The payment investigations series

6 pieces on payment exceptions and investigations — what they cost, why richer ISO 20022 data makes the work harder rather than easier, and what an agent can resolve with a person still deciding. Ordered as an argument, not by date.

  1. Agentic AI for faster E&I investigations

    When a cross-border payment fails, reconstructing it spans several banks and several messages. A demo of TxAgent investigating a beneficiary address exception.

  2. Ready to let clients go?

    Payment investigations still cost 3%+ customer attrition, 5–8 days to resolve and an estimated $1.6bn a year across the industry.

  3. Investigations won’t keep pace with ISO 20022.

    Investigations won’t keep pace with ISO 20022. At least not as-is.

  4. AI in fraud and financial crime is no longer experimental

    An ACI Worldwide and Finextra survey: 51% of organisations are live with AI, only 19% operate with full autonomy, and the blockers are structural rather than technical.

  5. The cover payment is missing: from camt.110 to camt.111

    A real camt.110 investigation request traced to its root cause: the cover pacs.009 was never sent. With the corrective message and the camt.111 closure drafted.

  6. Chat with your Swift data!

    TxAgent, our AI demonstrator for interbanking data: fewer false positives, faster analytics, and business users querying Swift messages in plain language.

Everything else, including company news, is on news and articles.