The Zwicky Transient Facility (ZTF), a state-of-the-art optical robotic sky\nsurvey, registers on the order of a million transient events - such as\nsupernova explosions, changes in brightness of variable sources, or moving\nobject detections - every clear night, and generates associated real-time\nalerts. We present Alert-Classifying Artificial Intelligence (ACAI), an\nopen-source deep-learning framework for the phenomenological classification of\nZTF alerts. ACAI uses a set of five binary classifiers to characterize objects\nwhich, in combination with the auxiliary/contextual event information available\nfrom alert brokers, provides a powerful tool for alert stream filtering\ntailored to different science cases, including early identification of\nsupernova-like and anomalous transient events. We report on the performance of\nACAI during the first months of deployment in a production setting.\n