Balancing Data Context and AI Scale with Sastry Durvasula of TIAA

In this episode, Sastry Durvasula, Chief Operating, Information and Digital Officer of TIAA, shares how the company leverages AI responsibly at a massive scale. He discusses the importance of data context, building empathetic customer experiences and preparing for a future where humans and AI work together.

‍

Key Takeaways:

‍

00:00 Introduction.

03:09 TIAA manages vast data across retirement, wealth and asset management.

08:00 AI strategy rests on four pillars: tech, client experience, operations, and services.

16:10 Human service agents handle complex participant needs with AI support.

21:42 Operating officers must blend operational expertise with AI and technology.

25:13 AI solutions help protect older adults from fraud and cognitive decline risks.

30:25 Complaints come through every channel and require timely resolution.

34:56 Beyond semantics, the real focus is on building the workforce of the future with AI.

38:55 AI agents shift economies by changing how consumers shop and services operate.

‍

Resources Mentioned:

‍

TIAA website

TIAA Institute research

‍

Sastry: [00:00:00] my job as it's defined today and my job in like three, five years from now, they'll be dramatically different. Right. So I, I think that's the other thing, like what we do today will be very different from, what we will do in the future.

‍

So I think workflows of the future. I think it's very important. 

‍

Anthony: Welcome to Data Masters. Today we're asking a big question beyond the hype. What does it really mean to be an AI driven company? And how do you do it when you're a hundred year old [00:01:00] institution responsible for the retirement security of millions? Our guest has the real world answers. Sory Dala is the chief Operating Information and Digital Officer for TIAA where he leads the charge on technology and operations.

‍

He's a true innovator. With deep experience in making transformation happen at massive scale. In our conversation today, sory is gonna ground us in what it really matters, what he calls the data context of ai. We'll explore the double-edged sword of having a century of data. How it's an incredible asset, but also a hidden liability, if not managed perfectly.

‍

And we'll talk about the hard, messy problems like making sense of 40 years of contracts for a single customer. More importantly, we're gonna talk about how this technology connects with our [00:02:00] humanity. We'll discuss the profound responsibility of using data to protect vulnerable clients and explore a fascinating concept, sory pioneering called the Empathy Agent a way to make or a way to use AI to make customer interactions more human.

‍

Not less. It's a discussion I guarantee will change the way you think about enterprise ai. Sory, thanks so much for joining us. 

‍

Sastry: Pleasure to be here, Anthony. 

‍

Anthony: So Sory in our planning call as we were thinking about this podcast you framed the conversation about AI around this really, I thought, really useful phrase, the data context of ai.

‍

And maybe we can start there and have you explain what you mean by that, TIAA is a obviously a hundred year old organization. Rich in deep, rich data. And so, I really can't think of anywhere better to think about data [00:03:00] context and ai, but maybe if, if you start us on what you think the phrase means and how you think it's more important than the model itself.

‍

Sastry: Yeah. I mean, data, obviously, you know, as a hundred year old company, that was formed in 1918, we have a lot of data and we have many businesses. We have three primary businesses, retirement, wealth and advice and asset management. And our asset management business is a global business. Retirement is more US centric.

‍

We have global operations. So you think about the breadth of the data that we have and the depth of the data we have. Coupled with the regulatory demands, we have, the privacy demands, we have the global regulations that we have to comply with. It's a pretty complex data ecosystem , and then the underlying technologies for this data has also generations of technology, all the way from the original mainframes that we're hosting this data to like the most contemporary data in [00:04:00] motion on realtime nodes with edge computing and everything in between.

‍

So. To me, you know, the old saying that we are all familiar with, which is garbage in, garbage out holds true. But in the context of ai, it's almost like garbage in, hallucinations out. Yeah. Garbage in, bad actions out, garbage in, scalable damage out. so the data quality problem and the context of data in terms of ai.

‍

I think it's probably 10 x more important than any other problem we had before, because if we don't manage the data properly, and if we don't have a good data ecosystem, at least when you have humans in the loop in some of these systems, as humans could at least smell out that data if you see it in a report or something like that.

‍

But if you just ask the agent. Agent AI system to handle that and you trust the keys with it, [00:05:00] the damage could be pretty bad. So I think that's on the I guess the cautious side of the equation. But then on the positive side of the equation, what better way to put good quality of data to use at scale?

‍

Right? Like, if we were making decisions based on samples of data as humans or systems were training their. Previous generations of machine learning models, et cetera, on limited compute and limited data because of various constraints. And now you can actually apply it to the entire corpus of data that we have in a company.

‍

And coupled with external data sources and solve interesting problems at scale for our customers and businesses, I think that's far more impactful. So I feel like, you know, in the. Data context of ai, we are gonna really test the boundaries on both sides which is great. 

‍

Anthony: Yeah. And I, I think another thing that people don't often think about is that a lot of these models [00:06:00] were trained on data from the internet.

‍

And so they know lots of very general things, but they don't know the specific things. They don't know. Your customer, your contracts, your relationship with that customer. And I think the opportunity to bring that context into these models is probably the number one theme I hear today. Say this repeated theme of like, how do I get context into these powerful models?

‍

Sastry: hundred percent I think, because, you know, I mean the, I mean, we know this, this is a big debate going on right now in the industry where, there's a lot of questions on what data have these models been trained on, and. Are we gonna rewrite the history of the data purpose? I mean, there's a lot of interesting debates that are happening.

‍

But at the end of the day, when models get to be available at scale, like what we are seeing now and what we will see, and they get more commoditized for enterprises to use. Then how you use it and what you train it on to [00:07:00] customize for your own business is where the real impact will come. And that's where I think the data context is far more important because then you're really training it for your specific customers and business needs.

‍

Right. And that's where I think the context is very important. 

‍

Anthony: no. Totally agreed. And now sort of following on that for a second, you are also responsible for operations at TIA. So you're not just wearing a data hat, but you're actually making these tools available. You're making it happen. With users.

‍

And so, I, I think of this as like, not just theoretical ai, but like real AI that needs to transform the way people work. Before we dig into the, the details of, of exactly what your, the use cases from a ground rules perspective, sort of table stakes, what do you think you need to have in place as an organization before you even start thinking about deploying AI at scale?

‍

Clearly good data. That's a good start. 

‍

Sastry: [00:08:00] Yeah. You know, I've got four legs of the stool as I call it, in my organization. And in my role I have global technology which obviously is one of the pillars for AI to begin with, right? Because the tech stack and where you're hosting these models and how you're computing them and how you're integrating with the rest of the ecosystem, that's one, one pillar.

‍

Second is our client experience. So the digital capabilities that we have, how do we define the end-to-end client experience, be it institutional clients that we serve, or individual clients. So essentially B2B, B2C, B two, B2C. Then I've got global operations across all three businesses. So how we actually operationalize some of these capabilities and how they're touching our end customer.

‍

And then lastly, the shared services organization. That's the fourth leg of the stool where. That's basically the middle office, the back office, a lot of analytics functions that support our global business services and capabilities around the [00:09:00] globe that will be AI powered. So if you now think about the four legs and you apply AI as you think about solving business growth challenges or creating delightful client experiences or driving new AI powered business models, sort of scaling operational impact and customer impact with the power of ai.

‍

For changing our own corporate functions with the power of ai, be it people, legal, hr, I mean, uh, risk and compliance or uh, internal audit, et cetera. It comes to this central organization that I have the privilege of leading. So AI for this organization is not theory, right? So it has to really, whatever we build in ai.

‍

We are the ones who are gonna be frankly, bearing the brunt of it if we didn't do it, if we didn't do a good job, because we are the ones using it on the calls as we receive from our customers or with institutional partners that we interface with. So [00:10:00] it's actually a unique role in that sense.

‍

Anthony and I really like it because we get to cook it and we get to eat it ourselves and make sure that it's the right food. So with that in mind. If I think about the types of use cases that we are trying to solve with the power of ai, we have basically focused in on all like three aspects.

‍

One is powering the business growth. How do we power business growth with ai? And how do we enable the next generation of impact for our retirement asset management and wealth management businesses? two is fueling innovation. Like what kind of. Business model changes or shared services, changes that we could be driving for the firm.

‍

And three, strengthening the core. So obviously there's a lot of opportunity, in driving a level of efficiencies that we've not realized, with the power of ai, this is like as big as the change to electricity or change to the internet, or change to the smartphone, right? It's that big of [00:11:00] a change that we experience.

‍

Anthony: Yeah, I think this goes to a a point I've made on the podcast before. In my humble opinion there really only three reasons that an organization invests in data and technology. It's either to drive the top line to save money or to reduce some asymmetric risk in the business which I think mapped very nicely to, to the points Yeah.

‍

That you made. Just to connect to ideas. TIA has almost certainly an enormous amount of data, not the least of which is that you've been around for a long time, but also, really at its core data is your business. And you made a point before about the. Your, the, the people that you work with that you're leading are really good at sort of like sniffing at data and saying, this doesn't seem right to me, and so I'm gonna maybe double check it, or I'm gonna go crosscheck with a different resource.

‍

And AI just doesn't do that. AI will just run with whatever it's handed. [00:12:00] And I'll extend something you said before a, a little bit. I think the, the danger with AI is not only that it runs with this stuff, but that it's trained. To seem as credible as possible with it. It's trained to, in a way be good to the, to its user by seeming confident and smart and thoughtful and, representing some piece of data as true, even if it's quite literally hallucinated or made up.

‍

But share your thoughts on this. 'cause, like, I'm curious how you've if you've seen this and, and how you struggle with it. 

‍

Sastry: Yeah, I mean, AI is definitely as confident as it gets, right? You go to any of the AI agents that we use and uh, the responses, you know, I mean, each ai agent that we use or, chat, GPT, like, options we have, I feel like they all have their own personas.

‍

Like. Plot could be a little bit more descriptive and emotional. And Chad g PT could be very confident and [00:13:00] encouraging. you know, you can like put your own persona tags on these, but they're all very confident in their responses. So, which is good because that's how humans we derive confidence and we would use that.

‍

So the adoption is actually driven by a level of confidence that you can, you can get from the ai. Then there is the hallucinations. I feel like, yes, we will throw humans at the problem in the beginning. Anthony and I do think that we are gonna have humans in the loop for the foreseeable future to make sure that the AI is just not hallucinating and portraying this false confidence.

‍

And then we are making bad decisions. But I do think that there will be a point soon where AI agents will be given that responsibility to do that on ai. So there's like a degree of separation. It's almost like how many cops would you employ on the freeways for speed checks versus putting speed cameras to do that?

‍

Yeah. I mean, there will be [00:14:00] its own issues but I think that we will have to evolve as, as, as the society evolves around these, where there will be layers of. AI systems and agents that will be checking on each other and creating a level of ecosystem because I don't think humans alone could scalably solve this hallucination issue and problem.

‍

So we are gonna have to get better at the algos and models, no question about it. We are gonna have to get better at the data quality issues, no question about it. But we are gonna have to also get better at checking and putting the right checks and balances. I think that some of that would be done by AI itself at some point.

‍

So yeah, and we are, I, in the beginning we have invested through our T ventures in a couple of firms that actually do this. So obviously it's early stage, but we are quite encouraged by the optionality that's like emerging right now. And we will just integrate that into our broader ecosystem at TI.[00:15:00] 

‍

Anthony: And I think the point you're making, you used the word scale a a a couple times there, and I think this is a really important concept. And again, in your role running operations, I think, I'm sure you think a lot about, the way, typically, or historically, the way you would scale an operation is bring more people on, you know, just hire more people.

‍

And I'm, I have to think that that has changed and is changing. And the example you shared previously on our prep call was an agent having to look through. 10 different contracts that you might have with a client that span over 40 years. That is an extraordinarily research challenge that a single human being might work on for days, if not, a week.

‍

And, using AI to. Quickly and scalably solve that problem makes the agent actually more efficient. And it's something that you [00:16:00] literally couldn't do unless you hired banks and banks and people. Is that the right way to think about it? 

‍

Sastry: Yeah. And that is a real scale problem solved with ai, right?

‍

And you still have the beauty of that use case. Anthony, we have a human in the loop. I mean, frankly, human is in the frontline here, right? So think about my client services, operations folks that are getting calls from. Millions of participants. And for those of you who are listening to this who don't know what TIA does, TIA was founded by Andrew Carnegie in 1918 to build retirement solutions and offer lifetime income solutions for higher education for healthcare and nonprofit sector.

‍

Now we managed close to $1.4 trillion of assets. And so if you think about the servicing aspect. Of our client services professionals, you know, a participant in our language, which is basically a customer who has a contract with us is calling in. Now, this could be a professor from one of the universities that has, let's say, [00:17:00] changed jobs four times and has done some research as well for a healthcare segment.

‍

And then they came back and now they're retired. Now they're calling with a question. The complexity of the contracts we have with all these institutions and the individual contracts contractual clauses that are there, it's one person at the end of the day that's calling us, so we have to treat that one person as one person.

‍

But the customer service professional has to sift through a lot of contracts and answer that question, so. Now what happens usually without this level of AI capability is we will say, yep, we'll research through our middle office function. We'll get back or, or we keep the person on hold and then we go through, now put AI in the mix.

‍

You got an AI agent that's gonna like go through these contracts, has the ability to come up with recommendations. This is real data. this is not like some, artificial data. This is real [00:18:00] data. And then it comes with synthesis on these contracts. You still have the human and you'd run with it, and then you're able to service the customer.

‍

It's a great classic use case of this generative AI technology, if you think about it, right? Because it is going through reams of data, synthesizing it, creating real actionable insights for a human that is serving a human in more real time, in more personalized, in, more customized in more, expeditious way from a servicing point of view. So I feel like that's a great example of, the use of AI. And uh, same thing. Like we do the same thing on the institutional side in our asset management business where we have to research. So our portfolio management teams research a lot of companies and a lot of data, right?

‍

So we have this research buddy, AI agent for our asset management business. Our asset management business is called Moen. So we've deployed it. Same thing. When you think about institutional insights that we need to [00:19:00] provide or portfolio decisions that we have to make in our investments from an asset management point of view, same thing.

‍

Lots of data, lots of documents, lots of unstructured documents that we have to analyze and come up with insights quickly. That level of scale. Is, I think what makes it really encouraging to have the level of capabilities that we're being to see with, with generative ai. 

‍

[00:20:00] 

‍

Anthony: So in this way, I mean, this is the real connection between your role as a as an operator and this AI and data context.

‍

Again, I think historically a, a chief operating officer would think about solving any scaling problem by scaling operations. Again, add a, a re more people to the research team and, or we're gonna add more people to the call center so we can handle these complex queries. And and.

‍

Now you've been given this ability to handle these problems using a technology approach and, and using a different scaling factor. I wanna also connect to something you'd said before. The population that TIAA works most closely with are retirees, people who are I would imagine that the kinds of calls you're getting are from somebody who has retired and has a question about their retirement plans and, and their relationship with, with TIA gr The challenge of course is this is an aging population. And, [00:21:00] and you brought up this question in our, prep call where. Many of these people are experiencing some level of cognitive decline. They, they may be confused about their plan. And you and, and your agents have a fiduciary responsibility to help guide that person to, to a good answer.

‍

This brings up a whole series of difficult regulatory complexities, ethical complexities. That's a really difficult problem. How are you applying AI to that? Challenge and problem. 

‍

Sastry: Yeah, I think, before I answer the question, you made an important point, which I think I should just double click on for a few seconds, Anthony, which is the role of the operating officer itself.

‍

Because if you look at the operating officer role, you have either people in these roles that have grown through operating roles who now have to. Learn ai, learn technology to a different degree, to put use of that [00:22:00] into the day-to-day operations. Right? Of course. Then there are people like me who have actually grown in those who have taken the scale jobs of running operations.

‍

So, so I think, either way you look at it, you're gonna have to learn the other side. So for me, frankly I feel kind of blessed to have grown in this path of technology, data, and ai. Now leading large scale operations because I could easily, conceive some of these ideas more because I've been trained to do that.

‍

But of course, we have to learn the operations of how does retirement services operations work. Like I've worked for financial services all my life, but it's a different segment of financial services, right? So I think at the end of the day, there's a lot of learning involved. It's quite an opportunity for, leaders who have large scale jobs like mine, who have operational responsibilities in addition to strategic functions.

‍

Now, coming to the use case that you mentioned, it's a great use case. It's one that Tia is uniquely positioned along with maybe others like [00:23:00] us. we serve generations of. Customers, right. Or participants in our language. So we have Gen Zs that are beginning to, like, start, accumulating for the retirement.

‍

Who may not call us who are very well-versed with digital medium, and they consume insights through social media and they make decisions based on that. So that's, there is that aspect of it. And they maybe using different phrases. I have teenagers at home, I hear all these different phrases like, huh, skip you know, and then if you have kids, so we have to converse with them and influence them to save for their retirement.

‍

And that's not the first thing, that's not the 10th thing even that they're thinking about. So there's that, that side. And then there is. The middle of like tenure participants who have been accumulating and who are now making d decisions based on where they are with their tenure and career and life, et cetera.

‍

And then you have [00:24:00] retired folks who now are Decumulating, who are getting their lifetime income paychecks from TIA, who may not wanna be on these digital channels, who may not want to go to their device, who may wanna just talk to somebody. So if that's the spectrum and you're serving those customers, each generation and each participant brings a different set of parameters on how we serve these customers.

‍

and you know, you could say, well, isn't that true for any business? Because every business has, these type of parameters, but decisions on what to buy in retail. Decisions on how to use your payment product are dramatically different from decisions when it comes to your own future nest of retirement money.

‍

The only thing that in my mind comes close to this is health. And health and wealth are like the most critical things from a decisioning point of [00:25:00] view. So when you now get into the phase of like, where we serve participants who are retired and. We have, America definitely has longevity increasing, which is a good problem to have, which is a great thing to have.

‍

But with that we have more older adults who are more prone to cognitive decline, who are more vulnerable for fraud, financial fraud who are less savvy with some of these, cyber protection techniques. That's the population. Now we need help, right? So, we've helped them all along to save for their retirement, but now how do you protect their financial wellbeing that we've now been working on AI solutions for that set of problems, right?

‍

Because if you think about cognitive decline can you put AI into work to help our participant, to prompt. Or to help the customer service professional in [00:26:00] handling these conversations, in a more impactful way or to engage their trusted contact. Because, older adults especially with cognitive decline, may have trusted contacts.

‍

So how do you engage them? How do you put AI in the middle of this to protect them from fraud actors? Somebody who's like impersonating as a government agent. like I'm from IRS and I'm calling you sir. And then, how do you respond to it? Or romance scams that are on the rise because of, senior like older adults with loneliness and there's all kinds of issues there with financial fraud.

‍

So I believe that this is an area where TIA make can make a huge impact. In fact, we just published some extensive research on this. We've done through our TIA institute a lot of research on cognitive decline and, financial protection and financial fraud prevention. So I'm personally passionate about it which is why I'm talking at [00:27:00] length on this use case because I feel like AI will significantly help.

‍

Of course we have to use it, carefully. We have to use it with all the privacy rules and regulations. We have to make sure that it's governed. We have to make sure it's checked and everything. But I think it's a good use of AI in terms of financial stability of older, older adults.

‍

Anthony: and you've, you mentioned well, there, there's actually a a fun podcast that I sometimes will listen to called money, death, money and taxes. I wanna say something close to that. And the idea behind the podcast is these are taboo topics like no one wants to talk about. Yeah.

‍

But it feels like, your agents are having conversations about death, taxes, and money. That's, those are really the only conversations they're having. It's all one of those three categories. So these must be extraordinarily difficult conversations. Not only do you need to. to our prior conversation, be concerned with the caller on the other end.

‍

Are they understanding? Are they tracking and following? Are they the [00:28:00] right person to talk to? Yeah. But you also need to do so in a way which is highly empathetic, which understands their needs and requirements. And it would seem to me that AI. Would automatically get in the way of that, that would be almost like a, a layer in between, like would turn you into a kind of robot answering these questions in a way that is, factually accurate, but devoid of any understanding of the context of the conversation.

‍

Somebody calls in. With a spouse that's passed away needing to understand benefits, and you're treating them like they're, doing an ATM withdrawal, that's not gonna go well. you've thought this through and I, I'd love you to share the way you think about how to actually have AI help with empathy.

‍

Sastry: Yeah, I mean, it's a great it's a great question and a, great framing of the problem statement, right, Anthony? Because. You know, Until now, I'll be honest. my view of AI is exactly what you just described, which is like what we've seen in the Hollywood movies, right? It's [00:29:00] like, or what we've experienced with chatbots.

‍

They don't have any empathy to say the least. I mean, do you just ask a question and then you get the response? 

‍

Anthony: Almost like you wanna prioritize speed, like you gimme a question, I'm gonna give you that exactly answer really fast. Right? 

‍

Sastry: So that has been my view, but off it, it's changed quite a bit actually because we've seen now some good use cases on empathy.

‍

And I'll give you the classic example of what, where we've deployed it quite successfully. So we built, first of all, you know, in our ecosystem, we built a platform called TIA Gate. Originally we started as generative AI technology gate, and now we. Just use the same acronym, but we are now using it as generative agent intelligence technology.

‍

And we put that in the hands of our colleagues through what we call my gate. So every colleague, basically, we have the ambition to put my gate in the hands of every colleague in the company. And we've been rolling it out to thousands of colleagues, By end of summer we'll have at least 5,000 colleagues [00:30:00] with, with migrate in their hands.

‍

So now they can actually use it for like citizen innovation essentially. So one of those innovations after we put migrate in the hands of our colleagues in our complaints department was this empathy, which is very powerful. and just to set the context, we get like every financial services institution or any institution, as a matter of fact, we get complaints.

‍

We get thousands of complaints, and the way we handle these complaints is a complaint comes in, it goes to the complaint. So the complaint could come in through our customer service division, or it could be coming through our online channels, through our mobile app. It could be as simple as somebody sending an email to me directly saying, I'm frustrated.

‍

I've done all these things, blah, blah, blah. So anything, or it could be in actual snail mail, which we do get by the way. Like actual handwritten letters as well. So everything is accepted. so a lot of research has to go [00:31:00] to handle this complaint and then to address it. And we have obviously timelines then, we have regulatory windows.

‍

The minute you start a complaint and we wanna make sure that, we, we, we handle this complaint and close it. So in the middle of all of that the complaint handling team. Their priority number one is to handle the complaint and do all the necessary homework and research, and then with facts, and then handle it properly.

‍

So it's hard to also be very empathetic in the middle of all of that when you have thousands of complaints and this small team has to handle all of that and all that stuff. so in a typical construct. Would've, as you said before, we would've thrown more people at the problem, and then it would still not fully scale because of the volume of some of these and the complexity of it.

‍

So now we've put AI, empathy agent for this team, and what it does is it's, it's more like a, an assistant for our complaint handling team for each person. Essentially you have a digital AI [00:32:00] assistant, let's call it the copilot for the lack of a better word. And it's giving them recommendations on empathetic correspondences.

‍

Or if they have written, like if I'm the complaint handler and I'm sending you a participant, a correspondence empathy agent is right there saying, Sastri, you may want to think about this this way because considering the context of this participant's issue, this is not the right tone. How cool is that?

‍

Right? I mean, like, one way to think about it is really, machines are now gonna teach us empathy as humans. The other way to think about it is they can assist us in the middle of complex processing to ensure that we don't lose the human side of us in our correspondences. I know it's counterintuitive, but I think it's a very, very impactful use case from an AI helping humans, assisting humans in frankly being better humans.

‍

Anthony: Yeah, love it because I, what makes an agent more empathetic is. Being [00:33:00] able to pull together an answer quickly. Yeah. And then having that check to say, what state is the person I'm talking to in? And maybe they're in a place where yes, they're filing a complaint, but what they need is resolution quickly because there's a challenge.

‍

And what they just want is an answer. And, or maybe what they want is just to be heard and to, having their feedback sort of responded to in a way more carefully, more thoughtfully. Is a way to have, have them acknowledge that, they've been heard.

‍

And in that sense, a system can really help guide an agent in, into pro providing that.

‍

Sastry: And I just to add to that, and we just, one point, it also helps us in training. Mm-hmm. Right? So, when we have like colleagues we hire a lot of interns. We have, we have, we hire a lot of people that are like.

‍

From schools that are going through these programs. And we train them. So it almost brings a level of uniformity to the way we communicate as well. And frankly gives them more empowerment with time [00:34:00] as they see this. So I think it's also a good training tool. Uh, Of course we are gonna use it in our training departments as well, but this real time production set of use cases, right.

‍

Anthony: No, that's a great point. And again a lot of this I'm sure comes through repetition and practice and and again, that can be really helpful. You've made this distinction between AI and you've mentioned a few times a agentic ai. I think this is. Maybe even an area where I struggle to clearly define the difference.

‍

And, and maybe in that spirit, if you don't mind, kind of casting your eye forward where you see TIA investing where do you see this going? Is everything moving to an agentic system? What is an ag agentic system and how would it differentiate from a traditional AI system? I'd love to get your view of where things are going.

‍

Yeah. 

‍

Sastry: Yeah, I mean, there's a lot of debate as we know. Like even like semantics matter a lot. There's like this religious debate on, is it [00:35:00] AI agents or agentic AI are the same? Are they different? What's the definition? And you open your LinkedIn, there's like articles on articles, on articles from all these influencers.

‍

At the end of the day, I believe that if you take all this semantics and all the jargon out. There is a workforce of the future in the making right now and how fast we will evolve into it. It depends on how fast this technology is evolving and the ecosystems are evolving and frankly how fast the companies are moving.

‍

that's kind of like almost the premise of the problem that we are in or the opportunity we are in. Whichever way you look at it . So workforce of the future and the workforce of the future will have people and agent AI and people and machines will have to work together, collaborate and we still have to figure out what that means and what it means to businesses, what [00:36:00] it means to business models, what it means to customer service, what it means to growth opportunities and.

‍

I mean, if you step back and if you think about the last big technological revolution like this go back to 2007 and Steve Jobs showing the iPhone and that moment, and we at that point knew a little bit about apps, stores, and apps and all that, and what it would mean and all that. Nobody at that moment predicted there would be this big gig economy, there would be this big sharing economy.

‍

There would be this, there would be that. And now here we are like moving, hundreds of millions of dollars through mobile phone transactions. And then people are doing gigs. And I feel like it's too early to kind of predict because it's moving so fast. But with a workforce of the future in mind, I think there is businesses of the future.

‍

That kind of come with it with this power of workforce of the future. Because, and I, this is a [00:37:00] classic example for comparison, right? You take a company of a hundred thousand people on one side that has been inventing or reinventing itself with ai, and then you take a company that has similar value proposition, but the composition of the company is thousand people and 10,000 agents.

‍

Well then. How do you actually compete in the market if both have similar set of products and capabilities, but one has the composition of agents that are working 24 7 essentially, that don't have the same emotional, power. But then you have people that are working their hours, but they bring a lot more emotional capital with them, but they don't have the scale of the technology agents that are running on the other side.

‍

Right. So I think. It's gonna be very interesting, to be honest. You know how this thing will evolve. So my first theme is Workforce of the Future. the second theme I would say is workflows of the future. so whatever the composition of the [00:38:00] company will be for any company that will finalize itself, with time what they work on is gonna be dramatically different. Like if I'm a marketing content manager or if I'm a developer, or if I'm a legal analyst, or if I'm a research analyst in our equities market, my job as it's defined today and my job in like three, five years from now, they'll be dramatically different. Right. So I, I think that's the other thing, like what we do today will be very different from, what we will do in the future.

‍

So I think workflows of the future. I think it's very important. And then of course, third is really, so what, so what is, what does that really mean to these big companies? Yes, there will be startups just like how we've seen in internet and social media and, remember this solo mode, social, mobile, local uh, ecosystems that we're surfacing.

‍

I'm sure we'll see different set of ecosystems that are coming up. We are already seeing that with ai, [00:39:00] AI powered economy, like agents doing the shopping. So if agents are talking to agents and then doing the shopping for you, then the whole emotional appeal of advertising goes away, right? Like we go to the website or app and we see this nice advertising for a jacket or a bag, and then we are like, oh yeah, that looks good.

‍

Maybe we should try it. But then if I have a shopping agent that does that for me, the shopping agent has no empathy towards that because it's ai. It's looking for value or brand or whatever the parameters are. And it's a very different thing. Or, professional services are changing as we speak, like ad agencies.

‍

There's, there's big thesis on it. Consulting, there's a big thesis on it. So I feel like, workforce of the future, workflows of the future, business models of the future as a result are the three like big problems or opportunities that the world will be contemplating with. And it's great to be part of this.

‍

Mega [00:40:00] revolution, you know, it's great to have a front seat, being part of a company of our stature and heritage to kind of shape it, from within with our teams here. 

‍

Anthony: I think that's a very useful framing the workforce of the future, the workflows of the future, and then how that impacts the business model.

‍

And what I really appreciate is it's clear, TIAA will be on the cutting edge of that. So, look, really appreciate you making the time to join and sharing those insights. And thanks for joining us on Data Masters. 

‍

Sastry: Thanks Anthony. Thanks for having me.