The Trifecta of Tech: Why Software, Data, and AI Must Work Together to Create Real Value

Speaker 0    00:00:00    
<silence>

Speaker 1    00:00:07    
This is a very special session because we've got a poll in this session. Mm-hmm <affirmative>. I would love for you all to get involved. We're gonna bring out our next speaker, Nisha, to the stage. Where are you at, Nisha? Hey, there you are.  

Speaker 2   00:00:25    
Hello. Can you hear me?  

Speaker 1    00:00:26    
Yes.  

Speaker 2   00:00:29    
Alright, thank you everyone. Good evening. Uh, but before I get started, I want to first do a big gratitude to all the organizers. Wonderful guitar. Thank you so much. Of course. The open source community and open house. Uh, special shout out to <inaudible> who invited me to speak today. And special, special shout out to Laura Anton. She's somewhere in the audience. She's my creative director who put the deck together. She, so you have to thank her when you see her on the chat. Uh, so on the next slide, um, I just wanted to kind of share, if you want to move the slide, uh, I wanna share a little bit of, um, how, how I got here, who am I and why are you listening to me? So I have, um, uh, there's no timeline here, but I came from India back in 1998. So dating myself, two plus decades.  

Speaker 2   00:01:29    
Lot of financial banking industry experience, but also in, uh, tech trenches from SQL Server 6.5. Many of you may or may not even have heard of that version to mainframe to most current, uh, last three years, I would say has been amazing learning journey on data, um, AI cloud, so much. So, uh, on all these, and that's probably the reason why you are hearing me most recently, I also had an opportunity to collaborate almost like an open source on this book that you see on the screen called, uh, the Secrets of AI Value Creation. I learned so much. There are 14 chapters. Each chapter is contributed by somebody. So extremely grateful to everybody who has taught me to this topic.

Speaker 2   00:02:21    
So this particular, uh, point of view for me, matters now is on the next slide. We are wild, wild west of ai, gold rush, if I may call it, and not to bore you with all the numbers, but the one that is amazing to me is this last one that McKinsey shared in their report. Uh, I'm sure many of you must have, uh, seen that report. It's like 10% <inaudible>. They have the data backbone for this very successful, and we have heard repeatedly, right? And all the present presentators today have shared that everybody shares data is the, uh, you know, the way to make this successful. So, having, uh, said that on the next slide, you know, my humble experience, again, you all probably in the audience are a lot more educated on this for this lifecycle, right? On any AI application, I feel like all the steps that are needed across, you know, these three disciplines, I would say, of data and ai, the whole lifecycle starts with, of course, and then going into wrangling the data that is needed.  

Speaker 2    00:03:38    
Um, sometimes we don't even know where the data is. And larger the organizations I've worked in, some large ones larger, the organization, you are just spending in amount of time to really get that data in the right shape so that it can, anybody can actually use it. And then on and on this lifecycle, and maybe some steps are missing as well, which is then brings me to the next slide, which is where, um, I feel my point of view again in 10 minutes I'll try to describe, is there is a massive disconnect across these disciplines. Now, software engineering, we know 30 years this, it's pretty mature from processes perspective.  

Speaker 2    00:04:24    
I was doing research and I found out that it is older, but the stack, and we heard from many, many people today on the, uh, conference, and we hear this all the time, if you take the data lifecycle from all the data to, you know, use the data, which is where some of the AI comes in, and AI engineer, that's like an absolute new, um, probably, uh, uh, function, a discipline that is being created, right? Um, and I, I, I'm, I have some music analog going on here. I feel like this is where a platform like open House can really do some value to all of us. So first time when I heard open house, I think we know about three, three years ago, or maybe two and a half years ago, I really thought of them almost like, um, you know, on a, on a, if I have a music sheet, they are the conductors on the podium and, and rest of what are all these instruments they try to play.  

Speaker 2   00:05:27    
And so anyway, disconnect and this point of view is what I want to share with you. Um, but before I go further, I wanna do a poll, quick poll. If you can, uh, have the poll and maybe the results for us here just to see if in the, on the next slide. The question is, in your current organization, how the three core engineering principles come together and are they collaborating on initiatives? And I think there are four options I gave you. We are a well-oiled, I call it trifecta, or you can say, I don't even know what is trifecta. So if we can get the poll going, we can see what is your opinion about this particular topic and how in your organization, I would imagine the smaller organizations probably don't have this problem, but <inaudible>.  

Speaker 1    00:06:35    
So we have, uh, people answering. I like it. It's cool. We've got it up on the, um, on the platform. Mines <laugh>. This is nice. All right, so we've got, um, B, C, and D. So basically we have pockets of collaboration, but silos exist. It often feels like separate teams working in parallel are the two with the most votes. And then what does a trifecta has a, a lot of votes, um, but not, not anywhere near those two middle ones.  

Speaker 2   00:07:12    
All right? Yeah. So trifecta is what we are discussing here. It's this munition of these three functions that have to come together to deliver an AI value. So why we don't even talk about trifecta,  

Speaker 2    00:07:35    
<inaudible>, uh, and again, if you haven't seen it, that I would love to hear from you. Your viewpoints on, on the point of view for me is like, it's a perfect time to bring these three, uh, functions together and really think of how we solve for some of the so-called bugs. So can we go to the next slide? I think first one I want to, uh, say is a conflicting priorities and metrics that we see. So from a value system perspective, if the AI doesn't have an actual value statement, what are we doing for that particular AI use case? That's where I see that software focuses on the app performance data focuses on the quality or bringing the data, and AI focuses on the model accuracy. And that's where I think the tension between these, these three disciplines. The other one I want to talk about is different life cycles and different stacks across the board, um, for software and data and ai.  

Speaker 2   00:08:41    
But we know one has to feed the other data governance and access is another point, right? How many of you, um, can confidently say that for all AI use cases, we have access to the data? So that's, that's another one. So over period of time, what I have seen is that these three disciplines are, you know, not not working hand to hand. They are very scattered, and they, they kind of, um, are working on their own individual teams agenda. Um, and so for, for the fixed part, I feel like there is a need for, um, a common, I would say understanding. It's on the next slide, the fixes on the next slide. It's a common understanding of the value system, the visibility of end-to-end, right? Right. Making sure that that data is available to, to all the functions, making sure the, the description of what we are talking as the value for a given company that is known to everybody.  

Speaker 2   00:09:46    
So that to me is the, um, thinking behind these, you know, the biggest hurdles and how do you think about all the platforms like we talked in data platforms today, Databricks, DBT, open House, snowflake, et cetera. But then there is also the software stack and the AI stack, which has to come together to be able to do that. So with that, I think there was another poll. If we can say, if you have seen these hurdles, what have you seen in implementing an AI use case? Are you seeing different tools and technologies? Are you seeing different life cycles, especially across these three dis discipline? And again, as I admitted, smaller companies may not see these three functions as a different function, but for larger companies, this has been my experience that we have three different disciplines which are trying to come together to generate that value. So quick poll here.  

Speaker 1    00:10:41    
Initial winner is manage managing data access and governance across teams.  

Speaker 2   00:10:49    
Yeah, which is, which is kind of obvious, right? Because data is, again, scattered all over the place. Um, some companies even don't even, I would say have data strategy. And for those who have the data strategy, the, the tooling, fragmentation in the data life cycle and the need for, at least in the banking sector, also probably health, healthcare and others, the need for, um, making sure the data is secure, only people who have access should have access. And all the, all the nuances of data management, which we can call governance is also another big, um, uh, you know, hurdle for, for, uh, ai. I had this experience in, in Capital One when I was there, the data scientists were always struggling to find, Hey, where is that data that I need? Or if I want to bring external data, how can I bring it securely fast and all, all that stuff.  

Speaker 2    00:11:45    
So, um, so on the next slide, I think I go into a little bit more on the value that I have seen when these three disciplines come together. In fact, I was just speaking to a commercial real estate EVP last week, and he was mentioning how even at the C level, they have created one role. I think they're calling it C-T-D-A-O, <laugh>. That makes sense, chief technology, data analytics officer. So I think it's the companies that have figured out, and those are the ones that are probably scaling to actually deliver value from, from ai. And some companies are mentioned here, my own, where I have spent a decade, uh, at and t some, some friends from at and t have been mentioning how some of these AI use cases are bringing value. But I think more and more I am interested in how's the operational and the scale of this ai.  

Speaker 2    00:12:42    
It's not just one use case that you put in production. It's about the entire, like I said, the lifecycle, and you repeat and rinse this lifecycle, right? Uh, I think next slide also has a couple good examples. I, I was, um, uh, talking to some of the folks in United, I think they have this idea about every flight has a story, and then how they bring that story together from an end to end perspective, right? So if, if, uh, Nisha is calling United, like how do they know everything about Nisha, the behaviors, the sentiments and everything, so that they can actually use, um, the, the predictive and, and the modeling part of it to be able to respond to me and still be that human touch. Um, many, many of their structures are also these principles of engineering that we have learned for last 30 years is all coming together for, for, for them as well.  

Speaker 2    00:13:39    
So I think on a final thought, bringing it to, to the next slide, um, I think, uh, it was Mark Anderson who has a famous code that software is eating the world. I think this is where trifecta comes in. For those who were asking me, to me, this is the trifecta that is building the future. And what is happening across this board, uh, is also across this, um, uh, you know, flywheel is that AI is also disrupting software engineering, data engineering, AI engineering. I think there was a talk just just before mine, somebody was talking about how they're using AI for the ETL and ELT right equation. So this, this synergy between software principles, data principles, AI principles, when we bring together, um, we, we see these experiences that hinges on that seamless integration and seamless way to generate the value. Um, another one was ar Airbnb that I was reading, that they have reduced like 15% of their customer inquiry.  

Speaker 2   00:14:42    
The, the question is not have they just increased, decreased, but the question is what is the customer experience across the board when this is, um, when this is applied? So yeah, bringing it all together, I think, um, uh, uh, in the, in the next slide, I have some questions. And the big takeaway for you is like, how do you sing from that same music sheet, right? So that that orchestration of all these things happen together in that shared data view, shared infrastructure, shared, you know, value that has been delivered for. So if, if anything from this talk that you can take away, trifecta is a, <laugh> is a word you can take away, but also, uh, if you are a software engineer, what principles of data engineer you need to learn? And same thing for all the other two as well. Data engineers. What principles can they learn? Because it's not just about their, your own silos, but it's also about how all of them work together for that scale, that repeat and rinse that next, uh, AI value can be generated out of it. Um, so with that, I think, um, uh, I don't know if time is left for any question, but, uh, I will leave that to you.  

Speaker 1    00:16:00    
Uh, I'm going to check right now. I think we got enough time for one question. We could do one. So a thought provoking question, but with the increase usage of AI for software engineering and having data engineering functions shrunk due to open platforms and ever increasing ease of ETL, won't we have a trifecta reduced to an AI engineer?  

Speaker 2    00:16:34    
Uh, potentially. That's a good one, right? But to me it's about the principle and the, uh, and all the things that we have to do for, you know, all the way from feature development and software to having the knowledge of the data. But you are absolutely right. That future might not be trifecta, but today it is. Clearly it is in organizations. In fact, um, I know organizations are creating this new function called AI engineer. So can you say that AI engineers will be able to do both the data and the, I hope so. I hope the learning curve for AI engineer is not just about that model, it's also about data. Like I am not, by the way you've seen my, uh, life, um, slide. I'm not expert in data or software. I've done many of these for people who have done data engineering, who know these formats that we talk about.  

Speaker 2    00:17:28    
I mean, Openhouse is a great example of a company, even Databricks and Snowflake, like people who are building these massive, massive platforms and products. They know this art of what does it take? I mean, I've dealt with, I mean, three digits, uh, petabytes of data, like right, we are talking large volumes of data. What does it take for data to land properly, store properly, not use the overuse, the compute? To your question, will AI engineers get trained enough on data? I hope so. I hope the models are also going to evolve. As we know, just this morning I was listening open, AI is already adding coding to your G now, so yeah, I do hope at some point the trifecta and this flywheel will hopefully become one word, but it's not there today. And the promise is, is little further out.  

Speaker 1   00:18:25    
Excellent. I must say that you were incredibly clear, and when I needed to change these slides, it was like, uh, it was perfect. So  

Speaker 2    00:18:35    
<laugh>, thank you for helping.