FCGForensic Capital GovernanceBooks and leadership
Concept 07All concepts

Human Authority in AI-Assisted Work

What may AI prepare—and what must accountable people retain?

AI can maintain and prepare evidence. It cannot inherit professional, contractual, acceptance, risk, or capital authority.

FCG / Decision architecture

Evidence before capital moves

  1. 01

    Funded belief

    The assumptions, outcomes, constraints, and value logic that justified commitment.

    Record object
  2. 02

    Current evidence

    The authoritative record of what has changed, been proven, or remains uncertain.

    Record object
  3. 03

    Four realities

    Business, delivery, technical, and financial evidence reconciled across boundaries.

    Record object
  4. 04

    Governing condition

    The material relationship that should control what must happen before commitment.

    Record object
  5. 05

    Human authority

    The accountable people who interpret uncertainty, accept evidence, and retain judgment.

    Current lens
  6. 06

    Capital decision

    Continue, pivot, pause, stop, or require proof—with the rationale preserved.

    Record object
Machine preparation may support the recordAccountable authority remains human

Decision brief / DB-07

For executive consideration

A governed reading of human authority in ai-assisted work.

01 / The condition

Human Authority is the explicit boundary between machine preparation and accountable judgment. Models may gather, classify, compare, test, summarize, and propose. Authorized people must still interpret material uncertainty, accept evidence, choose a disposition, and preserve the rationale.

02 / Why reporting misses it

Agentic workflows can continue because they have technical permission, not because continued action remains justified. When recommendation, authorization, execution, and verification collapse into one automated path, accountability becomes difficult to locate.

03 / Evidence to examine

  1. 01The provenance, limits, uncertainty, and disagreement within machine-prepared analysis.
  2. 02Explicit decision rights for recommendation, acceptance, authorization, execution, and verification.
  3. 03Stop conditions, overrides, preserved rationale, and the identity of the accountable human authority.

04 / Authority involved

Models may gather, compare, summarize, test, and propose. Authorized people retain professional, contractual, acceptance, risk, and capital authority.

05 / Executive question

“What may AI prepare—and what must accountable people retain?”

06 / Illustrative condition

A model identifies conflicting evidence and recommends pausing a release. It can show sources and explain its reasoning. It cannot accept technical risk, revise contractual obligations, authorize spend, or become the accountable owner of the decision.

Decision consequence

  • +Faster automation increases the need for visible decision rights.
  • +Model disagreement and uncertainty must remain inspectable.
  • +Chain of custody protects both human challenge and responsible use of AI.

Book connection

Both volumes preserve a firm boundary: AI may prepare the decision record, but humans retain judgment and authority.

Related research

The AI Workflow Had Permission to Continue. It Still Needed a Reason To.↗AI Found the Problems. It Still Couldn’t Change the Book.↗