2275 - Exploring the Role of AI Agents in Modern Business with Snow Leopard's Deepti Srivastava

0:04Well, hey, Thoughtful Leader. Welcome to the Thoughtful Entrepreneur. On this show, I interview purpose-driven founders, experts, and change makers who are making a real impact in the world. Want to be a guest? Simple. Go to upmyinfluence.com and click podcast. I'd love to promote your story to our audience of over 100 ,000 and across over 2 ,300 daily episodes and counting. Want to get booked on other great shows like this? Simple. Go to podverified.com and get your free guest authority score. Thousands of hosts are looking for high quality guests and Podverified connects you the right way. Now let's dive into today's episode.

0:58With us right now, it is Deepti Srivastava. Deepthi, you are the founder and CEO of Snow Leopard. You're found on the web at snowleopard.ai. Deepthi, it's great to have you. It's great to be here, Josh. Thanks so much for having me. Yeah, we're going to talk about what Snow Leopard is and can't wait to do that. Before we do, as a founder and you're in the Bay Area, what have you been doing for fun lately? What have you been geeking out on, interest, that sort of thing? yeah I think uh being in the bay area you geek out on technology that's really what you do it's been taken over by all the ai people um so there is a lot of like the fun part about that is that there's a lot of new stuff happening all over so there's you know non-traditional ways to get that information now of like what people are doing and how things are going whether it's podcasts or like you know blogs or you know uh instagram or linkedin or whatever so that's just been like the way I've been absorbing information and just geeking out on the new frontiers of AI and of technology in general.

2:03In terms of fun, we're lucky in San Francisco and in the Bay Area where there's a lot of like outdoor stuff to do, right? And the weather holds up most of the time. So my, like when I need to get away from the desk, what I love doing is just walking around San Francisco. It's actually really fun because every neighborhood is different. It's hilly. It's like every sort of, you have microclimate. So when I'm like, if I go for a longer walk, like, you know, one part, it's sunny, the other part is like foggy. So that's sort of my way to like really decompress. When you have friends or family come in from out of town and they say, ooh, let's do something San Francisco-y, where do you take them or where do they ask for?

2:50you know well they're like oh show me san francisco and i this is true i um actually have a two-hour tour uh like a full day tour a half day tour a four-day tour like a weekend tour yeah you know because you can you can do like just the golden gate bridge and the golden gate park in san francisco if you're familiar with those or you could do napa or you could do like there is a bunch of stuff to do. So we're very fortunate that way. I also have an elementary school kid. And so a lot of my fun time is actually just hanging out with him because as a startup founder, you either work or you spend time with your child and family.

3:28Yeah. All right. So you are a tech founder in AI. SnowLeopard.ai is your website. What is Snow Leopard? Yeah, it's designed to bridge the gap between AI agents in enterprises and your operational business data. So there's a couple of key insights. So I've spent like, I think backing up, I've spent like over two decades, you know, building enterprise data systems and data infrastructure, right, at Spanner, at Google, and like Oracle as an engineer before that, et cetera. And so, you know, as the AI sort of revolution started happening like a year or so ago, what I noticed is a couple of things, right?

4:16Like just having spent so much time building data stuff for enterprises, like there's three key insights, right? One is any new technology exists in the context of the technical architecture that already exists, right? Each enterprise has a unique ecosystem of technology and any new technology. It's not magic. It's like, has to fit into that system in order to, you know, truly gain value for the business, for the people. Side note, I care a lot about making technology have a positive impact on people. Ultimately, that's kind of why I'm here, right? I'm not here for the, you know, because tech itself is fun or, you know, technology for the sake of technology is fun.

4:58I actually enjoy building things that delight people and that actually help solve problems, Yes, make the world a better place. It's possible. I agree. I agree. So, you know, AI is a tool. It has to fit in the context of enterprises. Second is, you know, AI engineers, researchers, technologists aren't infrastructure and systems experts. And in my opinion, they shouldn't be. They shouldn't need to be. So there should be platforms and infrastructure that exists to help them focus on solving their problems with the technology. And in the world, this is the final point that actually leads to why Snow Leopard, which is in a world of agentic AI, we're really using AI to make decisions.

5:43And most decisions need up-to-date information. You cannot make accurate and positive moving forward decisions on inaccurate or stale data. And so Snow Leopard's job is to sit in between your AI agents and your operational business data. So your SQL databases, your warehouses, your API systems like Salesforce or HubSpot and fetch you the right information at the right time for that purpose. right so what that means is um you if you need to like in a world of ad hoc systems right like oh um you know do i need to replenish supply for water bottles because um there's a marathon going on right so do i need to do that that's sort of an ai agentic workflow will require you to actually know real-time information about what the um inventory looks like at a particular store Right.

6:47So using pipelines or ETL or RAD based systems just isn't going to cut it for that type of real time decision making, which is what we want AI to do, by the way. Right. Right. And just so we can make sure that everybody is kind of tracking along, you've used a couple of terms. Would you mind, one, I think is really, really critical that people understand. Can you describe or define agentic AI and why this is such a big deal? Yeah. So it's such an interesting question because, you know, on the one hand, there's so much hype around everybody's building agents. And on the other hand, the ground reality is like when I go to meetups or, you know, to conferences or just to like talk to, you know, potential customers, the first question is define agent.

7:41So clearly it's not such a huge like there is a lag between hype and actual adoption, I think is the main thing. But essentially, and the way I think about it is AI agents are just software that are designed to make autonomous decisions for specific or generalized use cases. So software like that already exists, for example. There's export systems, there are ML-based systems. There are just software programs that say, if X, then do Y, if A, then do B. And what we're moving towards is a world where there aren't predefined logic systems. They're like logic systems that can make decisions based on ad hoc inputs.

8:28But those ad hoc inputs, again, need to be the right information at the right time in order to make those sort of non-predefined workflows happen. Yeah. And so let's say that there's an SMB owner, maybe they've got an agency or something, consultancy, about 20 employees. What are some very common uses of agentic AI, which is becoming more and more and more available? Open AI, as of when we're recording this, just demo their stuff. There's other platforms that have been out. Google's doing that with A2A, yeah. Yeah, what are some really easy or common uses of agentic AI that you're like, you might want to be looking into this pretty soon because this has application for many, many business owners out there.

9:25If you were doing this, you could be using agents. If you're doing this, you could be using agents. Any ideas come to mind on that? Yeah, this is the interesting part, as I was saying, right? There's so much hype, but the reality is like people are still trying to figure it out. And ultimately, if agents are software, then any program, anything you're trying to build, any application you're trying to build now becomes an AI agent application. So, for example, support agents are the most common and most ubiquitously adopted AI agents there are right now. Right. Then there are agents that are helping people, you know, in the legal system or legal vertical or in the sort of medical vertical.

10:06It's starting to take shape right for unstructured data. So for PDFs, for, you know, website data, et cetera. right if you're looking at charts and graphs like as a um as a doctor or in the in the medical um like situations then it's helpful to have like ai agents that parse the prescriptions for example or parse the history and give you summaries um summarization of legal documents is a very common use case these days right so you have agents for that so i think we are still in this in this um phase of like agents being super easily understood and adopted unstructured information like PDFs and stuff.

10:46There's a huge uptake on using multimodal agents or multimodal models, LLMs for video and audio generation, right? Again, in the support use cases, like you can automate that whole workflow for support for various verticals using LLMs if you have the right sort of infrastructure in place for that. Yeah. DP, who are you working with currently? Who's currently using Snow Leopard and what have they been able to do? Yeah. So there's two things actually. One is we are working with financial services clients like SaaS, FinTech. They find a lot of usage internally right now for answering questions like, this is more of a sales and marketing internal use case, right?

11:37We're looking at people building AI agents for support where they actually want sort of, you know, their operational data, like where is my order? To answer that kind of question, you actually have to go look up one or two systems and, you know, actually look at them in real time, right? Things like, does somebody have a criminal record or not, like requires, again, up-to-date information. You can't just, you know, use that for, like answer that question or that type of question with stale data. So we're working with early design partners on both, on a variety of cases, use cases that are at this point internal, but we expect that as the agent systems mature, that they're going to start to use these kinds of data for external use cases as well.

12:24All right. Your website, snowleopard.ai. I see as we're recording this, You could sign up for early access. What are the next, what else would you recommend for someone that's joining us right now? Like what else they could do? Yeah. So we're actually working on publicly showcasing what you can do with structured data like Postgres and BigQuery and Snowflake and things like that. So by the time this goes out, like we should actually have, you get a sneak, like sneak insider information here, Josh. but we're actually going to release a Discord server where you can actually go and play with these datasets yourself, like public datasets, to see what you can do with Snow Leopard when you have a Postgres database or a BigQuery database or any of these kinds of databases, right?

13:12So we're really excited about that. We're going to launch it. I think it should be out in the next few weeks, which means by the time this airs. By the time this airs, it's live! Yeah. I have a pretty good substack as well. So I'd recommend for folks that are tracking AI and this intersection between data and agents. Great, great, great, great follow. All right. Yeah, our blog is blog.snowlepper.ai. And yeah, you're right. Like we talk a lot about data and AI and the sort of intersection of that and how you can actually use. We care a lot about making it useful, right? So operational data to make that useful in the AI context.

13:52So thank you for that shout out. Mm-hmm. All right. Well, Deepti Srivastava. I'm so sorry. I had it. You're fine. You're fine. Yes. Thank you. Okay. Again, founder, CEO of Snow Leopard, the website, snowleopard.ai. Deepti, thank you so much for joining us. Thanks so much, Josh.

14:18If you've enjoyed this conversation, I'd love to invite you to share your message with the world at podverified.com, where thousands of podcast hosts and thoughtful guests are connecting the right way. Podverified helps match you with the right stages for your level of authority. No more wasting time on brand new shows that just aren't a fit. Your guest authority score is free. And so are your first two podcast matches. Come hang out with me there. I'd love to learn more about your business, support your mission, and help promote you far and wide. Just head to www.podverified.com to get started.

15:02And of course, if you'd like to be a guest right here on The Thoughtful Entrepreneur, just go to upmyinfluence.com and click on podcast. Thanks for being a part of this movement. We rise by lifting others.