の公開 Opportunity は現時点で未収録のため、最寄り既存 Signal（Synthetic Apprenticeship / Future Senior Audit / Headcount Is Losing Its Meaning / Meaningful Human Review）へ接続する。