AI around the world

Thanks to the Evident AI symposium, I had the opportunity to learn about how leaders around the world are doing AI. Here are 3 leaders and few stories they shared with me that could help the broader community.

Tony Kim , Silicon Valley, USA

1. AI is eating software. Traditionally, we've needed a lot of software to solve business problems. With the advent of Big Data and now Gen AI to explore/exploit it we can replace quite a bit of software with shorter user journeys. Whereas the business problem itself might not change (book a flight, buy a shoe) the steps to get there are getting fewer.

2. Agentic frameworks while not perfect are showing promise for investment research, christened project 'Asimov', he and his team are exploring its use for reducing the upfront grunt work of collecting/collating data.

Dan Jermyn , Australia

1. To Notify or Not. They found their customers were using a 3rd party ATM to draw money which cost them vs free at their ATM. So they used AI and sent notifications to those people to say don't so it. Surprisingly, customers increased their spending at 3rd party ATMs! Their behaviour science team reasoned why. Customers didn't like to be told what to do, especially by a bank, they also didn't like to be told at inopportune times. E.g. they sent notifications when some customers were out having a good time and when they saw the bank was telling them what to do, enraged, they did the exact opposite. So, they changed the tack to just inform rather than prescribe how to not act, but also when to notify and whom to notify and whether to notify at all.
2. Doing it even when the ROI isn't there. When they launched a feature to allow their customers to send each other money and message each other about it, they noticed that some customers sent abusive texts and NSFW commentary. They applied AI to solve it by using graph neural networks which involved detecting, flagging and preventing abusive behaviour and texting. Doing the right thing was right for business.

Marco Li Mandri , Netherlands

ING measured their automation success using a single KPI called as Straight Through Processing (STP) between 0 (0 automation) and 1 (complete Straight Through Processing). This measures the amount of digital interactions a customer has with ING. Even if a single aspect of the interaction involves a manual step, then the KPI does not achieve a value of 1 but a number such as .95 or .99. While strict, adherence to such high standards and actioning them has enabled them to serve customer needs more seamlessly.

There was so much more that was discussed, but my hands just couldnt type faster. Thanks much to the team that put together this event and invited such eminent panelists to share their learnings with the world. Will be back for the next one.