Meridian's ops team wants to automate two different things: (1) copying passenger rebooking data from an old reservation system into a new one whenever a flight is cancelled, and (2) predicting which flights are likely to be cancelled next week due to weather. A junior PM says both should be built as 'unattended bots.' Where is this wrong?

CPMAI Task 2 requires distinguishing RPA (including attended vs. unattended bots) from AI, and this scenario is designed to test exactly that boundary. Task one — copying data between systems whenever a cancellation event fires — is rule-based, repetitive, and requires no learning or judgment, making it a good fit for an unattended RPA bot that runs on its own trigger without a human present. Task two — predicting next week's cancellation-likely flights from weather and operational patterns — requires a model that has learned from historical data to generalize to new situations, which is a predictive-analytics AI pattern, not RPA at all; RPA bots do not 'predict,' they follow scripted steps. The 'both are correctly RPA' distractor makes the common CPMAI-tested mistake of treating 'any task that used to be manual' as automatically RPA, erasing the real distinction between rule-following automation and learning-based prediction. The 'neither should be automated, both out of scope' distractor is simply false — both are legitimate, in-scope CPMAI topics. The 'reverse the assignment' distractor keeps both tasks mislabeled as RPA/attended-vs-unattended, missing that the second task isn't RPA in any form. Getting this right prevents Meridian from trying to solve a forecasting problem with a workflow-automation tool that cannot do it.