Business
The service problem is measurable, but the pilot case still needs an accountable sponsor and validated benefit assumptions.
Public assessment example
Follow a customer-support idea from clarification and evidence review through four specialist assessments, one public-guidance review and a combined result for a human decision.
Starting point
“We want an AI assistant to sort customer support enquiries and draft replies using our approved help content and order information. A support agent must review and send every reply.”
The assessment asks focused questions, records the supplied evidence and requires the customer to confirm the facts and boundaries before specialist work starts.
How it works
This public view condenses the complete assessment journey. It does not require sign-in, save information or use a free assessment.
Enter a short description or use the Starter Pack to prepare the basic information with the proposer and reviewers.
Resolve missing or conflicting facts, then confirm the use case, boundaries, ownership and evidence before assessment begins.
Identify unsafe inputs, unresolved blockers and the guidance profiles available for the assessment.
See the assessment scope and required stages before the specialist reviews start.
Examine separate Business, People, Technology, and Risks and oversight findings, with evidence gaps and limitations kept visible.
Bring the saved results and guidance review together as decision support, conditions and practical next actions for accountable people.
Specialist results
Each result is saved separately so its findings, evidence and limitations remain visible in the final report.
The service problem is measurable, but the pilot case still needs an accountable sponsor and validated benefit assumptions.
Human control is clear, but the design has not yet been tested with support staff or customers who may be more affected by mistakes.
A read-only prototype is plausible, but evidence for secure retrieval, quality thresholds and reliable operation is incomplete.
Keep the assistant internal until privacy, escalation, challenge and monitoring controls are demonstrated.
Separate guidance review
National AI Centre · Published 5 May 2026 · Voluntary public guidance
The concept includes useful human-control foundations, but evidence for accountability, impact assessment, transparency, testing and incident management is incomplete.
The source, version and findings stay separate and traceable. The review is not legal advice, certification, approval or proof of compliance.
View the authoritative public sourceCombined result
The app does not approve the use case. It brings the saved results together and leaves the next decision with accountable people.
What you can inspect now
This page exposes the flow, specialist views, guidance review, conditions and human-decision boundary without requiring an account.
Assessment report
A real AI-produced assessment report will be added here after its content and presentation have been checked for public release.
Report download coming shortlyStart with the information you already have, or use the free Starter Pack to prepare with the proposer and reviewers first.