Governing for Answers

Pillar 04 · Accountability

C1 Role Profiles

Someone must own what the AI tells people.

Form Document

What it's for

Data owners are accountable for the data an AI tool reads. Nobody, in most organisations, is accountable for what the tool does with it. When an answer is wrong, the fault could sit in the sources, the definitions, the model, the prompt or the interface, and every team can point at another. These profiles create the two roles the framework adds, the use-case owner and the agent owner, and the human reviewer and council that make them work.

What you get

  • AI use-case owner The new role accountable for an AI use case: the decision to deploy it, its controls, and getting failures fixed.
  • Agent owner The new role accountable for what an agent is allowed to do.
  • Human reviewer The supporting role that checks an AI tool's work.
  • Data or AI council Its terms of reference.
  • Appointment record Who holds each role for a use case, with a deputy and the date appointed.

How it connects

It appoints the owners S1 names for each use case and agent, and the human reviewer that C2's RACI and C3's charter rely on.

A flavour of the worked example

From the worked example APP-EX-04
Use case (S1)EX-04
Use-case ownerHead of Fund Reporting
Agent ownerFund Reporting Systems Lead
Human reviewerFund Controller
AcknowledgedYes

A few fields from one record of the toolkit's worked example, with long text shortened.

Next step

Guided build coming next.

See all the kits

Request access

Ask for access to the guided builds, including this kit's when it is ready.

Your name, email and organisation are kept for up to 12 months so Duncan can reply to your request and, only if you tick the box, email you about Governing for Answers. The internet address the request comes from is used only to count requests a day and is not kept with your details. To withdraw your consent or have your details removed, say so through the comment form on the Toolkits page, giving the email you used.