Scoped roles
People and agents are assigned defined responsibilities within the workflow.
Trust and governance
Trust should be designed into the work.
Important AI-supported work needs defined roles, approved context, bounded actions, human decision points and a record that can be reviewed later.
Governed by design
Make the operating boundaries visible.
These principles describe how roles, assistance, review and evidence are designed into an AI WorkBook solution.
Approved context
The solution identifies the records, documents, policies and history intended to support the work.
Controlled actions
Actions and handoffs are designed around explicit permissions and operating rules.
Human oversight
Consequential judgement, exceptions and approval moments remain with accountable people.
Reviewable evidence
Sources, workflow events, reviewers, decisions and outputs can be kept connected when designed into the solution.
Bounded change
Supported changes can be prepared, reviewed and verified before they affect operational work.
Engagement-specific assurance
Confirm the practical trust boundary for the proposed solution.
Deployment, security and data-processing arrangements depend on the solution and environment being proposed. They should be documented with the responsible customer, technical and legal owners rather than assumed from a general product page.
Information boundary
Identify approved sources, sensitive information, retention needs and permitted use for the workflow.
Access boundary
Define the people, roles, administrative responsibilities and environment access required to operate the solution.
Action boundary
Agree which systems and tools may be used, which actions are permitted and where approval is required.
Assurance boundary
Record the deployment, support, monitoring, change and recovery arrangements that apply to the engagement.