1. Confirm that the operation is suitable
- The activity repeats often enough and has a clear output.
- Volume, duration, defects, downtime and manual interventions are measurable.
- Exceptions are known: obstacles, unusual loads, people in the area and emergency modes.
- Automation will not preserve an unnecessary or poorly designed step.
- A business process owner is accountable for acceptance.
2. Establish the baseline
Measure cycle time, throughput, operating cost, quality, safety exposure and current availability. Use comparable observation windows before and during the pilot. Without a baseline, apparent improvement may be caused by seasonality, a different team or a change in demand.
3. Prepare the site and integrations
- Validate wireless coverage, power, lighting, markings and floor condition.
- Define sources of tasks and confirmation: WMS, MES, ERP, service desk or operator.
- Create test identities with minimum privileges.
- Approve zones, safe speeds, stopping, manual control and evacuation rules.
- Assign ownership for maps, configuration, charging, updates and spares.
4. Use operational metrics
Measure completed missions, availability, cycle time, first-pass success, interventions per hundred tasks, recovery time and safe stops. Record false detections, connection loss, route replanning and the time required to resume work.
Express the target as a range with a minimum observation volume. One successful shift does not prove repeatability.
5. Test failures deliberately
Exercise an obstacle, network loss, low battery, unavailable upstream system, malformed task, manual stop and post-update recovery. The team should know what the robot does automatically, what the operator sees and who authorizes resumption.
6. Decide by pre-agreed rules
Define four outcomes before the pilot: scale, improve and repeat, retain in a constrained area, or stop. Scaling requires safety, quality, economics, integration and support criteria to pass together. The deliverable is not a presentation; it is an evidence pack and a list of changes for the next deployment.