Cyber, AI, and identity: Elevating secure digital banking journeys [34:17-39:22]

in which whether it's for wealth customers or even wholesale customers, the amount of information and the way that that agent can be used to help a human service that customer far better is where we are focused a lot of our time with our clients right now. And we're seeing a lot of the benefit from it. I think to your question on how to make sure identity is really woven into that is certainly making sure that whatever however we are using the information within the FI that it is actually tied to that customer. I mean, certainly as you can imagine for a wholesale customer that has probably 30 different products and services as you are using an agent to gather data around transaction usage or finding out why a payment is stuck going to a beneficiary. You for that human to then use that information there has to be multiple checks in place that that customer data that is that the agent is using is actually for the customer for the account because we know that those wholesale customers have multiple accounts hierarchies and so forth with it. So I think that's really where identity trust has to play a role even if that's not an agent working directly with the customer but an agent helping a service a human service person inform that customer better because even if misinformation happens there it's it's a it can be a really poor experience for that client. Thank you, Michelle. But each you would like to share your thoughts on this. Yes, sure. You know, let me take the view from again going back to the fraud landscape right. If you ask me about the underestimated fraud risk right now within as a genetic AI scales, you know, firstly I would say is the abuse of tools and privileges right today you may have agents that have access to payments KYC information case management systems data. You know these agents can be manipulated via prompt injection or compromised connections and and now turned into this high-speed fraud channel where information is just pushed out going back into data extiltration right agents can inadvertently also disclose sensitive data whether it's summaries ticket information going into a stirred party you know look up or some connections are there where workflow spans different vendors APIs there could be a lot of sensitive information that could be manipulated and taken. This is a new concept also about agent versus agent where fraudsters and and this is happening criminals are using Jenny AI right now to find these deep fake IDs make more these social engineering more convincing and put a lot of pressure on you know the verification whatever's happening at the call centers fraudsters are using these type of agents to make things more convincing and also you know errors and feedback loop poisoning when I say feedback loop poisoning what it means is you know as AI is being used not having you know without strong controls you can teach the system bad things right and this is a reality but it's good to have again some good governance in place for AI and and this is evolving right you know for years we've always talked about model risk governance and now it's getting into this agent governance how are you controlling these agents what are the tools what are the permissions how are they allowed to operate like what is there a box that they're going to be under and and you know Michelle also talked about this is humans humans are important but when do you bring that human into the loop right and now started bringing in more risk-based analysis over there to say if I'm going to govern my AI I'm going to have some risk-based human interactions also and have this ongoing validation take place in terms of looking at the AI again going back to some recent examples let me pick two quick ones one is a g-server really large bank globally this they've claimed significant adoption greater than 75% of their employees using some of some sort of AI agent internally and another metric they've put out is over 50% reduction in IT called center or desk tickets right so a lot of it is internally being used we're also seeing AI agents being used in case management in terms of detecting emerging fraud themes helping case workers so so this is some real practical uses over there absolutely thank