Good morning, everyone. Thank you so much for being here.I cannot be more excited to cut this off, and to be in this virtual powerhouse, of Women in Tech Global Conference. A bit about myself, my name is Rachna. I'm at Oxford IIM, a woman podcaster, author, a lot of things, but most importantly, I work with Cognizant as data analytics and AI elite. Over the past fifteen years, me and my team have multiple large, scale digital transformation programs and the industries you manage to watch at the industries where you purchase from at the industries which you cheer for. So from consumer goods to retail to sports, that's that has been my, ramen. And overall, while I have been working in the space of analytics and AI, I'm still very much obsessed with data.

And, hence, today, we are going to dive in this brutal but beautiful world of data talking about the data, small patients, and now that's passion. What do I mean? What can be a better way to start off a women tech forum with a little bit of fashion talk? So we're going to dive into the AI trend from a fashion hat on and see that what exactly we need to do and why probably ABC of AI starts with, so without further ado, let me set the stage. As I said, it's a it's a best way to start a women's forum, have with having a little bit of talk around fashion. And just like any other fashion trend, trends come and go, in fashion. And similarly, AI is the latest trend as we all know. Is that is AI going to walk the runway and own it, or is it going to trip on the data? And why I say that?

It is because currently, every boardroom wants intelligent, AI. Every investor wants agent take, autonomous. The dreams are unlimited in the boardrooms. The dreams are unlimited in the investor communities, but the reality is still the same. Our data is still largely outdated. It's still very, very messy. And at the same point in time, there is barely clean data for AI to be trained. So on good for AI, let's try to dive deep into what can be some of the real life challenges AI would face and is facing and how data is very, very essential, to actually have any impact derived from AI. So onboard the hype cycle with me, and this is not my data, but, as per 2024 Gartner report, almost 80% enterprises are investing in AI technology heavily, which seems to be a very good thing. The problem is only 20%, and, yes, 20% are prepared with the right data.

That is one in five company, which is actually investing in AI, is actually prepared with clean data to support the AI function which they're trying. They don't have proper data governance frameworks around it. They don't have data clean enough for AI to be trained. And another stats, again, for 2025, it is said that 85%, again, 85 of the companies which are investing in AI and currently and next two years are actually going to have errors and results. And that's not just scary. It's absolutely tragic that you're investing in AI, and you're going to have errors and results 85% of the time. And how does that happen? That happened because we're trying to address our AI in this absolute fancy wardrobe without checking the fabric, which is the data, which makes that AI. Now we know a thing or two about fashion, and we know that nothing is better than a tailored fitted clothing. And what is the tailored fitting? That's a fabric.

That's the data. How well we have trained our data? How well we have cleaned the data? How well we have governed the data? And that is where a lot of onboard the AI hype cycle, there are real risks of it not being what we want it to be. Now, what happens if AI is real? Has it ever happened? Well, many a times it has happened, and some of these examples are very public, but I just want to bring it to an attention. What happens when AI goes absolute rogue? Yeah, that's the word. What happens when AI goes rogue? Well, Well, quite a few instance. One, for example, I'm sure a lot of you would know.

Microsoft, when they had this chatbot called Tay, back in days, the idea was that it would learn from the users, but in sixteen hours or seventeen hours of actually, being implemented, it started, putting racial slurs and misogynistic, tweets just because we know that it is learning from Internet users, and Internet cannot be the most classy place.

Another one, again, a very public one. IBM had Watson, as a diagnostic tool, which was when they launched it, it was considered as one of the most prominent technological advancement in the cancer treatment. Soon enough, it started recommending, unsafe practices, treatments. And why was it happening? Well, the root cause was that because the training data was largely hypothetical. The training data was not, real patient data, and the hypothetical data had its own biases due to which the diagnostic tool, which started with a lot of promise, did not perform well. Another one, imagine it's a scorching heat, and you just go into your Instagram feed, and some companies suggesting you a green coat or a good winter wear. Sounds bizarre. Right?

But this has happened, with as also when they have tried to launch their ecommerce chatbot just because of the wrong training data and the mismatch of their inventory data and the API is coming from the weather data. They started recommending the chatbot started recommending winter coats in summer. And not just that, in that quarter, just because of this measure, they had almost seven to 10% decline in customer satisfaction, and, of course, it impacted, the orders as well. And now you might be wondering that why in a nice morning at the beginning of, a tech forum, someone is talking bad about AI. Is it all bad? No. It's not all bad. What is the problem, and what do we really need to do is to solve a data dilemma. Let's put it. I call it a data chaos. Now what is that data chaos?

As I said, you can also only have your model going as smooth as on run rate as good as the fabric is, and the fabric is data. Or AI is going to be only as good as our data is. AI is a reflection of the data, and what happens is while everyone is focused on having that fantastic model of AI on the runway, no one actually has given the spotlight to data. When I say spotlight, is attention and rightly so investment as well to sort the data max. We're not doing spotlight to data because it's a messy blob to solve. There are a lot of silos. When I say silos, I'm sure. And I can bet my life on this that every one of you has certainly one file in your computer which says final underscore version five underscore version two, really final dot x l s, something like that.

That's how it is. Data is created by humans. It is entered by humans, and then it is duplicated by systems, by processes, and absolutely ignored, by budgets. And that is where data set in silos. There is inconsistencies. The information which is much more outdated. And what happens when we train the AI model on this data? Of course, garbage in is still garbage out, as cliches as it sounds, but garbage is and still garbage out. The The only difference is with AI, we are now teaching that garbage how to talk, which, you can think how to talk. So it is very essential to give data the priority, the budget, the investment, and the time to solve and be ready for it to stitch a beautiful fabric, which can then create our AI. So that's where we are. Now, again, I don't wanna sound all negative about AI.

Art of possible with AI is is is there a potential that, AI can be a supermodel on a runway? Absolutely. 100%. There are quite a few styles, but some staggering ones, for example. Zara, one of our favorite fashion retailer, they did something amazing. What they did was they merged. They used AI to merge, that POS data, which was a point of sale data, their inventory data, and the social media trend data and then created a model to help them reduce the inventory stage and they could reduce it by 40%, and that's a staggering number. It also not just help them reduce the unused inventory, improve margins. Another one of a football club, Manchester City. Everyone would know it. They implemented AI very strategically into the player performance. They merged their player's data, GPS data, performance stats, all the, body stats, all sort of data, and the aim was to reduce player injuries and peak performances.

When they could reduce injury by 26%, they could enhance player performance by at least 12 to 15%, and that's a massive win where the margins are very thin in elite level sports. Financial services, everyone loves it. JP warned me cheese. They deployed AI on structured transactional data and unstructured customer complaint data to actually identify compliance risks. What they could do with that is they could automate almost three hundred thousand hours of manual mental work. Absolutely. Manual mental work, I would want to call that because that's the work which is done by humans. It is hard to find compliance issues, and that's where AI has been phenomenal. So, again, can AI be the supermodel? Of course. If you're gonna keep off, amazing data, it can be a supermodel on the runway. What is important is what sort of data is needed?

Are you clear as an organization on what you want your AI to achieve? And most importantly, are we giving enough attention to what we are feeding the AI in? Now that sounds all nice and good, but how do we do that? Well, that's that's a question worth asking. Just like every fashion designer would want to build their wardrobe, I would suggest a framework, and that's a framework which would probably go very well with our theme. It's called dress framework. You can call it anything. But idea is very simple here. Post to diagnose, when I say diagnose, what do we need to diagnose? Is there weeks and weeks of data quality maturity assessments and weeks and weeks of discovery phases? Yes and no. Yes. Because it's very important to know what your data what is the data landscape currently, what are the gaps, what are the silos, what are the quality.

Everyone wants to invest in AI. Very little organizations really want to understand what is happening in the data phase. So that is is important. Not necessarily you bury yourself in the data because data can be a really messy place to worry. Be very, very clear what is the aim of AI and then diagnose all the data landscape, which is going to feed into that AI. That way, you'll be very focused that what you want to identify, what you want to diagnose, what does the discovery phase looking like, and then just identify what are the gaps in those data. and very important, in lot of organization I see there is no clear responsibility, which is who is responsible for the data governance. When I say data governance, it's it's a whole lot of things inside it. We have privacy issues. We have governance. We have quality issues and whatnot.

Who is responsible for overall governance and how they are going to feed in the AI of our dreams? So it's very important to set responsibilities. educate. I cannot stress this enough. Lot of tech organization lot of organizations, tech wing of the organization, really want to do some amazing work in data and AI. Well, actual reality is that business literacy data literacy is very low, and you cannot have it Because as I said, AI has to be everyone's job. It cannot just be a tech job. So everyone needs to everyone in our organization, every department needs to have a data literacy to understand that what is feeding that AI and what is possible and what is not possible. Also, recently, I realized that there is a new trend in the market. There are usually now separate departments. There is a separate data wing.

There is a separate digital wing, and the text sits outside of it or different of it. How does that work? That does not work. Data has to be something which is a function which cuts across your entire organization. It has to be something which cuts across organization, and it needs to be working very, very closely with the tech and digital teams or units to make sure that everyone from business perspective as well has a very good data literacy. Simplified. Again, we do not simplify ourselves. As I said, for one tech function, we have three to four different different teams, three to four different different silos. It becomes very difficult. And, hence, a normal business function like HR or marketing would actually be intimidated with data governance rather than want to help, into governing the data. We need to really simplify the structures, and there is a lot.

As you can see, there is a lot we can we need to do before we can actually start building the AI and then last, of course, escape. Why scaling is important? Because I know an organization which has almost 8,000 RPA bots have almost 30 conversational AI chatbots, and they are in process of creating five another AI. That sounds scalable. It's important to have an AI strategy and then invest into different different business functions where AI can be implemented, can be scaled at a larger, rabbit. Wardrobe, and that's how organizations needs to approach it. Now, one thing which I want to make very, very clear here is that AI cannot be just a tech job. AI cannot be just a data job or a digital job. It has to be everyone's job.

So your HR, your marketing, your finance team has to be there while we're having discussions about data and AI. They need to have certain level of data literacy. And, of course, every product team needs to have data governance in that product road map, not in the silos. AI can just not be a tech job. So what is the runway ahead? Of course, we would want to finish off with with some good thoughts. Is there hope? Of course, there is hope. There is no question. See, just like any other nineties fashion trend, AI is not a trend which is going to go away. AI is here to rule. So we better make peace with it and be ready. What is important is AI cannot move this, runway alone. We need to have pairing with AI. Human and AI need to pair to have a really perfect catwalk. Now some of the leads which is going to rule the market, explainable AI.

AI which can explain the decisions in human terms, who is logical, who gives reasoning, agentic AI, something which can explain and talk in a human language, Explainably, I would be absolutely there in the market for a long. There is certainly a lot of effort needed from a data perspective and more and more, we would go into this AI bandwagon. More and more, we would require the data and the frameworks of it. And last, again, I cannot stress enough. There has to be business integration of AI. AI needs to be standard in every business process in your finance, marketing, and supply chain, and everywhere. It cannot be just a siloed tech function. So let's just try to make sure that a quick question. What do we wanna choose? We want our AI to be fast and flashy, you know, fast fashion, or we actually want our AI to be smart and classy and probably fashionably late.

I would choose the fashionably late because what we want is not just trendy. It's certainly a timeless. So I'd leave you with that, on the closing thoughts that not just about making statement. It's about making right statement. AI is the need of an hour and by all means organization must invest and to have better. We onboard the high train between to self prepare that are we ready, and then we can process runway. That was my time. Thank you so much. Feel free to reach out to me, during this forum, or this is my email address, for any future conversation. Always up to have discussion. But thank you so much, guys. That was my time.