Supervision is the mechanism that ensures AI is infrastructure, not authority. The platform executes. The human decides.

The Review and Acknowledgement Flow

After execution completes, the output enters a supervised review state. The user reviews the draft and makes one of three decisions.

Draft
Output produced — not yet final

The platform has completed execution and produced a structured output. The output is available for review but carries no professional weight until acknowledged. It is clearly marked as a draft.

↓ Review
Acknowledged
Professional accepts responsibility

When the user clicks Acknowledge, they formally accept responsibility for the output. An audit record is created immediately with their identity and timestamp. The output is now edit-locked — no further changes to the draft content.

↓ Finalise
Finalised
Immutable — sealed to Vault

Finalisation seals the output permanently. The finalised record is immutable. It is saved to the Output Vault (shown in the current interface as Generated Vault). Billing settles at this point. The audit trail is complete.

If the output is not satisfactory: The user may reject it and re-run the app with corrected inputs. A rejected draft is not charged. The rejection is recorded in the audit trail.

When Is Supervision Required?

Not every app requires explicit acknowledgement. But any app that produces outputs with legal, financial, regulatory, or professional consequences must require it.

Legal outputs

Legal memos, case analysis, contract summaries, and any output a professional will act on without independent verification require explicit acknowledgement.

Financial outputs

Financial assessments, credit analysis, management accounts, and structured financial outputs must pass through supervision before finalisation.

Regulatory outputs

Compliance documents, regulatory submissions, policy outputs, and any output presented to a regulator or governing body require explicit human sign-off.

Related Platform Features