“Sibos 2026 showed an industry moving from asking whether AI works to working out how to run it safely at scale.” 

This Appledore Research note examines how data and AI dominated Sibos 2026, Swift’s annual forum for payments, securities, trade and treasury, held this year in Miami. Banking prizes stability and low error tolerance; AI moves in months and behaves probabilistically. That tension ran through the whole event.

We review what BNY, BNP Paribas, Citi, Deutsche Bank and Bank of America said about agents already running in production, and how vendors including Accenture, AWS, Cognizant, Databricks, FICO, FIS, Fiserv, Google Cloud, Microsoft, NTT DATA, Quantexa, SAS, Smartstream and Temenos positioned their data and AI propositions. We assess where the industry stands on data quality, control architectures and agentic commerce.

We also give our candid view of what was missing from the conversation: the ontology layer that turns raw data into something AI can reliably act on.

This note is for bank technology, operations and data leaders, and the vendors who serve them, who want an independent read on where AI in banking is genuinely heading.