A campaign sends each reminder, and the agent handles what people write back.
Send the run one way or the other, on a condition you write.
Sort the message into one of the options you name, and take that path.
Send what has been gathered to a model and take its answer.
Have a second model check a sensitive action against your policy before it happens.
Record a result in one of your systems, and keep trying until it lands.
Tag the conversation with an outcome, such as Paid or Wrong number.
Ask a person on your team to take over, tell the visitor, and stop.
The campaign sends the reminder. Each reply is sorted, and a second model checks a promise against your policy before it reaches your system.
Tell us what it should handle, and where.