Thoughts on AI for real-time transaction monitoring in emerging markets?

Been pondering the efficacy of AI-driven real-time transaction monitoring, especially when operating in certain emerging markets. The data can be so messy and fragmented, and traditional rules-based systems often throw up a ton of false positives, or worse, miss subtle patterns. Has anyone seen genuinely effective implementations for spotting AML red flags where the transaction volume and data quality present significant hurdles? Curious about what tech stacks are actually cutting through the noise.