Anatomy of an Agentic Transaction
Alex Johnson · OCT 14
Welcome back to the Fintech Takes podcast.Copy anchor linkCopied
I’m Alex Johnson, joined once again by Dan Murphy (partner at Allon Advisors and former CFPB official who worked on the Section 1033 open banking rule) to ask some annoying questions about agentic finance.
We start in Toronto, at the FDATA Global Summit, where AI agents effectively hijacked an open banking conference and screen-scraping veterans began to recognize a familiar fight.
From there, we follow the flow of a transaction as it might go through an AI agent, checkpoint by checkpoint.
Highlights include:Copy anchor linkCopied
- If we already struggle with KYB, how will we verify agents that can be spun up by the thousands and killed once they stop being useful — especially without a national digital identity system?
- Who decides what an agent can actually do? Dan walks through frameworks in the UK, Australia, and Brazil, and confesses he once wrote off variable recurring payments as just subscriptions. (They are not.)
- Dan’s hot take on the L word: whoever asks where liability lies is usually the one holding it and trying to get out of it. And why consumer trust has to come before the liability fights, not after.
- When an agent knows more than the person it represents, whose interests does it serve? What can fiduciary law teach us about a technology built around swarms of agents?
The agentic economy will land in financial services first. How these questions get answered in the industry will determine what agents do in healthcare, education, and beyond.
Big thanks to Ocrolus for making the conversation possible.
