There’s tons of research out there for all elements of ransomware incidents - eg Mandiant, Sophos etc publish yearly IR reports from real incidents.

Generative AI and AI malware don’t feature on any of them.

Initial access barely uses phishing nowadays.