The useful question is how much time a particular workflow could release after review, correction and ongoing costs. Start with a baseline rather than an industry average or a promised number of hours.
Record the work before changing it
Choose a typical week and log repeated tasks: the trigger, person involved, input, output, time taken and any rework. Separate hands-on time from waiting time. A shorter response delay may improve service even when the same amount of staff time is required.
Look for repeated drafting, document preparation, enquiry triage and reporting. Note exceptions and identify information that should not be shared with an AI service.
Choose a bounded first pilot
Pick one task with a clear definition of acceptable quality and a person responsible for approval. Use controlled samples first. Compare the manual process with the proposed assisted process, including checking and correcting the output.
- Email: draft a reply from an approved outline; check context and commitments.
- Meetings: turn approved notes into actions; confirm owners and deadlines.
- Marketing: create alternatives from a brief; check facts and brand standards.
- Reporting: summarise approved figures; reconcile the summary to its source.
Separate capacity from cash
Time released can create room for other work. It does not automatically reduce payroll or increase revenue. Include implementation, subscriptions, human review and maintenance when estimating the business case. Use the Opportunity Calculator to compare assumptions, then validate them through a pilot.
Make the scale decision with evidence
Track time per task, review changes, exceptions and output quality. Record the sample size and period measured. If review effort cancels the saving, change the workflow or stop. If the result is useful and repeatable, document it and train the team before expanding.
Try a sample in the Experience Lab or begin with an AI readiness assessment.

