TRANSCRIPCIÓN 🔗
WEBVTT
00:00:16.800 --> 00:00:22.320
Good morning and welcome to CFO Coffee Talk, brought to you by the Florida CFO Group.
00:00:22.559 --> 00:00:27.920
This morning's episode is titled Enterprise Semantic Continuity.
00:00:28.239 --> 00:00:31.359
Sounds like a real Hollywood blockbuster, doesn't it, Natish?
00:00:32.560 --> 00:00:38.560
Our special guest today is one of the tech leaders at Raymond James, Natish Gudapuri.
00:00:38.880 --> 00:00:39.759
Welcome, Natish.
00:00:40.159 --> 00:00:40.799
Thank you, Phil.
00:00:40.960 --> 00:00:42.159
Thanks for having me.
00:00:42.479 --> 00:00:46.479
You know, uh I saw you speak at uh AI Salon.
00:00:46.719 --> 00:00:59.840
And uh before we dive into it, uh let's let me give a shout out to that wonderful uh networking event uh they have every six to seven weeks ago or so, uh, sponsored by the Florida CFO group also.
00:01:00.320 --> 00:01:06.640
Uh and uh the next event's coming up in uh on September 23rd at Arc Innovation Labs.
00:01:06.799 --> 00:01:11.120
Uh go online and register and you'll thank me later because it's a great event.
00:01:11.200 --> 00:01:26.560
It's where I met Natish who was on stage on the panel talking about uh the$1 billion AI budget that uh his uh employer Raymond James has to uh gain competitive advantage in that competitive space of uh investment banking.
00:01:27.200 --> 00:01:29.760
Um there.
00:01:29.840 --> 00:01:34.000
Uh I love to get the uh our audience a chance to get to know you as well.
00:01:34.159 --> 00:01:36.319
Give us a little bit about your backstory.
00:01:36.879 --> 00:01:37.200
Sure.
00:01:37.359 --> 00:01:40.079
Uh yeah, I'm an associate director of technology.
00:01:40.239 --> 00:01:44.640
Right now I take care of uh uh primarily the platform modernizations.
00:01:44.799 --> 00:01:56.879
My work has been largely around uh enterprise core services, architecture, and modernization of uh large-scale mission-critical platforms, uh, particularly in the financial services.
00:01:57.040 --> 00:02:06.879
So over the years I worked across enterprise services, distributed systems, uh, integrations, modernizations, and more recently of enterprise AI.
00:02:07.200 --> 00:02:15.199
So, what became increasingly clear to me is that uh modernization is rarely just a technology problem.
00:02:15.599 --> 00:02:18.240
Organizations can successfully replace the platform.
00:02:18.479 --> 00:02:30.000
So I have learned over the period that uh it's about the process, the rules, the business rules, the relationships, and the intent, while the underlying technology continuously changes as we see.
00:02:30.159 --> 00:02:34.080
So that's the interesting part, and that's where uh I got involved.
00:02:34.800 --> 00:02:35.520
Oh yes.
00:02:35.680 --> 00:02:41.280
Uh and uh I can share a little bit about that tech background myself.
00:02:41.360 --> 00:02:58.159
I started off in the computer mainframe business, uh, brought in as a young, a part of a young team of college grads to
Load full transcript
uh do process re-engineering and system implementation to help a the second largest computer mainframe company in the world at the time, Burroughs, uh, uh modernize its processes.
00:02:58.319 --> 00:03:06.240
Um so we're gonna talk about uh how it's more than technology, it's it's processes and the most important element, people, right?
00:03:06.400 --> 00:03:08.479
Uh people, process technologies.
00:03:08.800 --> 00:03:16.000
Uh I know you from reading what you uh have put on the LinkedIn, really good stuff, by the way.
00:03:16.159 --> 00:03:30.479
Uh we share this mindset, this experience that you can be have a successful technology project with the technology works, but the results still miss the mark from a business perspective.
00:03:30.719 --> 00:03:31.919
Talk about that.
00:03:33.520 --> 00:03:40.879
So it's about the ability of an organization to preserve the consistent business meaning as its technology changes.
00:03:41.120 --> 00:03:48.319
A customer, a trade or a payment entitlement or a business rule may pass through dozens of systems.
00:03:48.479 --> 00:03:51.199
Okay, they have been there for quite a period of time.
00:03:51.360 --> 00:03:55.919
Those applications will change, the infrastructure will change, the integration patterns will change.
00:03:56.080 --> 00:04:00.960
But the meaning of those concepts cannot keep changing every time the technology changes.
00:04:01.120 --> 00:04:05.360
So semantic continuity is all about making that business meaning durable.
00:04:05.520 --> 00:04:10.400
So technology should be replaceable, but the business meaning should always be durable.
00:04:10.800 --> 00:04:11.120
Yes.
00:04:11.360 --> 00:04:19.279
And uh yeah, I what uh really grabbed me was that post on enterprise semantic continuity that rang so true to me.
00:04:19.360 --> 00:04:32.720
Uh, because uh you know my background was with a lot of uh mergers acquisition acquisition integrations, uh and even in the in those situations, it's particularly challenging to get everybody talking the same language.
00:04:32.959 --> 00:04:41.839
And if you don't get that solid foundation of semantic continuity down, uh why do you waste time arguing about uh what does this number mean?
00:04:42.000 --> 00:04:44.000
Uh what's the right way to define it?
00:04:44.079 --> 00:04:46.480
Uh so important for a business.
00:04:46.720 --> 00:04:53.600
Uh how has your experience shaped your perspective on uh enterprise semantic continuity?
00:04:54.639 --> 00:05:02.079
So a lot of thinking came directly from working with the large-scale, long-lived enterprise environments, particularly in the financial services.
00:05:02.240 --> 00:05:06.240
Uh, I've been in the same uh large enterprise for the last close to 10 years now.
00:05:06.480 --> 00:05:15.519
So when you work with mission-critical platforms that are really impactful for the business, you quickly realize that a system is much more than a source code.
00:05:15.680 --> 00:05:30.160
And over time,
these platforms accumulate business rules, operational exceptions, integration behavior, uh, recovery mechanisms, downstream dependencies, and the knowledge that may not even be fully documented anymore.
00:05:30.480 --> 00:05:36.160
So when you modernize them, the hardest question is often not how do we rewrite the application?
00:05:36.319 --> 00:05:45.279
The harder question is how do we understand everything this application does and its surrounding ecosystem that you actually know.
00:05:45.439 --> 00:05:53.040
So I worked in environments where enterprise services connect a large number of business apps and processes extremely high transactional volumes.
00:05:53.279 --> 00:06:01.040
At that scale, even a small semantic misunderstanding can propagate across multiple systems and business processes.
00:06:01.279 --> 00:06:08.639
For me, the hardest part of modernization isn't uh replacing the software, it's preserving what that enterprise has learned.
00:06:09.360 --> 00:06:13.759
Yes, and uh it often is uh a lot about semantics.
00:06:14.079 --> 00:06:28.959
Uh it is uh a key to the foundation of just solid data integrity and data discipline that without which doesn't matter how good your technology is, you'll fit you'll end up as confused as you were before you start the project.
00:06:29.120 --> 00:06:44.959
Uh I uh have been in a lot of discussions where uh we are uh instead of using valuable leadership team time to focus on how we can improve the business and gain competitive edge, we waste time talking about what does this number mean?
00:06:45.040 --> 00:06:47.839
Uh how come you're calling it this and I'm calling it that?
00:06:48.079 --> 00:06:57.519
Uh nailing that down is so important, and uh that's why I'm I'm uh was eager to uh uh shine a light on enterprise semantic continuity.
00:06:57.600 --> 00:07:05.519
It's just like a basic foundational thing for success, especially in the AI world, which is accelerating the pace of change.
00:07:05.920 --> 00:07:06.639
Absolutely.
00:07:07.279 --> 00:07:12.720
You yeah, and I see your post talking about uh architectural convergence.
00:07:12.959 --> 00:07:23.920
And uh I'm not sure I can even uh uh phrase the question right, but I do want to hear you talk about how it all comes together the governance, the data, the security, the identity.
00:07:24.079 --> 00:07:30.480
Uh give us a quick overview of uh uh how businesses should approach architectural convergence.
00:07:31.040 --> 00:07:31.439
Right.
00:07:31.680 --> 00:07:37.040
So today the traditional software largely executes workflows that we explicitly define.
00:07:37.199 --> 00:07:46.800
Whereas when it comes to the when how we interpret the goals, the reason, and the context, select the actions and potentially execute those actions through the enterprise systems.
00:07:47.040 --> 00:07:51.839
So immediately the question rises far beyond uh AI agent models.
00:07:52.000 --> 00:08:07.759
Um for example, an enterprise agent dependence
uh depends on much more of uh intelligence, it depends on trusted data, identity, delegated authority, policy, governance, observability and reliability systems of record.
00:08:08.000 --> 00:08:15.519
So if an organization's business meeting is fragmented across systems, uh AI will inherit that fragmentation.
00:08:15.839 --> 00:08:33.120
When it comes to the enterprise architecture, which is very core to how the capabilities are, uh I think the next generation of enterprise architecture will be different, less by any particular technology and more by a set of durable enterprise capabilities.
00:08:33.440 --> 00:08:59.840
The first question is the enterprise semantics, understanding of code because the concept and all the management, the second is the identity and uh delegated technology, uh the third is the mission-readable policy and governance, the fourth is really the real-time architecture, which means APIs, data products, events, uh, and finally, I think uh observability, which means uh which will expand beyond the infrastructure.
00:09:00.000 --> 00:09:09.039
Enterprises will need to understand not only whether a system is running, but why an agent is AI agent made a decision and what information influenced it.
00:09:09.200 --> 00:09:20.960
I just added AI agent because now the enterprise architectures are thinking with the with the flow of how AI agents can add into their existing workflows.
00:09:21.600 --> 00:09:42.960
You know, uh this whole uh discussion brings up that balancing act between tight governance to uh protect the uh organization from the uh data risks, uh, and uh yet instilling a sense of uh uh exploration and freedom that uh creativity that advances the organization.
00:09:43.440 --> 00:09:59.440
Uh how do organizations like Raymond James uh uh straddle that line and find the sweet spot between uh maintaining control but unleashing the human creativity that really takes the AI potential to its proper level?
00:09:59.840 --> 00:10:11.360
Uh as an organization, we start with the literacy first, and we make sure first the people, the culture is set, the mindset is set to a level of uh what our goals are as an organization.
00:10:11.600 --> 00:10:13.600
And then we build the policies around it.
00:10:13.679 --> 00:10:29.279
We build how we can uh empower the people around it, uh, what are the workflows that we could uh uh we could show, we could uh practice that makes their productivity increase, that builds the efficiency around their day-to-day workflows.
00:10:29.440 --> 00:10:34.639
So these are all the smaller iterative things that we do for an extended period of time.
00:10:34.799 --> 00:10:36.879
So they get used to the system.
00:10:37.039 --> 00:10:40.320
And then we apply the, you know, what does the business need?
00:10:40.480 --> 00:10:46.399
So, as the AI enablement for business needs, uh, the business on the field might require a different thing, right?
00:10:46.480 --> 00:10:55.360
A financial advisor might look for something uh very much whatever he's doing, uh, four hours might be uh can be done in less than 20 minutes.
00:10:55.679 --> 00:11:02.320
So there are some teams that actually work that help with the business, and there are some teams that actually work.
00:11:02.879 --> 00:11:07.279
Uh, in fact, every technology team is now slowly incorporating AI.
00:11:07.519 --> 00:11:20.080
And uh we are building our own studios, we are building our own workflows where you could just go in uh, and your role will define how you can use the AI within a firm.
00:11:20.240 --> 00:11:30.879
If you are an analyst, if you are a developer, you don't need all the tokens or all the work um uh use cases that uh another uh you know designated user can use.
00:11:31.120 --> 00:11:50.639
So we teach them, we train them over the period of time, and we show them what their existing uh PBI or a uh requirements document would look like before and after by using uh AgenTix Studio or you know the in uh the internal workshops that we do.
00:11:50.799 --> 00:11:58.080
So it's all about a continuous process of learning, teaching, and uh slowly applying it into their day-to-day workflows.
00:11:58.720 --> 00:12:14.080
And uh learning together uh has been my experience that we all uh tend to uh learn faster if we find learning partners to collaborate with and uh opens up our eyes to maybe some risks we haven't seen, some opportunities we haven't seen.
00:12:14.320 --> 00:12:20.480
Um being a chief financial officer, it's all about finding that balance between risk and return.
00:12:20.720 --> 00:12:28.879
And uh the whole uh uh field of AI is just so rich with possible great returns, possible big risks as well.
00:12:29.120 --> 00:12:43.519
So uh I love collaborating with folks like you, technical geniuses, who can uh help us uh guide our uh business strategy in a way that uh uh optimizes uh the upside potential uh while protecting our risks.
00:12:45.360 --> 00:12:59.840
So uh, you know, uh it's hard not to ask this question uh uh of uh to folks as bright and forward thinking as you uh what does your crystal ball tell you?
00:13:00.000 --> 00:13:03.279
What's coming down the pike five to ten years from now?
00:13:03.440 --> 00:13:07.600
How will how will uh the world of business be different in your mind?
00:13:11.120 --> 00:13:11.759
Interesting.
00:13:11.919 --> 00:13:16.639
So the competitive advantage won't necessarily be owning the newest technology, okay?
00:13:16.799 --> 00:13:26.480
It will be having an architecture that is capable of adopting, adapting new technology safely, coherently, and quickly.
00:13:26.639 --> 00:13:40.320
So these are all the foundations that we are building right now, and we should be ready for any kind of change because uh three years ago no one was ready for AI, and only since 2023, January, people start talking about it.
00:13:40.559 --> 00:13:48.399
Organizations have started uh uh talking about it and saw what is it for them if they
don't get onto the train.
00:13:48.559 --> 00:13:50.639
So this will always be there.
00:13:50.799 --> 00:13:57.440
So I tell that you need to have a system, a model that you could safely onboard into it.
00:13:57.840 --> 00:13:58.720
Very good.
00:13:58.960 --> 00:14:18.879
Um, and then as a financial guy, uh, one risks that I know is on my mind, time to mind of some of the business owners who uh worry about protecting uh their cash reserves is uh how do you how does uh really smart organizations like Raymond James protect against the uh cash drainage from token usage?
00:14:19.039 --> 00:14:20.960
Uh that's a big unknown, isn't it?
00:14:21.279 --> 00:14:23.440
That's part of the literacy that we are doing.
00:14:23.519 --> 00:14:24.799
So it's an interesting question.
00:14:24.879 --> 00:14:30.879
It all boils down to the the the whole concept is tokenomics, right?
00:14:31.120 --> 00:14:37.840
So it comes to how we should use what are the like I was explaining earlier, the workflows.
00:14:38.000 --> 00:14:40.080
What does that role identify?
00:14:40.320 --> 00:14:44.559
If a person is a business analyst, what does he use them for?
00:14:44.799 --> 00:15:14.240
Then we define them and we know as we train them, teach them the models, we know the exact number of tokens that would be used approximately, but we are trying to uh build those uh fragmented uh generative engineering, AI engineering, uh into our uh studios so that uh those are individually kept, of course, but uh we know a person uh really needs X number of tokens and we allocate them.
00:15:14.399 --> 00:15:24.879
And yes, there will be instances here and there, let's say you are on an innovation challenge, you want to build something in two weeks, you definitely can't survive with uh 400 credits of tokens, right?
00:15:25.039 --> 00:15:31.600
So uh there will be some exceptions as uh the leadership would take and uh you know uncap them.
00:15:31.759 --> 00:15:55.759
But uh by and large, uh the organizations, it's not Raymond James, it's all the large-scale enterprises will think about the cost that will come along the way, and uh they will try to tighten how the how these are being applied for, and then they will cap them out so that uh they really know who needs it and who doesn't need it.
00:15:55.840 --> 00:15:57.360
That's the most important thing.
00:15:57.600 --> 00:16:02.000
Yeah, it's why uh finance and tech leaders need to work so closely together.
00:16:02.080 --> 00:16:06.799
That's uh the new big risk area, but also it's a new upside potential area.
00:16:07.200 --> 00:16:12.320
Uh Latisha, I'm uh glad you could join us this morning.
00:16:12.480 --> 00:16:15.759
Uh looking forward to seeing you at more AI Salon events.
00:16:15.840 --> 00:16:18.159
Will you be at the September 23rd event?
00:16:18.720 --> 00:16:19.519
I will be.
00:16:19.840 --> 00:16:20.320
Very good.
00:16:20.480 --> 00:16:23.759
Well, uh our audience can uh see both of us there.
00:16:23.919 --> 00:16:27.600
And I I urge you to follow Natish on LinkedIn.
00:16:27.759 --> 00:16:31.200
Your posts are just excellent, uh, thought-provoking, informative.
00:16:31.440 --> 00:16:34.399
Uh, I love the quality of what you've been posting on LinkedIn, Natish.
00:16:34.799 --> 00:16:35.360
Thank you, Phil.
00:16:35.519 --> 00:16:40.080
Really appreciate your time and uh opportunity for me uh to be on your podcast.
00:16:40.399 --> 00:16:46.799
Well, Natish, we end each episode uh with the guest leaving us with a one-sentence key takeaway.
00:16:46.960 --> 00:16:48.000
What's yours?
00:16:48.720 --> 00:16:56.320
So the next way of modernization isn't another technology migration, it's architectural convergence.
00:16:56.559 --> 00:16:59.600
So that's going to be my one-line quote for the day.
00:16:59.840 --> 00:17:00.639
Oh, very good.
00:17:00.799 --> 00:17:08.799
Thank you, Natish, and hope we see you all at future episodes of uh uh CFO Coffee Talk and future AI saline events in St.
00:17:08.880 --> 00:17:09.359
Petersburg.
00:17:09.680 --> 00:17:10.720
Bye, everybody.