A (3:51)

Can you talk a little bit about how you maybe are using AI to personalize services, customer experiences, handling complaints through automation, intelligent automation in the back office, or maybe even advanced fraud protection being in financial services?

B (4:05)

Yeah, so the way I think about our whole technology strategy and that basically trickles down to the AI strategy is a three legged tool. So we want to power the business growth, we want to fuel innovation and we want to transform how we operate or strengthen the core. And so AI is being implemented in all three legs. And in fact our investment framework we used for return on Investment and CBAs on these use cases that we come across through the process we have, it basically looks at these legs and the impact that these use cases are making. So for instance, in the way we serve our customers in delighting them in our business lines as we power the growth agenda in our asset management management, we've implemented an AI powered agentic solution called Research Buddy. And you know, if you think about our asset management industry, it's a B2B industry. We're working with institutional, you know, institutions in our institutional investment partnerships. And we need to do a lot of research. It's hundreds if not thousands of hours of research every week that our analysts have to do, our portfolio management teams have to do. Right. And it's across a lot of documents, publicly available, privately held documents. And I mean talk about the power of generative AI and agentic AI. Now we have now transformed that whole function to the delight of our institutional clients. Same thing if you go back to our retirement services business where we serve our participants, which is individuals or it's a B2C construct essentially, or institutional clients, which is a B2B client. The way we Handle these contracts. You could be a professor who has changed jobs six times, for instance. And every job you had, you may have had multiple contracts with us. Now you're calling in, you've retired, and you're availing your lifetime income paycheck, which is the promise of TIA's fundamental mission. And our reps have to go through all this information to serve you. Now put generative AI in the middle of that. So we have created AI solutions on the desktop of our colleagues, and they can get insights through the AI solution to serve you better in a hyper personalized way. So that's that. And then of course, talking about your fraud, you know, we have like what we call cyber for AI and AI for cyber fraud for AI and AI for fraud. So essentially, how do we protect our customers using AI from a cybersecurity and fraud perspective, but also how do we use cyber capabilities to protect us from AI itself? So one actually good example would be cognitive decline. Lucas. So we have a lot of our participants are aging participants. Longevity is actually on the rise, which is a good problem. But with longevity, there is cognitive decline. So with cognitive decline, there is extreme vulnerability to fraud and cyber breaches.

A (7:17)

Right.

B (7:18)

So how do we assist our participants, you know, from these AI attacks or fraudsters or cyber attacks, using the power of AI? So we've been actually, you know, doing a lot of work in this space. We published some research as part of our institute studies, and we've some patents in this space on how we can leverage these capabilities to help those that are suffering cognitive decline.

A (7:44)

Can you give me an example of how it might work?

B (7:46)

Yeah. So, for instance, if there is a fraudster calling a participant and trying to do like an impersonation fraud, like acting as the IRS agent or an investment agent or an investment, our own advisor, it could be, you know, they could be impersonating our own advisors. Right, right. And AI detects that. It could be a fraud signal. And then this capability kicks in to the point where we could even engage now the trusted contacts. So the participant may have like a trusted contact, which is. Could be their. Their kids or their grandkids or their niece or their friends or their lawyers. We could activate the threat with them. We basically look for those signals and actually intervene in the middle of it, really. And then we engage our fraud prevention unit, obviously, through this process. So that AI is doing the job before fraud analyst is engaged and alongside the fraud analyst as we go through the case. So it's, you know, you can see like deep fakes are on the Rise impersonation, attacks are on the rise.

A (8:55)

This is a big deal. That's the issue.

B (8:57)

Right. Don't know who, major issue. And you know, there's a lot of studies on this topic on how many, you know, hundreds of millions of dollars are lost in the industry, especially in financial services. That's why, you know, we've been quite focused. We're partnering with other nonprofits as well in this sector. We've been doing partnerships with universities to do advanced research in this space. And you know, we just published our marquee research on this through our institute as well.

A (9:22)

I mean, to me, that's fascinating. Using AI to combat AI.

B (9:26)

That's right. And you know, the other thing also is it's a great use case of AI. I mean, we always talk about, well, AI is now taking cognitive workloads, what's going to happen to white collar jobs, et cetera. But how cool would it be to use the cognitive power of AI to help those that are suffering cognitive decline? Right, right. I mean, the other use case, which is a very counterintuitive use case, is the empathy agent we built. So if you asked me two years ago when we did this conversation, like, dude, would you think that AI would ever actually help humans in creating empathy? I would have said absolutely no. Right.

A (10:04)

It's a machine.

B (10:05)

It's a machine at the end of the day. But you know, we do a lot of these cases where complaints come in. You know, you mentioned the word complaints earlier. And then we have to research a lot of these complaints. It's very complicated. So the, the reps that are assisting these complaints want to get the transaction done as quickly as possible. Operationally, they don't have the time to write empathetic communications in that process. Right. We have some templates, et cetera. But on the receiving end of this, a participant is reaching out because they have some major financial hardship. Maybe they're trying to avail their funds because they have some sickness in the family. And so we need a real empathetic communication in the middle of this regulatory driven SLAs we have. And so we've created this empathy agent and it actually drafts the communications, it even edits the communications and it gives you prompts on what to say during the conversation. So it's early stage, but it's promising from what I can see.

A (11:10)

I mean, it'd be nice to see a financial services organization be more empathetic. So that's, I think that's a great,

B (11:16)

especially in our space. Right. I mean, we're dealing with you know, retired participants on one spectrum and then like Gen Z's who are just beginning their retirement on the other side who have an entirely different set of focus areas, you know.

A (11:30)

Yeah, you mentioned the word regulatory. Healthcare, financial services, maybe legal, transportation, those are the big three that are highly regulated. How are you dealing with the crossroads of deploying such a new technology as artificial intelligence while still being regulatorily compliant?

B (11:52)

Yeah, so we created a responsible AI policy. Obviously we are highly regulated from various entities. Responsible AI policy tries to exceed the regulatory requirements we have because there is also new regulations, especially in the global community, because in our asset management business we are global retirement services is United States obviously. So we have regulatory privacy requirements that we have to cater to. But more importantly, the security itself biases in our decision making capabilities. Privacy, set of requirements that we have, data security, network security. So all these things have to be incorporated. So we have a very comprehensive responsible AI policy that we hold every use case and every capability development to that standard. But more importantly, I think with the advancements now we have in AI, some of this policy is also codified so to the point where during runtime responsible AI agents can catch to see if what's being done runtime adheres to the policy or not. So it's not just a enterprise architecture standard that's sitting out there that we expect every project to kind of review. That's like 1.0. We are probably in 3.5 stage right now as like technologies. Right. So I expect the responsible AI agents to monitor for proper deployment of AI in our workforce, in our work streams.

A (13:25)

You've mentioned agents several times now and there's a bit of a conundrum in that even people selling agents aren't really selling agents, they're selling applications that act like agents.

B (13:38)

Yeah.

A (13:40)

How do you differentiate the two? And how would you describe your agents?

B (13:44)

Yeah, I mean it's, you know, agent could be, it could have a million forms, you know, in a definitional state. And right now there is this like those whole. And you probably see in the studies like how many agents should a company have? Right, That's a big question. Like there is a school of thinking that oh, we need to have thousands of agents, you know, that will basically do the workload of any workers and you know, augmenting and helping them or replacing some of these jobs potentially on one side of the spectrum, the other side of the spectrum says, well, you just need like a few super agents. And some firms have come forward and said that you don't need like thousands of agents. I think it still has to evolve, frankly, at this point, Lukas. And from my perspective, you're right. Some of these applications are really all the way from APIs to autonomous agents that are fully self sufficient. We have a few that I would really characterize as the true agent that could do most of the work end to end. But we are still in the early stages as the rest of the world is. There is a level of human element for the right reasons, to be honest. Yeah. So I would say, you know, regardless of the number of agents that these companies claim, whether the producers are consumers of these agents, I think the plot is still in the making. It's really the impact that it comes down to and how much of autonomous function that an agent is able to actually do. Right.

A (15:17)

And how willing are you to hand off those functions?

B (15:19)

How willing are you to hand off to that in a governed way at this juncture?

A (15:23)

So how are you aligning your AI initiatives with the overall business strategy?

B (15:28)

I mean, the business strategy for TIA is very simple. Leading lifetime income, delight our clients and strengthen how we operate. And that's what Tishonda and the ec, we've worked with the whole organization and declared at the beginning of our journey as a team like three and a half, four years ago. That's the strategy for the company, that's the strategy for technology, digital operations, shared services, my organization. That's the strategy for AI. So which means that AI has to drive business growth. For instance, we, for the first time in the history of tia, have transitioned to a model where our product travels. So our product, the marquee product in retirement is called 403B market, which is basically healthcare, education, higher education, nonprofit sector. Just recently we moved that product to be available in the 401k market, which is corporate retirement. In fact, one of the CIO hundred awards that actually TIA received this year, my team is here is for the retirement gateway we built so that it's called TIA Gateway. And the reason why we built the platform is to make sure that our product travels. So as the product travels to a much wider market, we are now talking about a $3 trillion market in the 403B space to a 9 plus trillion dollar market in the 401K space. And our products being available, I believe AI will play a significant role in that journey for us. Same thing as we expand our global market penetration in our asset management space as we target to multiple trillions of dollars in assets under management and administration. So same thing, AI will play a role then of course, delighting our Clients. We've talked about a few examples, obviously in this conversation and frankly, strengthening how we operate. My mantra has been I want AI to be available in the hands of every colleague in the company. So we built this platform called gate, Generative Agentic Intelligence Technology and we've created a wrapper on it. We call it the MyGate. So MyGate is now available to half of the workforce.

A (17:45)

Okay.

B (17:46)

And they basically, it's like their version of call it. It's not ChatGPT, it's multiple LLMs that we've embedded in it all the way from the clouds to OpenAI's to Metas of the world. We've added agents into that spectrum as well. So it's one stop shop for our colleagues. And I fundamentally believe that that's going to transform how we operate as a firm and drive efficiencies and frankly, release time for our colleagues to focus on upskilling and reskilling.

A (18:16)

Last question, I'll make it fun. Be personal or professional. What is a tech gadget that you really love and you've been having fun with or an application for that matter?

B (18:27)

I mean, right now, if you ask me, it's vibe coding.

A (18:29)

Okay.

B (18:30)

I mean, initially I was like, yeah, right. You know, I mean, I'm a hardcore coder growing up, so I'm like, yeah, right, AI is going to do my job now. I mean, of course I don't do development anymore. My team won't approve it, but I've been wipe coding quite a bit, you know, using rapid or cursor at home and just playing with it. I mean, the speed with which you can actually do some of these cool development tasks. As an engineer at heart, yeah, I think it's quite fascinating. And frankly, building agents, I mean, I believe that, you know, it's a massive change. Everybody has a lot of anxiety in their heads, what's going to happen to my job? How difficult is it? Should I be really AI equipped? How do I get AI equipped? It's almost the equivalent of the shift people I'm sure made when electricity came in and jobs changed from like mechanization to electrification. And a lot of blue collar jobs became white collar or blue, white, white, blue collar jobs. Massive change back then, like around 1960 in the United States. That's what we're going through right now. So I think, you know, getting as hands on as possible, I would say is probably the most easy way to like get a hang of the technology itself.

A (19:45)

I mean, most of the studies that I've read shows that it's going to create more jobs than eliminates.

B (19:50)

I, that's my, I'm a utopian believer. You know, when it comes to tech transformation that has been the case for all the previous industrial revolutions. I mean I have my 8020 rule that 80% of the jobs will have at least 20% impact with AI and 20% of jobs will have as much as 80% impact with AI. So you know, as a ex McKinsey consultant I can come up with these hypotheses but we'll see. I believe that there's going to be a better state for the humanity overall but not without a lot of these issues and risks that we're going to

A (20:25)

have to work through some of the kinks. But let the machine do the grunt work.

B (20:29)

Yeah, exactly.

A (20:30)

Awesome. Sastry, I can't thank you enough. Again, great insights on this stuff. Always a pleasure to talk to you.

B (20:37)

Same here. Lucas, great to see you again.