Captain Obvious chiming in here, but as one of those boring douche canoes whose concerns are data privacy, data locality and generally the ability to understand how AI trains off of and manipulates the data we feed it, I'm resistant to the tech. We're testing the use of AI to aggregate and explain patterns in the data we have, but this is limited to our ticketing systems and Slack. Until our current choice of AI provider can guarantee that our transactional data won't be stored outside of the EU, the people driving this internally and those keen to make the sale externally can take a long walk off of a short pier. I can almost smell the aroma of the coffee machine from the chamber in which an EU subcommittee is working on the Artificial Intelligence Act (AIA) (EU) 2025/xxxx to add to the raft of regulations that financial entities in Europe need to concern themselves with. Maybe they'll be nice and just append it to an existing act.
AI might be great. AI might be terrible. I'm not all convinced that most data aggregation features baked into AI and used by most normal companies couldn't be implemented in R or SQL (disclaimer: I couldn't code my way out of a wet paper bag if the tool I was told to use was an axe). It's just wanted so that someone can crawl over data sets to ask simple questions like 'how many merchants exceeded n number of transactions between a and b date' or 'My customer needs an eIDAS certificate. What do I ask them to send us without having to talk to Captain Obvious?'. I mean, we're busting a gut on revamping our developer docs, but given that the spec is already public, I'm pretty sure that developers can already vibe code against that. Going to test that and see how it gets on.
AI might be great. AI might be terrible. I'm not all convinced that most data aggregation features baked into AI and used by most normal companies couldn't be implemented in R or SQL (disclaimer: I couldn't code my way out of a wet paper bag if the tool I was told to use was an axe). It's just wanted so that someone can crawl over data sets to ask simple questions like 'how many merchants exceeded n number of transactions between a and b date' or 'My customer needs an eIDAS certificate. What do I ask them to send us without having to talk to Captain Obvious?'. I mean, we're busting a gut on revamping our developer docs, but given that the spec is already public, I'm pretty sure that developers can already vibe code against that. Going to test that and see how it gets on.