Good morning. Good afternoon everyone. My name is Rajni k I'm a senior support communications manager at Snowflake. Welcome to this talk. This talk is about AI driven customer experience, how to start scale and succeed in the enterprise. Let's get started. Okay. So customer experience is rapidly changing. Like all these companies and enterprises were able to serve customers in a reactive mode until yesterday, meaning customers if run into any problem, they'll come to. The enterprise support portal, or they'll call the customer service, they'll open up a technical ticket or support case with the company, and then somebody will be looking into the problem after they happen and then they try to solve and help the customer. So that was the yesterday scenario, how the customer's experience was. And coming to today, in this, today's years and world of ai practically anticipating customer needs even before the customer asks, right? It's becoming increasingly common by, by leveraging AI tooling and the technologies. So yet these AI tools are delivering. Delivering, sorry, personalized experiences to millions of people simultaneously in real time across the globe. You take an example of retail or healthcare or banking and finance, whatever industry it is, they are using AI to deliver personalized experience at scale. So I'll talk about that in the upcoming slides and I'll show you some examples from different industries. And also, yeah, mixing AI with the human expertise is very important not to replace the humans. One has to leverage AI to amplify that kind of an expertise from the human side of the things. So that's that, that's how like the customer experience is changing in today's world. So here are the three pillars of like artificial intelligence driven customer experience. The first pillar is hyper hyper personalization. Like tailoring experiences to individuals based on their preferences, behaviors, or context. For example, if you take any e-commerce business so like Amazon or. Any other e-commerce retail, so they know like what the customer's browsing history was. They know what the customers purchased during what time, what they bought in the last winter, or like what kind of a colors they like, what kind of a models they like. So they know everything, right? They know everything about their customer. So by leveraging AI on top of the data, whatever they have. They can personalize everything to the customer. That particular customer, not 10 thousands or a hundred thousands of customers. It's personalized to that particular individual who is browsing during that moment, during that day, during that festival season or holiday season, whatever it is. And the second pillar is like intelligent ent. These AI powered reagents, like they, they are very capable enough to try solving complex problems. They understand the patterns. Even if you have. Mountains of data, they know how to, dig through the data and come up with like solutions with, for those complex issues with mixing up with human empathy. So these are going to be very powerful agents, these AI agents. And the last one is predictive insights, right? Anticipating needs and resolving any issues that may happen even before customers notices them, right? For example, in the, i in the telecom industry if the company companies are using AI to proactively monitor the signal drops. Or if it is in fiber internet or something, they even. Getting to know if there are any pockets, drops or something like that. And then they are proactively fixing those issues even before the customer noticing them. So that's pretty solid pillar. For the, aI driven customer experience. Alright, let's go to the next one. And the next one is an example from the retail industry. I talked about this a little bit in the previous one as I said if you take a real world scenario as soon as a customer walks into your store or they started browsing online. Shopping online. So AI recognizes who that customer is and what their browsing history and what kind of purchase patterns the customer had in the past. And then they'll there, and then they'll provide those recommendations based on all these points. And the second one is the AI will enable dynamic pricing for the customer based on their individual preference and demand. It's not like a. Generic solution or 20% off across all the customers or something like that. But it's basically pointing or categor like delivering the pricing strategy exclusively designed for that particularly individual during that day, during that time. During that holiday season. So that's the hyper personalization in retail. And the third one again, as I said showing product recommendations based on the weather. Hey, this customer has purchased a winter jacket during last year, during November. And if the customer is browsing during November, but this year around the same time, probably let's show him like some different winter jackets based on his color references, based on his choices that he made in the past. So that's hyper-personalization that we are talking about, and this is all made possible because of, like using AI into the regular workflows. So that's the power of AI and coming to the hyper-personalization. And if you take finance or healthcare, right? So finance financial services are like companies who offers like investment plans. They are basically using AI to tailor. Tyler to provide a tailored approach based on the risk tolerance of the individual based on their personal goals. Based on the, based on like their age, where they are in their life during wall stage. And, like it's a powered personal financial advisor for that particular design, exclusively for that individual. Based on these, all these factors, a 25 year world startup founder. Maybe looking at the different options compared to a 55-year-old person who is nearing retirement. So the system understands using the ai about the, about all these like data points, and then comes up with the, like financial plan, investment plan exclusively designed for that particular customer based on all these factors. That's about finance. And if you come to healthcare, again, the same thing. AI is powering, delivering care at speed based on the medical history, previous, past episodes of the customer their references, their demographics, or if they, based on, even based on their literacy level. Because some of the educational materials they are very important for the patients, let's say. If your patient knows that, if they're the first time diabetic patient, they wanted to understand what are the, some of the things that they have to take care and what are the dietary plans and all this kind of stuff. If you just hand out your education. Educational papers or something like that, that they have to go through, it would be difficult for them to understand. So that's where the AI can help you. Like based on all this information to serve patients better based on all these, like inputs it's exclusively like patient-centric. I, as we know it's not like one size fits all, right? It has to be. Category. It has to be specially designed for that particular patient based on all these factors that we talked about. So that's another example. In the healthcare we talked about retail, finance, and healthcare. So in the next slide, probably I'll talk through the case study that we work at Snowflake for processing thousands of customer satisfaction feedbacks that are coming in. So the part one in this particular slide I'll explain the challenge that we have faced, and in the next slide I'll talk about the solution that we have put together. Okay. The challenge was, as I said as soon as. Like a support ticket is closed. We ask, or we send a survey link to our customers, ask them to provide their feedback, right? Based on the service that they have received. And as part of the survey, we are providing a open text form, open text box, asking them to provide their response whatever they're thinking about the survey, and if they would like to see any improvements or if they wanted to understand why certain feature is designed in such a way or something like that. Or even if they wanted to say thank you to the customer engineer, whoever helped them to fix the issue. So that's the kind of open text box that we are putting in the survey form and asking our customers to fill the form as soon as they have some time. So we are receiving definitely responses thousands of responses are coming in every week. And we are a we have a team of three and manually going through each of those responses and categorizing them by sentiment like positive, negative, or neutral. And also mapping them to issue type. Like this is a product feature issue. This is a service quality issue. Or this is a technical issue, something like that, right? So we need to bucket those feedbacks into various issue types so that we can have some kind of a closed loop process so that we can fix those issues and improve our overall customer satisfaction. So manually categorizing those thousands of feedbacks that are coming in and putting them into various sentiments and issue types is a very tedious tasks. Literally it takes weeks. If we are receiving thousands of feedbacks this week to process those thousands of feedback manually, it takes weeks for a team of three or team of anybody to to, do that kind of a work. So of course, like the teams were burned out doing a repetitive heavy manual data entry kind of a work. It's not going to scale out, and it's by the time the feedback reached to the particular team it's already scaled out, right? We cannot really make use of the feedback. So that's the problem that we, we dealt with in my company where I work right now. And let me show you the solution, right? What's the solution that we came up? Of course we used the ai but what we did was like, we plugged the AI into the whole workflow, as I said, like we have a platform survey platform that is getting all the survey responses and we are feeding the survey responses into your snowflake table, capturing that information into your snowflake table. And then we have plugged in the AI deployed AI model. To read all those, like feedbacks and at scale and classify them into sentiments, into issue types and everything right. At scale. We don't need to manually do this. Yeah. The workflow will automatically. Call the ai model a AI function, and then it'll process all those feedbacks we, which were came in the last week and automatically process them. So what's the result? Earlier it used to take weeks, as I said, like to process those thousands of feedbacks. Now it is, AI is doing it in overnight, maybe within hour, within few hours. So that's a tremendous change from the effort standpoint, and then it's, it is definitely saving millions of dollars by, like eliminating that kind of manual work and freeing up the teams so that the teams can free up from that kind of a robot type of work. And better utilize that time in strategic problem solving, building relationships, and working with cross-functional teams to actually implement the feedback that we are receiving from our customers. So what's the end result here? Of course, we have an improved csat customer satisfaction. And also pretty faster response times, right? We don't need to wait for weeks and we can solve those issues pretty fast. And also happier teams, right? So even like the internal teams are pretty happy because they get to see the customer responses, customer feedback in any faster time so that they can act on it. Then they can close the loop. That's this kind of a solution. So again, like we didn't build it from the scratch. We used our existing workflow and instead of doing a manual categorization, we have we have plugged in an AI model a classification model into the existing workflow, which is like reading, feeding sorry, reading the data from a snowflake table and processing it and putting it into another table. And then that table is of course having all the information, the sentiment and the issue type, categorization, everything. And so this entire workflow is scheduled to run every week. We don't need to do anything every week. Midnight Sunday night, it kicks off and it starts processing all that information by Monday morning as soon as the teams are in, and they will be able to see all the information and start working on those. Yeah, this is very useful. Our teams are happy, as I said, like we are using it. In our company right now, and it's pretty good. It's going pretty good. And like even when it comes to in this slide, right? Even it comes to we, we all, at some point in time in interacted with chatbots for sure, right? These chat bots, right? The traditional chat bots, more or less are designed to. Answer frequently asked questions based off of script based responses. Something like, what are your store hours? Or what's your written policy? Or something like that, right? So immediately there will be. There will be an answer available based on a script. So it's a traditional chat bot that we have that we are we have all seen, and interacted with. But AI has changed that game as well, right? So the modern AI agents are so powerful and they're pretty good at solving complex issues. And, like understanding the context, the full context, what the customer is talking about, and they can also signal the customer frustration and like they, they try to, empathetically respond back to the customer based on the questions that the customers are asking. And even it can pull in a human, right? If it cannot solve. The problem there and itself. And it'll provide the full context to human and then the human can pick it up there and there itself, right? Without needing to go back, start over everything from the beginning. So that's the power of modern AI agents. So just to give you an example, right? If you. Look at any airline if you missed your flight, and then you are looking to rebook your flight, right? So if you imagine you have a ai chat bot or a agent. Available within your flight provider app, which will automatically understand your situation, compares prices and everything rules and everything. And it comes up with your new new flight booking. All. You just need to do an accept, click on accept, and then you're good to go. You are new boarding pass will be ready and you're ready to board your next flight. So that's a kind of an, a good realistic example that I can talk about other than what we have implemented at Snowflake. With the modern AA agents, you don't need to do anything. You don't need to talk to any person, any human, so everything will be taken care automatically. So I think that's where like the, a AI agent, short signing the agent AI is becoming so popular especially for these type of use cases. Let's go to the next slide. So this is where like I just wanted to touch upon various industries where they can start, where, how they can scale and succeed with the ai. For example, as I talk a little bit about, like the telecom or the internet companie. They can predict predict so churn even before customers leave, right? Let's say the usage is going down or, if the customer is not using their service pretty much any much pretty much, then probably, like it can practically. Send out personal retention offers through their, to their emails based on the behavior or something like that. So that, that is one thing. And also, if the, as I said, if the network is not working as expected let's say a customer has signed up for, like a hundred megabits per second. But they are end up receiving only 20 s. But they're not noticing anything. Major difference. Like the, these systems can predict and then they can fix everything back in the backend automatically, even before customer notices it. And also in the SaaS side of the things we know, like the SaaS is a very popular, like a lot of product, a lot of companies are spliting up their SaaS products. The AI can help you. Identify feature adoption gaps. And also, let's say O one SaaS company have 10 different features, the customer is only using one or two probably, like the AI can automatically send a personalized tutorial based on the customer use case or the customer environment. And then they can explain like how they can use other eight or nine features, which the customer is not using as of today. And also it can help respond to customer proactively even before they come and open a support case with us, right? Support ticket with us. And when you take a look at the manufacturing, the predictive analytics kind of a. Use case for the manufacturing industry is basically anticipating equipment failures. Scheduling, maintenance even before. And they go down IOT sensors by using IOT sensors and like the systems will predict and then they'll send in alert. To the technician, and the technician will automatically take care of the, of those, missions missionary before, before they go down. And also as we all know ai is helping financial and all companies, maybe mostly this is applicable for all industries, right? Where payment is payment processes involved, so it can automatically detect payment issues and issue auto refund in case of any failures or something like that. And even communicate before the customers can complain, saying that, Hey, your transaction is failed. We have automatically refunded every amount back to the original payment method or something like that. So that's the power of predictive analytics, and that's where the companies can start scale and succeed without, needing to put too much of manual effort just by tapping on the AI and including AI into their regular workflows and adding human in the loop. All right. Let's so what are the, some of the ethical challenges or what are the, some of the challenges that, that anybody or everybody can face during this whole journey? Number one is and also I'll touch base a little bit about how to navigate them. Number one is like the data is always residing in silos. So that is the biggest challenge that they all these enterprise companies have, like they have their CRM systems, they have their customer transaction systems, they have their royal TP systems. Maybe they have their like data warehouses, data lakes, so that the data is everywhere, right? So we need to invest in, how to overcome the challenge is simple, right? We have to. Invest in, in data integration and a good data governance before plugging in yay AI on top of the data. So that's the number one challenge. And and how to overcome that is is having a good data governance and data integration in place. And the second one is by us in training, right? So if you train your model only on some type of data. It'll always try to, bias us towards that, right? So you have to ensure that your training data is well diversified and also you need to audit data for the training data for data quality. Are there any issues or is the data, is the good data or does it have any bad data? Missing emails or missing missing, like some other information, which is very key for the AI model to predict and take action. And also how to test models for fairness across, different segments of people, different segments of, like geographies or something like that. So that's one thing that we have to be very careful. And the second one is and the third one is like our personalization is sometimes it causes. Too much of a problem for customers. We have to be always transparent and we have to tell our customers that saying that, Hey, we are using AI to provide you a better service. If you are not into it please opt out. From that, something like that, right? So we have to provide better controls to, to our customers. And also adding humans in the loop for any critical decisions is always best. Even though AI is doing a lot of heavy lifting, that shows some accountability on our side. Which is very important in this slide. So these are the five best practices that, that I can, and that I can share with you all. The number one is of course data is first. I always put data at the first because, like quality data foundation even before building your EA models, training your EA models or testing them. So you have to always have your data, quality of data available in a commonplace and also. You need to integrate AI into your existing workflows. You don't need to, create or rebuild everything from the scratch. Just identify areas where your teams are putting a lot of manual effort and see naturally v AI can be a best fit in that area. And then start including ai there, right? So you just build a AI model or try to plug in your s and then input your, data into the LLM and it'll automatically, start taking care of that heavy process lifting or something like that. And of course, you have to the third best practice. You have to always. Add a human in the loop. As we know, AI tends to make some mistakes, so it's always good to have some human oversight and feedback into the model. The fourth one is, like measure continuously. In our previous example that I have shown you, like csat. So we have before and after, and also we try to see like how the CSAT scores before and after, how the resolution times are looking, like, how the retention rates are looking like, so you have to track and iterate on the same thing every now and then. And the last point here is the, transparency, as I talked about it in the earlier slide, you have to be very transparent with the use of ai. So you have to tell your customers that you're using AI to improve their overall experience. It, it definitely builds trust and you have to provide them options, right? If they don't want to use ai, they can they should be able to opt out of that. So that's the fifth to best practice. These five best practices are a very important, and we have used these five best practices in our day-to-day regular work at Snowflake building the, especially the examples that I have talked about. So what is the path forward? So the path forward is pretty simple, right? So these companies are, that the leading tech companies are using ai for serving better providing best customer experience. Come up with like the features that the customers are asking for, improving their overall products and services. So those are the winners. These winners aren't replacing humans for sure. With the AI automation or A TKA are using AA in regular day-to-day life. But instead they are enhancing the overall experience by adding AI into the regular workflows and having human connection with with all these intelligent systems. One, one small thing that I can suggest definitely to everybody, whoever is listening to this conversation this talk you have to start small. You don't need to build everything from the scratch, as I said, see observe your existing workflows and see where you can include ai. So that's where like you have to measure your impact and, and also see that if the scaling also works from if AI can best serve that kind of a scale always keep the human in the loop and feedback to the model. Thank you all so much for. Like joining in if you have any questions, feel free to reach out to me and you can find me on LinkedIn by my name Rajni K, and you can connect with me. Thank you so very much, and you happy? You have a great conference. Thank you. Bye-bye.