How to Get Hired for Applied AI Roles in Fortune 500 Companies (Target Sr Data Scientist POV) - w/ Sowmya

What if the most important AI skill at a Fortune 500 company is knowing when to not use AI? So it's not about which enterprise of which company has the best AI or the best platform. It's about who has the best judgment in terms of when not to use AI actually or even when to use AI. Joining me today is Samya Podila. She's on the centralized generative AI team at Target. The team that literally decides what AI gets built and what gets killed. Cost was not even a discussion point on the table.

You can go experiment. You can have your own model. You can spin up a DD. Everything is self-service and you can just have credits. You have free reigns, right? You can choose your model, do it however you want and then go talk to the person that's going to be using it. So it's your customer quote-unquote. Clearly you get access to these models even prior to the public release. So we know that four points. iPod 4 is coming at least one or two weeks ahead.

We get to benchmark it on our own data set. So by the time it's already a word out there, we know what it actually knows to our people kind of on LinkedIn or you know. Lances claim these things say now this is the next AGM. Not really. Samya is literally seeing the future before it drops and her take is that most of the hype is actually just noise. It empowers 100x engineers to build 1000x faster. It also flows down the enterprises by 10x or even 100x because who is reviewing all of this to ensure that it meets the company's standards.

There is a code I can just submit the PR. I'm not even reviewing the code. And then some engineer comes and also just add an automation style push it to CI/CD. It's in production of the normal has reviewed ADT. Subscribe on YouTube 45 or wherever you get your podcasts. Here we go. Hey, Naman. Nice to meet you virtually today. The first thing I want to talk about is what is implementing AI or you know, agents look like in a retail giant.

The size of target, right? It's obviously, as we know, the biggest retail companies in the world. And for those of our listeners that aren't super familiar, that includes me. What does it look like to be in the weeds trying to figure out how AI works with enterprises? Because obviously I'm assuming that looks completely different from like a solo pranor, like myself, just opening up open claw and having a bunch of agents go rogue.

So can you help set the stage for us in terms of maybe just a day in the life and kind of what sort of projects you get to work on in this place? Wow, diving right in. First of all, thank you so much for thinking of me to have me on your show. That's a real pleasure. One thing, I mean, I'll start from where it started with target and why I chose to join target actually that might set the stage for some of these conversations.

When I interviewed for the first round with my hiring manager who is the director of the generative AI team at Target, gen AI is still pretty new. Other than chat, GPT and a few other chatbot kind of experiences, people have developed. Agents is really not the area. Agents is still not that big of a hype as it is today. And everyone is still figuring out what it looks, what it means to have a gen AI strategy for a company as such.

And I have that question because I was interviewing with Fortune 500 companies and I was figuring out what they're doing. I was figuring out what their AI strategy is or how I can be part of that. And I only had questions in my preparation process and very few answers are very few clearer parts of other enterprises having a strategy in the first place. So I was even figuring out if I don't know what their strategy looks like, I don't even know what I can bring to their table.

So I had only these questions working into these interviews basically. And I put the same question across to my director on what does it look like, what is the enterprise strategy for Target. He kind of gave a very clear roadmap that I kind of still remember to date. And that first question he answered for me actually gave me that interest or passion towards preparing and you know for the rest of the rounds and want to pursue that opportunity with Target.

So they talked about having like an enterprise wide platform to experiment basically. Hey, let's bring all the models out here. Our own internal platform hosts them because we don't even know what govern and security is complaints. All those things, no one even knows but they already talked to that. They want to build out their own on-prem platform to bring some of these models purely for experimentation purposes. They all had these enterprise license contracts.

And they had this full phase of you know having a centralized Gen AI company that will go and it works as a startup inside this company that will go and embed with various product and business teams inside Target. Both in terms of enabling them to have AI literacy in the first place to understand what AI even means. Get the basic terms right, understand this technology, what it can do for them. Then also understand certain use cases that could be good candidates to leverage an AI and what their pain points are, how could a POC look like.

So it's our team's role is to kind of do that literacy basic enablement to understand what use cases would even make sense and bring some of these use cases and do a POC. Work to understand the business process, the tribal knowledge these teams has. What's the pain point and do like a four to six week very quick POC to even see if this is a use case candidate for AI versus not. And then there is value will work with the team to see them build them the minimal viable product.

And in the process also hand hold the team so that they would do some of these stuff and they're in a good place to take it forward and maintain it or deployed in production and continue enabling and maintaining the product so that we can take a step back and go into other teams you know doing the same work over and over. So instead of letting every team go rogue they had a central team helping them kind of delete rate and also hand hold them in identifying cases you know doing testing the waters and also understanding where the value comes from versus not and also scaling and so on and so forth having that central team who kind of is an expert team in house but also giving them the range to experiment you know fail fast and also learn fast so that system that framework they had.

It is not an answer to what is your enterprise strategy but I think that was the right first step in figuring out what could an enterprise strategy look like because no one has figured out at that point. I don't think given companies have figuring out every day even today. So the approach kind of felt like amazing. And given they are a very big enterprise they have the bandwidth to kind of build their own platform bring models give rains free rains to experiment or be okay with failure.

And initially they did not also impose as much hey if the use case has to bring an ROI it has to show value that emphasis was very low. The cost was not even a discussion point on the table. Can we experiment you can have your own model you can spin a PDB you can everything is self service and you can just have credits. I mean again it's not unlimited credits but of course credits do not even question cost on the table to start with.

So I kind of like having that level of enterprise privilege to experiment freely is one thing and also having that expert in house team where people really are still figuring out this technology. They are able to bring the right talent on board building central AI team that will help all these smaller teams inside the company. So to bring the powerhouse of talent also kind of you know is one thing enterprise good effort.

So that's how it feels like to be part of a enterprise in AI world is to have that privilege of experimentation and connections to a better talent network and people who are curious also to experiment and learn. Yeah I mean in many ways I can relate to a lot of what you're saying because at my own job and the only difference here I guess is that my job has nothing to do with AI yet. But so I work at the center of excellence at cushion and wake.

But it's the same exact model that you mentioned right so in our case or at least when I first joined we were just working with a lot of low code no code tools. And we were just kind of injected all over the business so tax brokers accountants. You know your anyway like just people that are not related to technology that don't quite frankly need it outside of your usual excel stuff. And then we used to be parachuted in and then we teach them how to automate like a certain thing that's taking a lot of the time.

And then we kind of just exit. So in your case it's sounding a lot like it has shades of that same thing but here you're explicitly focused on a generative AI or maybe as you said maybe going forward more agent type experiences. That's already really cool and my favorite part about this is almost the entrepreneurial angle that just seeps into this right because you have free rein right you can choose your model do it however you want and then go talk to the person that's going to be using it. So it's your customer quote and quote.

Yeah and I think that is just yeah I love that you got to experience with that. I am curious though when you are interviewing I'm assuming you must have spoken to a few other companies like this. What were some of the other responses that you got when you asked them that hey what is your organizational AI strategy and to further set this stage if you could also share around what timeline this is happening in. I think that would make it much more richer in context.

Sure this is like late 23 early 24 so you know one to two years two to three years into this gen AI experience as such where companies are taking it home this is not a far this is going to stay so we should have a team we should have a strategy. We should go all in so they were all making those calls. And there were also like this is at the same time where companies have already started doing the first round of layoff it's at the post covid post chat GPT that kind of level setting for the over financial commitment a lot of companies have done.

So they were level setting so they were layoff's happening but also hiring happening and no one has five years experience engineer. But people were actually looking for such companies actually looking for someone who has who could bring their five years of experience even the technology is two to three years in the game basically. So the market was very competitive specially for Janaeros because everyone is trying to claim their prompt engineer and want 500 game salaries so that I remember that era. Exactly so it was that era.

And a lot of companies still were saying you know various things about having an enterprise license versus just wanting to use chat GPT versus you know trying out different open source close source models just for experimentation. So a lot of companies had a broader roadmap or access to the right kind of enterprise licenses as well. So there are different companies on that note but I also work interviewed for other fortune players for example Google was one of them.

Google still at that time I kind of felt like oh my god Google is still one step behind catching up with these AI providers specially with open AI it kind of felt like you know they still didn't have that full AI suite of tools that they have today. And they were still on the verge of figuring out those things so yeah I company at that moment but cut fast forward to 2026 they caught up and like now need it. I agree with that I think they started off stumbling a bit because I think they what was it called Google bard I think is how they started.

And then they'd like discontinued it and then restarted Gemini I think like a few months later so I do remember them getting off to a rocky start but I agree with you that yeah they have caught up since. Sorry I didn't mean to interrupt. No no no no interrupting I love a conversation so that's exactly letting you got the things right and I when I interviewed for example Google had a product called dialogue flow which is their conversational UI AI. Conversational project is not AI as such.

So they had a lot of customers using their dialogue flow for automating customer service interactions it's more rule based heuristic based they have conversations redesign. Given they had a suite of products that they have invested into it's not like one day a company like Google but will discard all of that and saying hey let's discard dialogue flow let's go conversation AI with full AI models right they didn't make that call.

So the interviews were still for some of those roles that they want someone to understand AI but still use the traditional products and maybe eventually figure out you know a road map if there is potential here and there. So they didn't fully feel like I was really interested in a more genuine kind of driven role but Google kind of company still had a lot of traditional work products work teams in place with at least the roles I interviewed were in that direction.

So I was not sure of the opportunity I would get to full scale experiment with Janae but like target said I am part of a centralized in 18 we have building our own platform we have access to all these models full range to experiment. They feel like oh my god this would be a gold mine to play with you know so that excited me more but I think yeah there were people who are figuring out and it was across the spectrum in terms of where enterprises or even small companies had their strategies.

They were across the spectrum but everyone at least had that vision that this is not a fact this is going to take off we need to figure out how many have figured out where at what stage they are was you know here and there actually. Yeah I think what I'm hearing is that there was almost like a giant market wide FOMO and everybody just wanted in even though they didn't really know what was the best way to get started they would just like let's just do something.

And we can figure out what they should be doing and I still knew I didn't have the guts to say that hey I don't think no one had the guts to say that they were probably proven wrong because no one had the vision into how this is actually going to shape you know so they wanted people who could shape the vision and yeah maybe I could have been more confident and told myself better I said like this is how I think but I don't think I also know I was being honest in those interviews and yeah.

Yeah I mean there's always always value in being honest in interviews right I'm sure you know that much much much better than me given your more experience have had just more time in the in the corporate space. I am curious so given what you just said that just sounds like that would throw off my interview prep completely because then what are you preparing for right like it always becomes like a chicken and egg problem because you don't know what you prepare because there is nothing to prepare but you still have to prepare so how did you go about that process and then maybe if you could also some shed some light around what sort of questions they asked maybe in terms of technical abilities.

But did they look for things in particular or mostly just projects. I guess what does that entire world look like because I think for me and a lot of my audience which is not exposed to like the senior middle career type roles in this basis. So I'm very curious actually how this entire experience was for you in terms of preparing. Sure again again we're talking about two years ago where to do more than two years ago two and half to three years ago but this technology still nascent right correct yep.

So there were people like I said claiming that they know things they have done projects there are people who have done to but there are also people who are claiming that they have five years experience in this technology and they can do a lot of things. Given I still didn't have that hands on I don't want to boast about that I tried to do some experimentation on my own so that I also don't sound like I don't know anything you teach me so then you're not going to learn the rule right.

But that was the reality for most people but again you need to package yourself and also see inside in terms of hey what are the existing skills that I have that could add value to someone figuring out Jennay at this point. So I did that I had background in NLP machine learning deep learning so all these fundamental technologies and in gendered way was a natural extension to all of that work. So I kind of made sure that I know my fundamentals right when I say fundamentals the entire deep down drill down into what a transformer architecture looks like for example and even if you were to build a transformer model from scratch what would that look like.

So having I think even to tell a lot of people do not have that clarity is what I feel people focus simply on like prompt engineering and all of that so that's applied AI. There are rules for that people are doing that great but if you want to build something when especially when something is new having that strong foundational layer is really important is what I thought and that was in my hands because it was my wheelhouse I had the experience in ML and deep learning understanding transformers you know all those things definitely you know I'm suited to do that right rather than just focusing on applications.

So I did the understand all the fundamentals very well and I try to do small projects when I say prompt engineering you try yourself what has worked what has not worked what is your view or perspective on prompt engineering rather than that one block that our official vendor have put out that maybe everyone has read but again when you implemented that might not have been your experience you know. So just that little bit of fear and there hey what has worked for me what has not worked for me when I tried it myself so having that first hand experience when especially everyone who didn't have experience or claiming to have experience.

It helped the interviewer separate that you know that say that noise versus true signal or true passion for implementation so people everyone was trying to weed out people basically at that time on who actually has done something versus not or who has the capability to even do something so because everyone is applying for generals interviews where interviewers were even trying so hard to figure out that signal you know from a massive pool of candidates.

So having fundamentals right at least some real time experience own passion project experimentation anything that you can get your hands dirty with that will give you first hand experience rather than just reading from someone else's experience. So I think that help and also like having a imaginative vision again I kind of had thought okay I don't know what other companies are doing. I don't know what even I would do but let me imagine again most of it could be just be a side but I try to have my own imaginative vision on what could it look like you know can I think of a very futuristic use case five years down there and what could it even be your company like this how could even leverage

again all of that might not make sense because no one knows the answer or maybe people have that vision can at least you know analyze or connect to some of my visions so I tried to have a visual what could look like for a company like target or Google you know what it could look like so I kind of was doing that and also obviously doing a thorough search into any of their existing work what exactly have they done so far anything out there in the web that I can pick and say that hey I know you guys are working on this so at least I have done my homework you know.

Yeah honestly the it's so interesting that I know we were talking about this before we hit record here but the thing that what you just said it reminds me of is one of my guests was we were talking about chess right like this person just taught themselves how to play chess and got really good at it and they said something that stuck to me which was that even a bad plan is better than no plan. And I think that's kind of what you're laying out that yeah no one knows right but you have to at least be creative enough to think of what it might look like you can't just be like oh I don't know you figure it out. You have to change it better.

Oh yeah. No no but like you were the one that put out the info what I'm more curious about and you touched on this briefly but maybe you can expand on this a little bit more but I get the sense of the field two and a half years ago. What is it like right now because I'm sure you're involved with probably at least in some way with the hiring and such and I guess even if you're not. If somebody is trying to get into applied AI currently at fortune 500 or even like smaller companies. What are some of the requirements that exist is is it mostly the same or has stuff changed since the last two and a half years.

I would say things have definitely changed a lot the space at which AI is evolving is just crazy. And the number of tools out there every other day there is a new model update that meets all of the benchmarks you know that beats all of the benchmarks so the pace of evolution is crazy. And I kind of feel like now it comes to this question. So it's not about which enterprise of which company has the best AI or the best platform.

It's about who has the best judgment in terms of when not to use AI actually or even when to use AI. You know I kind of feel like it's coming to that so everyone has the infrastructure everyone has the models everyone have the capabilities now everyone kind of at least got a sense on how to use this technology. So you know the how but where does it make sense where is where is the ROI and where exactly you still need human judgment you know and if everything is automated who exactly can consume all of this in this world so there are all those questions on the table.

And what is the right to use case or a business problem that needs AI that otherwise that doesn't need AI is a simple automation problem that can be solved with human that can be solved with rule based or heuristic technologies automation methodologies and there is that space where you still need that expert judgment human judgment the creativity that you know AI systems might not be able to bring as of today. So identifying that way not to use is more important than way to even use in the first place because a lot of companies in that mean we have all experimentation is good but in that light lot of companies have pursued a force that might not be great use cases for AI and it's good to identify that shut them down we prioritize we strategize understand where it makes sense to use this versus not.

So in that light right so people are also expecting candidates who kind of have that insight into this is an AI problem versus this is not an AI problem. And also if it is an AI problem how does it look at an enterprise scale or a production scale when you want to put it in front of actually real people using it. How does it look like because now that experimentation is slightly coming to a saturation phase and people are more interested in actually using them in real time.

So what does it look like when AI systems are put in front of users at a production scale that's what they're looking for so we're not to use. If you're using it how to make it production great ready so that people can actually use it and what breaks in production so people are looking for companies are looking for candidates that will have insight into those kind of questions these days. Yeah something you said around I think it was curious to hear you say that if there's a new model dropping every day right because I actually I'm not sure why but until now I was just under the impression that that's only a problem that exists when you're

not in like a really big enterprise established fortune 500 organization. It does make sense what you said I guess what I'm curious about double clicking on is, is do you feel the need to use the latest and greatest models or is it okay as long as it gets the job done, especially in your situation where you have free rain, you have almost like a siloed organizational structure that can talk to any other vertical as and when required.

Does that enable you to kind of not chase the newest shiniest toy and just focus more on the delivering value piece of it. Or do you still also just feel like oh man there's a new cloud. So let's like clue that in just to see how we do. No I think you already you almost answered the question yourself. So we have an internal team where we benchmark these models against our own use cases. So we have a suite of you know benchmarking data sets that we built and we know where the gaps are so there are certain use cases where even an open source model is actually doing really well so we know that we almost had a 90 to 100% accuracy within an open source models are like a lightweight smaller models you don't even need the powerful model out there for certain use cases.

We know that data we have that data sets actually that we could use when a new model comes you know we can test it on. So when we run our benchmarking actually we get access to these models even you know prior to the public release so we know that. High point for is coming at least one or two weeks ahead we get to benchmark it on our own data set so by the time it's already a word out there we know what it actually knows.

We know that it's coming and we know what it looks like on various use cases at an enterprise scale so when people kind of on LinkedIn or you know the lens has claimed these things say now this is the next AGM not really. But yeah it's a better model on certain use cases so we are already with a different view by the time people are hyping about it. So that's it we also know where there is a lift here we already know that there are these certain challenging use cases inside our company that are still at 40 to 50% accuracy only there are a lot of hallucinations or production grade failures and there's a lot of human in the loop intervention happening we know those use cases.

And when a powerful model comes in we are at across all the speeds but we look if there is a potential improvement that these models are bringing in to those challenging use cases. We'll definitely look that and if not hey we know that we do not even need the shiniest, most powerful model close to AGI for a very simple use case you don't need to do that so we will decide on which use cases would get that lift from this new model.

So yeah it's definitely you know finding what use cases where these models will add value is the direction. Yeah earlier you said that for agents I think maybe I sensed it incorrectly but I almost sensed reluctance slash it's not there yet slash almost like a denial maybe by denial I mean just that they're not good enough. So talk to me about agents like why haven't agents caught up yet at big enterprises and I guess what's the gap because again from my vantage point for my little pet use cases you know that are not serious at all that are just like finding guests for my show pitching me on guests as a guest on other shows like these really toy examples agents seem to be doing a pretty good job.

So why has there not been a similar rate of adoption at maybe I don't know like maybe you can't speak for all of them but at least sounded like at Target it's not there yet. So why do you think that is. Well so it is there yet but again we had done our own experimentation so I can elaborate on that and again use a lot pet use cases they're all valued use cases I think AI is helping individuals leverage rather than just organizations and that's the power.

That's the power behind this technology on why it's kind of you know call the next revolution after electricity internet where they're calling it the next revolution because it's not just impacting bigger organizations who have lots of money. Everyone anyone can use it and should be using it and are using it so that's the magic of this I revolution. That's their enterprises are using agents for example we have done individual agent workflows where there is one agent instead of you doing one single turn interaction or one pass you have built an agent system it has access to external tools and all of that.

Where it's a complex flow we have done those projects for sure. And what we have been hearing from external markets is that you know hey there is also a lot of hype in this direction that there is that multi agent the universe possible where it will take an enterprise scale problem actually would take you know 100 people but can all solve all of that together that complex universe Christopher Nolan universe of agents is out there and it's possible. So you know even as we have been hearing that signal from external markets and I'm like why are we not doing it then so they kind of that question back to us. Do you think that multi agent the universe where you know there are for example 10 teams inside target has built 10 different agents can now there be a super agent where that takes all of these 10 agents and can solve a very

popular problem on its own market say so is that a reality you know they had got that question to us and we have built like an agent simulator on which we have published a paper as well and got accepted at authentic conference in Cyprus. So we have built our own simulator kind of environment where you can plug and play various agents that are built in various frameworks. For example a few frameworks out there are Langraff crew autogen semantic framework there are different frameworks out there on which you can build agents. Again these are not the consumer focus but what enterprise create you know frameworks that enterprises use to build agents when we kind of took intentionally built one agent in one framework and another agent in another framework and all of them will solve smaller problems.

For example we are interested in seeing if all of these agents together can plan the entire assortment catalog of target what should the next season catalog for the online store can look like. There is an inventory analyst there's a sales analysis analyst you know there is some other vendor analyst you know if there are all these smaller individual capabilities are different teams have built out can we build a super agent that can go all of these agents as needed and can solve like the full retail problem what should target be selling the next season you know.

So we tried doing those experiments and like I said different agents were built in different frameworks just for the sake of experimentation to see the chaos actually it would create. And what we realized is that for example for these to be successful enterprises need to have data in a way that AI systems can understand that data with documentation and we say data there's a table out there with numbers but again what does that numbers mean what does that column mean the column description of business use case of this table and how does business decisions are made for example how do we even choose up to vendors how do we decide something is a trend but target might not adhere to all the trends right maybe you will try.

So there are different business tribal knowledge decisions that happen is if that's all not captured anywhere now it's hard for someone to sit down and do all of that unless someone does that right. So that business process gap is there so a agents does not know all of these knowledge is that companies use so how can they make the system so that is one hurdle. And the second hurdle is how different agents communicate then something fails how does the retry when there is hallucination how to detect things so there are all these technical challenges because AI is a non-deterministic system it cannot give right answers all the time.

So that evaluation you know and especially when companies like retail that have a lot of consumer data there's a lot of personally identifiable information. So there are a lot of regulations around compliance you know how to safely use that data is also there so you cannot just take all the consumer sales data and just simply give it to AI systems. So there are a lot of processes around which we have to you know ensure that you are using the data right.

So there are all these enterprise scale concerns that come into picture but most of them all it's kind of inter-Asian communication especially when you're talking about multi-agent systems is a problem these frameworks are still evolving. And the data and the business process is not actually captured right for the AI system to leverage it they're all existing in share points confluence something to make sell it's kind of hard to bring them all together.

So it came down to those nuances to be able to leverage agents at enterprise scale as of today. It's a it's a possibility in the near term companies are now focusing on how can we make our metadata and documentation AI ready. So they're now looking at the foundation layers needed to kind of see this multi-agent tech universality based. That is so cool so basically what you're saying is if we found a way to normalize all of our data into the same common source whatever that looks like.

That would already put us much more closer to having a completely agent run like organizational wide workflow than we do currently so the bottleneck isn't the ability of the agents as much as it is the state of normalization and such of the data itself would that be accurate. Yeah, yeah I think you rephrased it exactly right so yeah. Wow that is so interesting actually yeah I had never once considered that and again this is the like the part of the reason why I asked you the question because.

Yeah it's just different right it's just completely different when you do the same thing at a place like target versus you know that like a no name two person company even though you're using quote unquote the same tool. It just everything is different it's almost as if like it's not even the same league that that you're operating in which it that part of it just really fascinates me and even something you said earlier was something I was having like a random shower thought about that if apparently these tools allow every person to 100x themselves should they also allow enterprises with a thousand people to also 100x them, but that has not happened yet and I think that is something that we will just kind of maybe have to wait

and watch in terms of how that plays out because the way I understand it and maybe you can educate me is that like organizations that are much smaller are actually at an advantage here because there is less sinking to be done for this 100x steps happening versus having many many more people because then you run into the very real problem of how do you synchronize all of the out productivity increases that have come across as a result of these tools.

Any thoughts on that piece is that something that you have an opinion on. You have packed a lot of interesting questions into this thing. All great question maybe I'll take segues into answering each of these individually. First thing is that you know a smaller company the league is different. Actually even though the league is different I kind of feel like they have the same problem as well. For example, I want to relay the question back to you if when you are using your open flow or clock over your own agents.

You want it to think like you and act like you you just do not want another regurgled response that looks like every other response out there you know where people might have it and just post on LinkedIn. You don't want to look like that you have a better thought process you have your own perspective you have a personality you want that to be reflected. People are trying to write all these skills agent files you know build elaborate problems with your own brain your own thought process and personality and infuse it. But how many were able to actually achieve it.

Similar to targets having the data and business processes in place. You also need to have your past assets beat your resume your past projects you know everything that you have done or your past podcast as a reference data set. When you are generating you using a to generate a script for your next podcast you the idea is to understand who you are so creating that data and the business processes or that judgment call you have in brain.

That also needs to translate and power your AI system friend that's when they'll actually be useful otherwise if I do a podcast versus your podcast if you're doing the same thing using a how does even differentiate you and me and 10,000 other podcast is who could actually now leverage. So even individual people will now have that problem maybe the early adopters they were just just fast forwarding it. But now millions of people are using it everyone actually and they are the AI systems are also now getting normalized to a point where if you ask the same question to chat GPT versus Gemini versus Claude.

And it was getting very similar response. And eventually if you and I asked without that memory without that customization also might get very similar response then creativity personalized part is lost unless you put a lot of effort to customize it as for your world basically. So individual people will face the same problem as enterprises so because it looks like AI gavel response so you need you cannot exactly use it as is it halfway through would feel like I could actually have done this myself because it doesn't like me.

And yeah, I feel like that like 60% of that I am not gonna like why am I doing this is like just let me just open Gmail and just write it. It's just not much easier. Yeah. And even if it can even if these systems can understand you personalize and still automate at scale work around the clock right millions of people are producing podcast blogs research papers. So the spirit which everything is produced is so fast. Can anyone catch up are they see the audience were willing to catch up or it's just your agent creating my agent summarizing it nothing good. Nothing that's getting created is actually you are you know garbage in garbage out right.

Exactly. No one is seeing the full podcast reading a two line summary then why do you even create a podcast using a because no reading it. It's just that resources are infinitely wasted in a loop is what would happen. So, eventually that would break even at a two people's a solo premier creator level. All of these problems would become real and they also have to take a step back to not out of proportion. You not use these technologies out of proportion you will actually resort to using it when needed with productivity key and it matters only. So that will happen.

That said right. So the other question you are asked us how is it enabling people to be better be 100 X right I think at a big organization scale. Yes. So people can like I said read the same problem again can be uploaded and enterprise scale. For example, even a lot of open source code bases are now going through this. There are hundreds of agents out there submitting PR. Open source code base similarly to even an internal code base. They have given cloud code access so we can just generate amounts of code and we're like my job is done. There is a code I can just say that here I'm not even reviewing the code.

And then some engineer comes and also just at an automation style push it to see I see it's in production of there. No one has to do anything. So if that is the world. Okay, enterprise is tackled when an enterprise put something even if a small mishap happens in production grid, it makes it to headlines as the number this song goes down. So they cannot afford to do that. So yeah, it empowers 100 X engineers to build 1000 X faster. It also slows down the enterprise by 10 X or even 100 X because who is exactly all of this to ensure that it needs the company standards, the policies are that brand image because you do not want to put anything out

of tone for the company. Right. So now that regulation compliance reviewing monitoring that has become the bottleneck for the systems and that's why companies are not maybe going as rogue as they could have actually. Wow. Yeah, that makes a lot of sense. Yeah, it just increases. Yeah, it's just like constantly shifting goalposts. I feel like you try to increase productivity, but then you're stuck with compliance or you know you make sure that stuff that's getting pushed is actually even stuff that's needed. Because a lot of my listeners are you can think of them as kind of being on the fence with a lot of these technologies. I am curious. How do you just approach.

Like, so you know, so suppose like somebody listening to us isn't even from tech. For these type of people, what do you, I guess what is your personal take around the doom staying that happens around AI like is it a bubble will it burst or is it actually useful. It will continue to even just grow more and more. I know usually it's just like two sides, usually polarizing that people end up on. So I'm just curious because you're actually in it and you work with it just day in day out and you're, you know, basically an expert at this.

How do you see these tools irrespective of target right irrespective of any other employer you work for but personally you saw me how do you feel about where this stuff is headed. No, I think this question has been on my mind for the longest where I have asked people who kind of are more in the thick of this technology than me and I also have built my own view and have answered in other places. So, like you said, there's a utopian view and the dystopian view where you know, AI does everything.

But again, if AI does everything, what a human's going to do and how they're going to effort it is one question I would have. So it's I kind of use the word example of nuclear technology. We have that power, but if that power is not in the benefit of humans, it's not in the service of humans, they are just, you know, boxing it. We have the nuclear capability, but we do not know the right way to channel it or use it unless for certain kind of electricity generation, we don't know how to even use this technology. So it's out there, but no one is using it, right? Because it's not in the human service at scale.

So even AI, the artificial general intelligence where AI could do everything, but end of the day, this plan is inhabited by humans. And we would like to still continue inhabiting this planet and live a more richer life, more happier, fulfilling life. If AI is not servicing that AI at an artificial AI, like where AI can do everything humans do scale. If it's not in our service or if you don't know the right way to shape it and use it for our own good, you might just bottle it because you know, yeah, I can do everything, but who can effort that who is maintaining that who's going to shape all this in human service.

If we do not have that, you know, a scale body that is kind of this and it's not in our service, we might bottle some of those ultra powerful capabilities because we just don't need them. So there is that view of the dystopian view. I don't see humans leading our own selves to doomsday. If it's coming to that, we'll shall wait, we'll bottle it, you know, we'll not let us let us get the point is what I feel. On the other side, the utopian case is like, you know, I kind of feel like there are a lot of places where AI can actually be helpful.

You know, I don't know if I get old and if I'm better, I don't know if in this generation, we are only having one or two kids. I don't even know if it's reasonable for them to take care of everything when we are sick as well. So, you know, if there is a humanoid robot that can help the elderly the sick, you know, when people actually need if a humanoid kind of AI capability can help us live in those environments in a better way.

I would love that technology to come to fruition and help me, right? So I am really scared of my old age in general and if there is a technology that would help me, I would like to have it. So more than just for coding, you know, or other kind of things, I would like that to happen. But again, they say that if you, if AI can code, it's basically AI can think and create an algorithm. And that's the basic foundation for figuring out humanoid robots or, you know, AGIS scale.

If you can think, take a small problem, break it down, build an algorithmic solution around it. That's the essence of solving any problem at scale, right? So when we say AI can code, AI is figuring out the foundation piece as of now that could actually lead for all these, you know, bigger problems to be solved. So I see it in service of those bigger problems. But in the interim, a lot of productivity gains and all are just like consequences, you know, where I can also do this and we are all leveraging for that.

But AI can actually solve a bigger problem when I'm waiting for that problem to be solved sometime soon. And we are just enjoying the side effects or consequences of it for now where, you know, I can create this. I can enhance this. I can draft the video. I can create another. All these fun things are just side picks up the bigger AI revolution is what I feel. Yeah, that reminds me of this meme that was recently been wild. Maybe not decent. It's probably been more than a year, but it was just that I wanted AI to fold my clothes and do my laundry while I write poetry.

But what has happened is I have to fold my clothes and do laundry while AI writes my poetry, which is the opposite of what I want. So that I actually kind of that room has been created a lot of creators. I shared one such version because I am a new mom in the thick of it where I need to do laundry every day watch bottles. There's a lot of I have to do, even though I have interest in doing all these higher order things, writing poetry or analyzing like a new sci-fi movie or even creating content.

I can do all of that, but I am like pulled back to doing all of this given I have a new mom and I'm like, yeah, I will take my job and I'm folding my clothes. So I'm wondering when you know I created that created that watch not showing a woman doing a household shows wondering when really I take her job actually. I come from that same boat as of now. That's awesome. And to also to touch on that, I know you've been on so many podcasts, you write on LinkedIn, you know, you're a content creator by any definition of the word.

So I'm curious, how if at all has doing that type of work just on your own time because obviously I can relate to that, because I know how much hard it is and how much hard work goes into consistently reliably doing these things, especially, I mean, I can't even imagine what I would do if I had a little human to take care of. So it's already hard for me. So I, you know, I don't even know what it's like for you. But I am curious, I guess, for folks that are in these companies doing these roles, going about their lives, working with AI, building with AI.

Why should somebody think about content creation? I guess what would be your motivations for putting in this hard work? And then, yeah, I guess what drives you to do stuff like this. Sure, thank you so much for asking this question because I have that question back to myself at times, but I also have these answers, you know, why I even started this in the first place. Given I kind of started working in a thick of it, even people like my friends and family who are in the IT, like they also working data science data edges and but not exactly.

I asked here by this technology, will AI take my job? It says it can code. I see this cloud code. I do not have not even used it, but will it take my job? So there are all these kind of fears among my own friends circle and given I work on it and I use all these tools to code and do all of things. So, those are looking for me from the answers. And people who have no clue about this technology are on the sidelines and I kind of feel like this is going to become mainstream especially, you know, those are always the marginalized community, especially more immigrant community or women or like even the children of the next generation. Sometimes because now we have not even figured out, they'll suddenly grow up and are out in the society in a place where they have not equipped to navigate this world.

So I kind of feel like, again, I'm not like the one building it or controlling it or have the influence to shape where this technology is headed, but at least I have some insight into how this is unfolding. And if I took it upon me as a responsibility to share that insight and kind of help people who are either scared about this technology or lost in this technology completely, you know, have no idea about this technology.

I cannot feel that responsibility upon me to kind of share my two cents to help these people navigate this world better. And also as a new mom, I had this feeling that how the world will be when my baby grows up. What can I do for her? What kind of AI curriculum school? Exactly. What kind of AI knowledge should I instill in my baby? How should school curriculums change? I have all these questions only, but again, there are very few people answering it or trying to answer this.

So I really am interested in understanding how AI should be incorporated in the curriculum of the next generation or even be as parents if you want to continue providing better life to them till they catch up. If my job is going five years, what should I be doing? How should I be repositioning or upskilling myself to navigate this world better? So those are the motivations. Like you said, right, I should be learning, I should teach my baby, but also for people who are out there wondering if I have something that I could help them, I want to help them in my own two cents way, nothing like a...

I don't think I have the power or influence to shape anything at scale, but you know, just my two cents to make people understand this technology. Yeah, I mean, it starts with one, right? Anything of any scale ultimately starts with just one person being exposed to an idea or a thought or reading one of your posts. So I can relate to that sentiment where it never feels like it's enough, but there is always that angle of the silent majority, which cuts both ways unfortunately because majority of the people that you're helping, you will never hear from them.

You won't even know that they exist. Actually, that's just the nature of the internet and how stuff works. But on the flip side, it's especially bad around what you were saying earlier that anytime there's a new model, there's just people engagement farming. Oh, this startup went out of business. This thing is now obsolete. They haven't even installed the thing, right? They're just out there yelling from the rooftops, literally just spreading misinformation.

So that's the curse and the blessing of the silent majority. But I guess my reason for sharing that is you're doing like incredible work and we need people like you. That's actually informed that puts in the work that understands how these things work to be our arbiters of truth versus these other nameless anonymous Twitter accounts that are against straight up just rage beating about these type of things, because that is just the nature of social media and we need that balance. So, yeah, I guess just wanted to add on to that.

Thank you for saying that I just didn't want to point anyone, but that's also one of the internal motivations because I see a lot of false information, misinformation, engagement, farming, all those things. And even when I was new to this technology, I wondered that I was behind 10 years when the technology was one year, so they make us feel like that, right? And I also could now leverage automated pipeline where, you know, clog generates something notions if you're posted.

Like, I could do all of that automation thing. Again, why am I posting that? If that's what it comes to as a new model, I could do all of that, just sit back and use some of the credibility that I am getting to kind of Englishman farm, become an influencer, whatnot. But how would that serve anyone or how would that serve me is the question. So I am trying my temptation not to resort to any automation, any points soon.

I'm using it for productivity gains, but nothing without my own idea inception. I am not starting anything. I am looking out for real world things to even get inspiration or idea and whole. This is what it is. I love something. This is worth saying. I wanted that to be the starting point rather than asking plot to generate 10 ideas for me to post 10 posts. I don't even want to start there. If I don't have any idea of my own, I don't even want to share anything out there.

And when I have an idea, it needs to look aesthetic for the web these days. So for that part, I definitely want to take some help. Editing or finding good stock images or writing more, you know, catchy hook lines because outright my thoughts are very technical and boring. That will not even take it. Sometimes I need to resort to some creativity somewhere, you know, some inspiration outside. So only for those parts, I am leveraging the capabilities, but not for the idea, sourcing or the full blown write-up of a time.

I am making sure that I would never get to that place. Even if I don't have time, I don't want to get there. Yeah, yeah, I can definitely relate to that approach. I will lie or I won't lie, rather. I have been inducted by the dark side where I have definitely shared stuff that I have not personally tried out myself. Just to, you know, give myself that position of, I think it's just authority, right? Like we all, I think as humans, it's natural to seek authority. But thankfully, what I did realize pretty early on in that journey is that this is not the type of creator that I want to be.

And so thankfully I was able to pull off from doing that and I no longer do that. But I, yeah, I think it's, I love that you call that out in terms of it is a very real temptation, right? Because there is no barrier. There is no harder to doing that. Like literally somebody that has never even installed cloud code can go out there and just have, just pray 15 crazy use cases that you just mentioned, like connect Google ads to any 10 that then there's this that then there's that UGC clip farming.

Again, it's just a word salad at this point, but it's not helpful. And that's my not star is that is this helpful, which I just, you know, hopefully this can at least inspire one person to not do that. And do the useful kind of content, which, you know, is kind of a goal that I have with this, with conversations like this. So glad that we are sharing this perspective. And like I said, that it comes to the now these days, it comes to that judgment of not to use.

Actually, that's so true use case to use AI for, you know, yeah, I love that. Yeah, honestly, I think that I'm already like my content brain is already trying to workshop that as the title of this entire conversation. Because in a lot of ways it does serve as the backbone for a lot of things that have that we have talked about here today but somewhere this has been so incredible. Thank you so much for taking the time.

I took away so much from that conversation. And I'm sure all of our listeners did as well. If people want to get in touch with you, what would be the best place to contact you I'll be linking your LinkedIn in the show notes. But other than that, if there's any space that you would want to prefer to be contacted, could do mind sharing that with our audience. Sure. Yeah. Thank you so much for having me. I really enjoyed this conversation. Glad that we share a lot of views and you're also trying to identify the right use cases in the right way to use AI and thinking about the broader societal implications.

I think that's what I want everyone to be doing that on their own and build their own perspective, rather than just be flooded by everyone saying things out there. So that's that. Yeah, LinkedIn is the best place for me to reach out. Please do like that. I'm also available on Instagram. I'm not resisting the temptation to use AI for content creation. I'm only starting to put out. Maybe more work will be shared on Instagram. My Instagram handle is indigirl.ai. Unfortunately, I had to use the AI.ai because that's how I get found through searches is what I realized.

Yeah, I'm on Instagram. Awesome. Yeah. And I'll be linking that link as well for anybody that wants to check out your work. But yeah, that is everything I wanted to cover. Thank you so much for taking the time. Yeah, it is a pleasure talking to you.