Choose one workflow before choosing AI.
Define a testable AI opportunity in planning, deployment, service delivery or customer care.
Audeamus Group · Practical guidance · 21 October 2026
Start with a recurring piece of work
Write down the trigger, the person doing the work and the output the next person needs. Candidate workflows might include organizing planning inputs, assembling a deployment work pack or preparing a draft customer-care response. These are illustrative examples, not client case studies or claims of completed deployments.
Establish a baseline
Sample completed work before introducing a tool. Record volume, handling time, rework and the checks required. Keep the definitions consistent during a pilot, including handoffs between planning, field teams and service or care teams.
Work through the operating constraints
Identify permitted data, exceptions, responsibilities and any outputs that require human review. Resolve unclear data permissions before testing. Choose stop criteria that fit the workflow, and make service quality part of the assessment.
Leave with evidence for the next decision
A workflow map, test set and measurement record should show what changed and what remains unknown. Potential hours released are a planning estimate until observed against a baseline. Include review and correction effort before treating a time estimate as an operating benefit.