BOUNDED AGENTICINTELLIGENCE

Free 5-minute self-assessment

Is your agentic AI pilot
ready to earn authority?

Answer ten practical questions about ownership, permissions, human approval, data access, evidence, and failure handling. You will receive an immediate readiness band and a focused next step.

No account required. Your answers stay in your browser and are not transmitted or stored.

The governing idea

Capability is not
permission.

A model can be technically capable and still lack legitimate authority to act. A responsible pilot makes the boundaries visible before deployment: what the system may do, what it must never do, who approves consequential actions, and what evidence proves the controls operated.

01Is one named business owner accountable for the workflow and its outcomes?
02Have you documented the exact actions the AI may perform and the actions it is prohibited from taking?
03Are consequential decisions assigned to an authorized person with a recorded approval?
04Is the AI limited to the minimum data, tools, and credentials needed for its task?
05Can you identify which person, service account, model, and tool performed each material action?
06Will the workflow retain inputs, outputs, approvals, exceptions, and tool results as reviewable evidence?
07Does the design define when low confidence, missing data, or conflicting evidence must stop or escalate the work?
08Have you tested denied actions, bypass attempts, tool failures, and unsafe or ambiguous inputs?
09Are baseline measures and pilot acceptance criteria defined before the AI is introduced?
10Can an authorized person pause the workflow, revoke access, and return to a safe manual process?

What strong readiness looks like

Clear boundaries.
Reviewable evidence.

Authority is explicit

Permissions, prohibited actions, approval points, and escalation paths are written into the workflow.

Evidence is designed in

The pilot produces records that let a reviewer reconstruct what happened and why.

Expansion is earned

Broader authority follows measured performance and control evidence, not capability demonstrations alone.

Need a deeper review?

Turn one workflow into
a governable pilot plan.

BAI's Workflow & AI Authority Assessment maps the work, clarifies boundaries, identifies control gaps, and defines the evidence needed for a go, revise, or stop decision.

Review the assessment

Begin with a general, non-confidential description of the workflow.