Your customer has clearly an action-video level of understanding of what hacking actually is. An operation as complex as overriding an AI-controlled droid cannot be conducted ad hoc via random and chaotic gesticulation over a terminal. Hacking a hostile droid is usually achieved by planting a virus inside a "consumable" data packet - like an update package or through tapping into an existing data stream like a communication transmission. Hacking a captured and immobilized droid is about unlocking its service mode and modifying its software. There is also a crucial difference between working with a human built AI and the self-rewriting and self-redesigning AI often found in the deep Omicrons. Finally, there is a tremendous difference between hacking an isolated unit and a hive-conscious artificial being.
Your customer has to provide us with a detailed explanation of the use case scenario and desired effects. We need to know whether the software is intended as a mean of repeated electronic warfare or just a way of performing a one-time infiltration. We need to know as much as possible about the target and whether it is possible to capture and disable one unit for the reverse engineering process. Finally, your client should be made aware that repeated use of the same algorithm will soon become ineffective as the AI will develop its own safeguards.
P.S. I hope your client understands that a dedicated software will cost him a fortune.