The situation
Harbour Lane Accounting wants to use AI to prepare client follow-up drafts without handing decisions to a system.
Lesson 1 of 12 · 7-minute lesson
Follow one AI use case from an initial idea to an accountable decision, owned actions and review.
For: AI leads, accountable executives and people evaluating the product
This lesson uses National AI Centre guidance: Planning for AI, Essential AI practices.
The situation
Harbour Lane Accounting wants to use AI to prepare client follow-up drafts without handing decisions to a system.
The problem
The idea touches client information, staff responsibilities and supplier controls. A promising idea is not yet a safe or accountable use case.
The choice
Maya creates one shared AI use case and connects its policy, preparation, assessment and follow-up before testing begins.
What you will learn
The complete journey
Moment 1
What the person does
Maya records the purpose, limits and owner.
What the workspace records
One shared record brings preparation, assessment and follow-up together.
Notice
The proposal still needs preparation.
Why it matters
The team shares one starting point instead of running an informal experiment.

Moment 2
What the person does
The team answers seven screening questions.
What the workspace records
Answers identify areas needing attention without creating a score.
Notice
Screening is an early check, not an approval.
Why it matters
Relevant preparation work becomes visible.

Moment 3
What the person does
Maya assigns relevant activities and owners.
What the workspace records
The plan connects each activity to its reason and responsible person.
Notice
Not every tool is required for every use case.
Why it matters
The organisation does only the preparation this use case needs.

Moment 4
What the person does
Maya reviews which records are ready to include in the assessment.
What the workspace records
The workspace shows what is included and why anything is left out.
Notice
An attached file is not automatically reliable evidence.
Why it matters
A reviewer can see what informed the assessment.

Moment 5
What the person does
Maya prepares a proposed decision from four separate reports for the accountable decision-maker.
What the workspace records
The proposal records the reasons, uncertainty and conditions.
Notice
A proposed limited test is not approval for testing or live client use.
Why it matters
The organisation can see what is proposed and what must happen before approval.

Moment 6
What the person does
Actions, evidence needs and review triggers return to the shared record.
What the workspace records
Owners and the evidence needed to close each action stay connected to the proposed decision.
Notice
Completing a task does not prove that a safeguard works.
Why it matters
Governance continues while the proposal moves through approval and testing.

Official guidance used in this lesson
National AI Centre
Frames the practical work used to understand an opportunity and prepare an organisation.
Last checked 30 September 2026
National AI Centre
Provides the six practices used to organise safe and responsible AI use.
Last checked 30 September 2026
The complication
Meaningful human review is appropriate, but the team has not yet demonstrated that people can detect, reject and escalate unsuitable drafts.
The outcome
Harbour Lane records a proposed limited test using agreed sample data, people and stop rules. It remains unapproved until the named conditions are met and the evidence is reviewed.
Check your understanding
No. Screening identifies attention areas. Preparation, assessment and an accountable decision still follow.
Key points
What happens next