Skip to main content
Intelliger

Human control

Put authority at the point of action.

Configure what the worker can read, prepare and execute, then place approval and escalation exactly where professional or regulated judgement is required.

Authorised professional reviewing an AI-prepared case before approval

The question

Who has authority when the case, risk or proposed action changes?

01

Authority assigned

02

Approval gates enforced

03

Exceptions routed with context

04

Actions attributable

Control at runtime

Human oversight works only when it is designed into the process.

01A generic approval button at the end of an automated workflow is not enough. The worker must know which decisions require which role, what evidence that person needs and what happens when authority is unavailable or the case is unfamiliar.

02Intelliger applies those rules while the case is running. The worker can continue routine preparation, pause before a consequential action or escalate immediately when its conditions no longer hold.

Authority path

Apply control before the action, not after it.

The runtime evaluates authority and risk as the case moves through the process.

  1. 01

    Identify

    Resolve the worker, operator, role and case context.

  2. 02

    Evaluate

    Check the proposed action against policy and authority.

  3. 03

    Prepare

    Assemble evidence and unresolved questions for review.

  4. 04

    Approve or return

    Let the accountable role decide the next state.

  5. 05

    Execute and record

    Perform only the approved action and retain its history.

Accountable authority

Match the control to the consequence.

The organisation defines which decisions remain human, which actions can be delegated and which conditions always trigger escalation.

Worker responsibility

Define authority in operational terms.

Controls are assigned to roles, actions, case types and risk conditions rather than buried in prompt instructions.

Role authority

Specify which accountable role can approve each decision or action.

Action limits

Constrain systems, operations, values and time windows available to the worker.

Approval context

Present evidence, reasoning, uncertainty and the proposed action together.

Exception routing

Send unfamiliar or high-risk cases to the right specialist with context intact.

Accountable role

Human judgement and approval

Decisions with material, regulated or professional consequences stay with the named accountable role.

Named accountable roles
Case and risk-based gates
Time-bound delegated authority
Separation of preparation and approval
Fallback and escalation routes
Recorded approval and action history
Control performance

Measure whether oversight is timely, informed and effective.

Good control prevents unauthorised action without turning every routine case into manual review.

  • 01Approval turnaround
  • 02Escalation precision
  • 03Unauthorised actions
  • 04Returned cases
  • 05Context completeness
  • 06Override rate

Define the boundary

Bring one process where authority cannot be ambiguous.

We will map its roles, approval gates, action limits and escalation conditions before any automation is introduced.