- Published
- 8 October 2026
- Last reviewed
- 8 October 2026
Describe the existing process
Write what starts the work, who performs each step and what marks its completion. Use a small process that you can explain clearly. Automating a poorly understood sequence can reproduce confusion more quickly.
The fictional example is moving a practice enquiry from an incoming tray into a review queue. It uses invented records and does not send messages, change live data or make decisions about real people.
Identify inputs and rules
Record the fields the process requires and the rule for routing each case. Explain what happens when information is absent, duplicated or inconsistent. Distinguish fixed rules from judgement that needs a responsible reviewer.
For the fictional exercise, a topic category might select a review queue. A missing category should reach an exception queue rather than disappearing. Do not assume every enquiry fits the normal path.
Design the checkpoint and exception path
Identify where a person needs to review or approve the next step. Explain what information they will see and what they can change. Record who owns failures and how a stopped process will be noticed.
Consider repeated triggers and interrupted runs. The practice map should state how the team will recognise an item already handled and how it will investigate an incomplete run. These are planning questions, not a claim that one workbook implements a reliable system.
Test a fictional walkthrough
Walk through a normal case, a missing-field case and a repeated item. Record expected and observed behaviour. Check that the final outcome and any exception are visible to the responsible person.
Download the workbook in PDF or Word. Finish with the approval needed, questions for the implementation team and a next review. If AI is part of the proposed workflow, also define how generated content and tool actions will be checked before wider use.
Source and practice note
The UK Government AI Playbook provides context for human control when AI is included in a workflow. The mapping exercise also applies to rule-based automation. The fictional examples and exercises are original RDE learning material.
Cover image: AI-generated illustrative scene created specifically for this resource.
Download and complete the exercise
Download the editable Word document
Printable PDF and editable Word document. Version 2.0, 8 October 2026. Personal and internal organisational adaptation permitted with Jefferson Otonbara Imgbi author credit retained.