Regulatory change management is the standing process a bank, insurer or other regulated firm runs to keep its rulebook current. It starts with horizon scanning, reading what every relevant regulator publishes, and continues through an applicability decision (does this rule bind this entity, licence and business line), obligation extraction (what specifically does it require), mapping to the policies and controls that carry each obligation, and remediation of any gap before the effective date.
The volume is what makes it hard. Thomson Reuters counted more than 61,000 regulatory events in a single year across 1,374 regulators, and most compliance teams still run the process in spreadsheets and inboxes. The expensive failures are rarely a rule nobody knew about; they are a known rule that never reached the right control, or a control that drifted from the rule while nobody was comparing the two.
AI fits the reading half of the job: triage, extraction, mapping and drafting, each output carrying the clause it came from. It does not fit the interpretation half. The regulator holds the firm to the interpretation whoever produced it, so the applicability decision stays with a named person, with the source text open in front of them.