Choose the right
process to automate.
A useful starting point is a repeated task with a clear owner, consistent inputs and a result you can check. The task causing the most frustration is worth examining, but it is not automatically the best first automation.
1. Describe the work before choosing a tool
Follow one recent example from start to finish. Write down what triggered it, where information came from, who handled it and what counted as complete. Include the work people do outside the official system, such as copying details into a spreadsheet or chasing a missing approval.
2. Establish a baseline
Count how often the task happens and measure the active time needed for a representative sample. Record corrections, waiting time and the proportion of cases that take an unusual route. A weekly estimate is a starting assumption, not a verified saving.
3. Check whether rules are enough
If a task follows explicit conditions, ordinary automation may be easier to test and maintain. AI may help interpret variable text or prepare a draft, but its outputs need appropriate review. Do not add AI merely because it is available.
4. Compare candidates
- Frequency: is there enough repeated work to justify a change?
- Consistency: are the inputs and desired outputs reasonably stable?
- Access: can the required systems and data be used appropriately?
- Exceptions: can the team recognise when the workflow should stop?
- Ownership: is someone responsible for reviewing and maintaining it?
- Consequences: can errors be caught before they cause harm or irreversible actions?
Use these questions to compare options, not as a universal numerical score. A frequent task can still be a poor first pilot if mistakes are costly and difficult to detect.
5. Pilot a narrow slice
Choose a bounded task, a reviewer and an agreed test set. For example, prepare suggested categories for incoming requests while staff retain responsibility for routing. Include incomplete and ambiguous examples. Run the pilot alongside the existing process until its usefulness is demonstrated.
6. Measure the complete result
Compare preparation time, review time, correction effort, failures and ongoing costs. Time freed for other work is valuable, but it is not automatically a reduction in payroll or cash expenditure. Agree what benefit matters to the business before expanding.
Turn the choice into a trial plan
Use the free AI pilot planner to compare preparation and review time. If you have several ideas and need help choosing, start with an AI assessment.
A useful first conversation
Bring one redacted example, the systems used, a rough volume and the person who owns the work. Astir can help map the problem and scope an appropriate next step.
Business process improvement · AI workflow implementation · What affects cost?
Discuss a repeated task ↗