The Future of AI: Agents Over Tools?

Full transcript

Is AI a threat

So welcome back to the podcast, guys. Today, we are joined by Smita Kulkarni, who's the AI product manager. Smita, welcome to the podcast. Thanks for having me. So why don't we start by you telling us a little bit about yourself? Yeah. Absolutely. So I'm Smita. I'm an AI product manager at Accenture. My journey into AI began back in 2020 um, when I started pursuing my Master's in data science. At my current role, I'm building an AI assistant, uh, at an enterprise level that helps users, uh, get answers for answers from structured and unstructured data real time. And so any complex queries that they have, they just go ahead, ask this assistant, and they get a response. And I also inter it's interesting that, uh, prior to this, I also explored, uh, the startup world. I cofounded a company and helped launch an MVP. And this experience was great, uh, for me to learn how to take a product from, uh, zero to one. And, uh, yeah, outside of work, I personally enjoy, uh, using my personal projects, uh, as AI sandbox,

and I'm currently building a book tracker for myself that helps me track the books and recommends me the books that I would be interested in reading based on my personal ratings. So, yeah, this has been a great journey for me so far. Uh-huh. And when you said, AI, like, are they voice powered, or is it chat? As of now, it is all chat. Uh, the AI assistant that I'm building, it's all based on chat. The users go in and, uh, put in their query that may be related to the database, like, the their data related questions, or they have questions pertaining to, like, hundreds of documents out there around policies, user guides. They just go ahead and ask the questions and get a response real time that helps them in getting their decisions faster. Mhmm. And what are some of the trends that you're seeing right now in AI? Well, uh, that's an interesting question. Uh, so one trend that I'm seeing is we are moving from, uh, building tools that are basically helping us answer questions to agents where agents are reasoning

and taking actions on behalf of a human. So this shift has been amazing to see, uh, so far. Mhmm. And how are you currently leveraging AI in your day to day? Yeah. That's that's a great question. So I, uh, I'm a strong advocate of, um, making sure that you use your personal use cases as AI sandbox and see what is a good use case to build with AI and whatnot. Right? This is a great learning experience even outside of your work. And, um, I, uh, I have multiple tools that help me in, going forward with that. Like, for example, um, I did try exploring no code tool. So if there there are people who are not very comfortable with coding, we have lot of AI tools like, like, make, and so on. Right? So you just go ahead. You register yourself and try using the no code tools to use your personal use cases. So that's one aspect of it. Otherwise, even if you are familiar with little bit of coding, right now, we have coding companions like, uh, GitHub Copilot or Codex or, of course, Cloud Code.

AI literacy is vital

Right? Uh, you just go and have a conversation in a natural language, and you can build. You can go about from idea to prototype in such a short duration using the coding companion tools as well. So these are both are my go to tools at this point in time based on what the idea is in my mind. What's your favorite AI coding tool? You mentioned some of them. So, uh, for me, I'm, uh, I'm comfortable with coding. I had I have a master's in data science, and before that, I had a bachelor's in computer science. So I'm familiar with coding, and GitHub Copilot is my favorite at this point in time. And there is a, uh, tool that I built that I was talking about called as leaf note. Right? The book tracker tool. So the whole idea to, uh, to moving it to prototype, to releasing the tool for other users to use it, it just all went about in one to two days. Just over the weekend, I had the idea. I built it and released it. So GitHub Copilot has been great, like, helping me debugging,

if I'm stuck at some point. And, uh, the whole like, once you have the PRD ready, again, using the AI tools, you just go with, uh, GitHub Copilot to help build the product. So GitHub Copilot has been my favorite so far. Is there any technical challenge that you encounter? So, uh, when you are the solo builder, right, you are the architect. You are the developer. You are the product manager. Like, when you are building solo, of course, you will get into multiple technical challenges, like, um, how like, which, uh, tools to use, right, and what database to use, what is the database design that you want to model for this particular use case and so on. And, uh, yeah. So, like, uh, all that architecting is not my genre, So that was one technical challenge that I was facing. But, again, uh, I was able to, uh, work on it based on brainstorming using different AI tools and what would be the best, uh, systems and tools to use for my use case. Mhmm. I'm curious. Do you think that AI is a threat?

Wow. Uh, well, that's a great question. I think most of us think about as AI a threat. Is AI going to change the world? But more interesting question to ask would be, how will humans change because of AI? Right? So, uh, and I believe, uh, the people who are getting comfortable with AI at an early stage learning how to collaborate with agents and build skills around, uh, multi agent collaborations and so, uh, so on will be the ones who are going to define the future. So my recommendation would be to start getting into AI. AI literacy is not an option anymore. So it's a fundamental skill that everyone should start learning, and that's that's how it's going to help everyone, uh, to take the future forward with AI. For sure. I'm also learning AI as of now, so it's, uh, it's been a fun experience. I'm interested if you you are obviously in the AI industry for quite some time right now. What would you change if you had to pick something about it? Wow. Interesting. So, uh, I think more, uh, hype is around using powerful models,

From tools to agents

learning, uh, all about the latest models that are coming out there. People start focusing and chasing what's new tools, what's new models in the industry. But what I would recommend is the focus should be more on still the fundamental skills around, is this the right problem to solve? Is this the right product direction that we have to take? Uh, so, basically, uh, the underlying foundational skills is something that you should be focusing more on with AI around us, than focusing and chasing the latest models that have been that are being getting released out there. How do you actually envision your career evolving over the next few years as AI is so rapidly changing? Yeah. Uh, that's interesting. So I think in the future, um, the focus for me should be getting comfortable and working along with agents. So the productivity for myself will not be measured

based on how much amount of work I'm able to do it myself, but instead how effectively I'm able to delegate my work to the agents and collaborate across multiple agents to get the work done. So that's the shift that I'm going to see for myself as well in the future. Mhmm. Do you think people will get fall behind? Do you think that people will fall behind when they are not using AI and these kind of agents? That's a tricky question. Um, I definitely think the learning AI skills has become a fundamental skills for professionals out there, and it is very important, um, to get familiar with AI tools and you need not be a machine learning researcher as such. Right? But being familiar with what are the tools out there and how can the AI help you to get more productive, this is something that I definitely think that people should start focusing on. Start Mhmm. I'm I'm not sure if you mentioned it, but besides your master's degree and bachelor that you did, how did you actually get into AI since it's like, I recognized AI maybe one and a half year ago,

And you said you started in 2021, if I'm not Yeah. 2020. My journey has been interesting. So back in 2020, uh, I started hearing about machine learning models that are predicting based on the patterns that exist in the data. And that whole concept itself was extremely intriguing for me. So it was magical. Like, you learn patterns and you're predicting the future. How cool is that? So this idea itself started making me more curious, and I decided to pursue my master's in data science, uh, that too with a full time job and taking care of a two year old. But still, it was challenging, but the curiosity itself was so strong that I wanted to go deeper into that space. And once I started pursuing my master's, of course, the shift started happening from traditional machine learning to generative AI to agentic AI And, um, actively being involved and partis like, being part of the whole industry shift has been a great learning experience for me. What advice would you give to someone if they are just starting out this year and they would like to pursue a career in AI product management? Yeah.

What would that advice be? That's that's a great question. So as I was saying, uh, that AI literacy is not an option anymore. Every professional, no matter at what phase, uh, he or she is, whether entry level or intermediate or a professional, they have to be, uh, familiar with AI. So my advice would be very simple, uh, with the amount of ease that's out there now. Right? So you can go, uh, about from idea to prototype in just a matter of few hours. So the advice here is start learning with start learning with AI, start experimenting, start building. And it is very simple out there. Once you get into it and start looking out for help, it it is very easier than it seems to be. Great, Smita. Thanks for joining us on today's podcast. I will leave your LinkedIn so they can check you out if they would like. And thank you guys for watching. Thank you so much for having me.