Hello, everyone. Uh welcome to our session. My name is uh Vafa Ahmadiyeh. I'm one of the AWS Principal Solution Architects. I'm working in uh global financial sector. I'm part of a large team that we are looking after JP Morgan Chase uh with their migration to AWS.

I'm um honored to have Amy with me on the stage from JP Morgan and we are, we have created a fantastic talk to go through JP Morgan Chase's journey on how they are using AI/ML services to improve the productivity in the contact center. So let him introduce ourselves and then we get through the talk.

Hi, everyone. I'm Amy Ellenberger and I'm the CTO for Machine Learning for Operations within Chase. We focus on predictive modeling. We focus on NLU and NLP solutions and really contact center optimization, which is why we're here today. So let us walk you through this journey.

Customer service is important. We all know that customer service is important but how important is it really? Let's look at the numbers. On a PwC survey that we have done, 75% of the people said that having a good customer service is a really key factor for them to choose a provider. I personally am happy to pay a bit more every month to make sure that either is my bank or is my phone provider or internet provider. I'm happy to pay more so that when I've got an issue and I'm grabbing the phone calling the customer service, I receive what I'm expected.

But the more interesting number is that half of the people, or almost half the people, they will turn around and find another provider after just one bad experience. That's, that's quite astonishing, that number. But on the positive side, 85% of the companies are telling us that they actually can improve their customer service through the call center when they introduce technologies and when they are bringing AI/ML to improve the productivity of their agents.

So we are going to dive deeper into these. If you haven't done this at this Re:Invent, we are trying something new. So uh we got a polling. You can grab your phone, scan this or you can go on AWS Re:Invent app on the right hand side, click on More. And the third is a live polling. Just put that number 307177 and let us know which of the following is actually your contact center challenges or your customers are facing. I'll give you a few seconds to see how the room is feeling about these challenges.

Ok. So all of them basically, that wasn't surprising. Um when we were doing other sessions, it seems that these are the challenges that most of our customers are facing. I think in equal number, 32%, 18-25%, that with your existing call centers or with your customers, let's move on.

So, um what you can introduce with the AI/ML to your services is to basically not only tackle the time that the customer is on the call by two ways. One is that you want to transfer the call to the right agent. I think at Chase, we call them specialists. Is that right? So you want to find the right specialist, especially for example, for a large organization like Chase. If I've, if I've got a credit card problem or a debit card problem, there are maybe two different specialists. So while I'm calling the call center, I want to talk to that specialist first before going through the department to department.

Not only that we want to make sure that we empower the agent to resolve my problem much, much quicker. So the other issue is that we find the result is high agent turnover. But introducing AI/ML, you can actually make their life a bit easier. You can, what you can introduce is things like chatbot or self agent assist to answer some of those questions earlier or having a good knowledge based system to bring all the information that that agent needs to be able to resolve the problem much, much quicker and obviously got tons of data that you can find out what are the areas that you can improve in your customer service.

More importantly, cost saving, you can do a lot of cost saving when you introduce AI/ML. Let's dive deeper. At Amazon, we got two options. Amazon Connect, I'm sure almost maybe all of you or majority of you, you know what Amazon Connect is. It's an easy to use cloud contact center with built in AI/ML in it. So that's one solution.

Um if you have worked, I have worked in the past with designing call centers. Sometimes it takes six months to 12 months even to just design networking and everything else with the call center. With Amazon Connect in five minutes, few clicks, you got a call center. Literally, you got call center that each of you can use. You can just go on your AWS account, if you don't have an AWS account, create one here, just have Amazon call center within five minutes.

The other option that we have got is what we call CCI. How many of you by raising hand heard about CCI? Ok. Not many but quite a few. So today we are going through the CCI solutions or AWS Contact Center Intelligence.

What we heard from the customers is obviously, Amazon Connect is a good service for really easy built contact center for many, many of our customers. But many of our large enterprise customers like Chase or many other financial sectors, they already have got a call center, either it's Genesys or Cisco or anything else.

CCI is actually a combination of the services that we created the solutions to integrate to all of these, no matter what call center that you've got, you bring AI/ML services into it. So CCI is not a product, it's not off the shelf product that you buy and it's not a new service. It is just a combination of the services that we already created some solutions because there is no one t-shirt fit all with every enterprise customer that we have got, got different needs. So we had to make it flexible enough that you meet your goals with the CCI solution that we've got.

So there are three use cases with CCI. Self service virtual agents, that for example chatbot, we talked about it or IVR solutions when you would like to resolve the customer issues before even you transfer the call to a live agent. On the right hand side, you got post call analytics. So that's where you are going through all of your data, either is for a month or for the whole year or for the week. And you want to go through all of your data and see what are the common problems, where are the areas of improvement where I can focus, what are common problems with my customer, what are the common problems with my agent that I can improve.

So we got post call analysis CCI solutions, we got self-service virtual agent or the middle one, which in this talk, we are going to dive deeper into that, which is real time call analysis and agent assist.

Behind the CCI is basically these AI/ML services. If you need to create an IVR, if you need to do a conversational AI, you will use Amazon Lex. Obviously, as soon as you try to do IVR, then based on your business need, you need to do some text to speech. That's where Amazon Polly is coming. If you are in an environment that you need to do multi-language, obviously, you need Amazon Translate. So that's come to the picture. You can think of them as Lego boxes and we created a solution to put them together.

You might say that, well, I don't need a translation, but I do need this and that for example, a knowledge base, you might need a knowledge base to help your agent. I call the bank and say that I've lost my credit card. Then you can immediately tell the customer that, ok, if it's to do with losing credit card these are the things that you can tell the customer for that. We have got the Intelligence Search, which is Amazon Kendra that you can use for your knowledge based system and Amazon Transcribe.

And I'm so happy that we have got one of our senior product managers from Amazon Transcribe toward the end of this talk. We are going to dive deeper into the architecture and we have a Q&A to answer your questions about Amazon Transcribe as well.

So, and we go through the use case of the JP Morgan Chase.

So if you need to do speech to text, Amazon Transcribe is one of the services that you are going to use. Who has heard about TCA Transcribe Call Analytics? Ok, perfect. That's actually an announcement we had yesterday. TCA was enabled Transcribe Call Analytics. We had it for a year or two and it was doing fantastic things for you.

So with TCA, what it was doing before, it was doing the batch. So you would put all your call recordings into an S3 bucket and then you were running a job against that S3 bucket. You would say that ok, these are a month of call recordings that I got and out of the box, they will give you three things, which is really, really important.

So if you wanna know what is the main issue of each call, you cannot go read all the text or listen to all your calls. So one of the things that TCA will give you is main issue, what is the main issue with that call that gives you out of the box?

The other thing that TCA will give you is which is actually very, very good for improvement in your call center is that, has my agent promised anything? Is there any action that I told the customer that we will get back to you and then you can put them in a category and then go back to say that have you actually done what you promised to the customer or not? That's the other thing that TCA is giving to you.

The third one is a call summary. It gives you a call summary of each call. But more importantly, TCA has got an engine that you can a rule engine that you can create a rule. You can create as many rules as you want. I'm sure there is a quota limit. I can't remember off the top of my head, but you can create many rules to say that for example, I want to know all the calls that there is more than a minute's silence in it.

I want to know all the calls that for example, in the last 60 seconds, the sentiment of the call was negative. I want to know the calls that ended badly. One of the things that our customers using a lot is I want to know all the calls that the agent at the end didn't say "Can I help you with anything else?" I want to know all the calls, which is, that's one of the most used cases, that the customer on the phone says that I want to talk to your manager. You wanna, you wanna know those calls, put them in a category and then take some actions.

So this was existing in TCA or Transcribe Call Analytics. What we announced yesterday is you can do that all live. You can do that all in real time now. So imagine a call center. I'm an angry customer. I ring and I say that I wanna cancel my subscription live. You can send it, you can flag that call and you can send your supervisor to go and help that agent before that call gets escalated. I'm gonna show you a demo at the end about this.

Gonna ask you another question. Let's have the phones ready before I hand over to Amy. I quite like to know what are these call centers you are actually dealing at the moment? Let's see where we are with the results.

Oh, wow. Again, we see that that's not unexpected. That's what we were thinking that we get today and that's what we are actually experiencing with most of our enterprise customers as well. In exact same, I've got one. Perfect.

Let's go back and I've got this is the last question. I promise. I'm not gonna do that again. So I'd like to know what you have got in the pipeline for next year. Let's check the result.

Oh wow, that's interesting because on another talk, it was a lot more on the self-service visual. It seems that that that's perfect. So we'll go through at the end of the slide, we'll go through how you can get help from what we have built in, in CCI solutions and how we can help you. But let's go through the fantastic journey that JP Morgan went to build live agent assist.

All right. So I like to start at the top. Technology at JP Morgan. There's a lot that JP Morgan Chase has to be able to do. We focus on scale, on reliability and security and then we overlay the ability to innovate using AI/ML as a part of our key investment strategy. This is across all pillars and it is something that we're intentional on.

When I break it down to Chase, we actually stood up a focused team a couple of years ago to start solving operations problems through automation and applying AI/ML. And the reason we're focused here is some of these key pain points that you see today.

Right now, Chase has 66 million households and 5 million small businesses that we run and service on a day to day basis with those customers. 62 million are digitally active. 48.9 million have our mobile app. These are double digit growth year over year, every year. I was gonna say every day, it's not every single day, but we are continually pushing new self service capabilities out into the market. We want to make sure that we're meeting the customers where they are and where they want to be.

Personal finances and banking become emotional. This isn't something that I can drop the ball on when they try to make a payment, when they want to do a transfer...

This is affecting their day to day lives and we have to be there for them. One of the other driving factors for the kinds of work that we're after is the fact that not only is money personal that people are using multiple channels, 50% of our customer base are starting online going to the app.

They, they interact with us in all facets, every channel they come into our kiosk, they're at the atm, they're in our branches and then they call our call center, bringing everything we know about them together allows us to personalize an experience and understand what they want to be able to do.

So with all of these digital safe service capabilities coming to life and new capabilities coming live, not every day but really frequently. How do we keep up and why do i still see 32 million calls coming into a call center? So let's take a look at the chase call center experience.

I don't know how many of you here are chase customers if you are. Thank you. I'm hoping to continue to earn your business. But while you're here, the first thing we do is put you through ivr. Sometimes it's great. Sometimes it's terrible, but we are focusing on improving using nlu and other automated prompts. And right now i can cut out probably two thirds of the calls by allowing self service and automation to take over in that space.

So once you're authenticated and if i haven't gotten you through, uh or self haven't delivered an ability to self serve in that channel, we'll connect you to a specialist, but the specialist will then validate your identity. But problem identification is the piece that continues to add swirl.

How do i know at which level of granularity? How do we be able to knit through everything they say, there's a back story for how they lost their card. There's a reason they need to be able to transfer things. What is the disambiguation we have to go through in terms of clarifying the right intent so that we're creating the right task and workflow for our agents.

And then finally, we wanna make sure that this is handled and dealt with and closed out first contact resolution is our goal. So after the call, we do a wrap up. Now, the the quality of the wrap up really drives the insights and the analytics capabilities of our teams, as well as the models and the accuracy of the predictive power that we have using the entire data chain. And not only just thinking of it as an exhaust, but thinking of it as a product, how can i consume it and make more, more value?

So what can i do with a iml again, self service automation? Just take it out. How can i make it easy for a customer to do what they wanna do? Uh second understand why someone's calling. This is, this does more than just making sure that the agent is ready. How do i feed that back in to the capabilities and feature function and our our backlog for the future roadmap for uh digital?

How do i make sure that our chat bots can handle these things? How do i make sure that live chat can handle these things? And how do i make sure that these are easy and seamless workflow processes? Omni channeling. Um one of the things we haven't talked about yet is making sure they have the right answer at the right time.

When you're a company, the size of chase, we do have a ton of different product offerings out there. How do i make sure that the right one with the right answers are in front of the agent at that time so that you don't have to sit on hold for 90 seconds while we're going through a, a 500 page document to find that one answer about that one. Apr about the third card that you have in your wallet, making sure that this is at their fingertips is key.

And then having accurate data. I talked about that call summarization capability. And we're very excited about the announcement yesterday. So a couple of years ago, we went through a pandemic, we had just launched this idea of we're gonna sta stand up in a i focused organization. We're gonna optimize the call center. Then the world changed and our agents started working from home hybrid environments two days in the office, two days back somewhere else and the ability for their natural ability to team work together and ask for help went away.

So it created this opportunity for us to really understand and prioritize the development of a virtual agent assistant tool. You can no longer just tap your buddy on the shoulder and say, hey, have you seen this before? Hey, which article do i need? What is the form that i have to fill out? How do i press that button? You can't hear the heated voice and tone of the customer and raise your hand so that your team lead can come and help you out.

So we decided to build a virtual agent. So what we wanted to be able to do is make sure the agent had the intent at their fingertips as fast as possible. We wanted to make sure that while we were doing that, we would automatically surface the article. So they didn't have to search more than once. We wanted to make sure that we had next action. So the right next service action to avoid the second call back and wanted to make sure that at the end of the day, customer mp s went up.

Now, it's also great if the employee experience goes up. So i created this tool and its side seats are virtual agents. So now when you sit there, you can see the transcript go on live, we are passing an information from the ivr. So you know what the customer thinks their intent is, we have the ability to pop up other things and it cues off of both our agents and our customers voices.

So you can hear chase speak and different entities and ideas will pop and the associated knowledge article will pop and populate on the right hand side, we are measuring success and we have a ton of experimentation going on right now. We're learning into what's valuable to the customer, what's valuable to the employee experience as a whole.

So you see some of the metrics we're tracking at the bottom, whether it's first call resolution, average handle time reduction in transfers. What are the things that we can do to improve the end to end experience for both sides of this coin, the agent and the employee.

So this is what we came up with. If you take a look, we've done um our audio is streaming in it. It was to be fair. The first time we did this, we did not use managed services. We built a ton of homegrown manual solutions that took a lot of care and feeding and optimization. And it took us a while to get it to a point where we were comfortable exposing this to some of our agents.

We learned a ton along the way and that's really what i'm hoping you guys get out of this talk. So yes, we were able to have customer audio streamed in real time. It was transcribed using our existing incumbent engine that we used to store a lot of our recordings.

We have the ability to after the call was transcribed, take it through some homegrown n lp models, extract intense entities, pop them up on the screen and then use that to queue off an elastic search query in the back end through our knowledge management articles. And in doing this, it created a lot of churn in, in the experience itself, these new entities were popping as we were refining some of these things.

The knowledge pain would change in the experience, although not optimal was progress. So i am a big fan of progress over perfection and you'll see that in some of the iterative development approaches that we take here at chase. Um the last piece is making sure that we had the ability to not only pop the intent, but if that knowledge article had some guided content that we could show them the scripting and do some coaching in real time.

What we found is that transcription and the accuracy of it was really important. This was the foundation that we were building a lot of our intent models, our entity models, our uh knowledge base, we were looking at call reason and volumes to even prioritize the things that we would want to be able to execute, contain and self service and we need to be able to get it right.

So we cut over and decided to start using amazon transcribe and we went live this summer. Um what we saw was an a approximately depending on line of business, 12% reduction in word error rate that is important for two reasons. One, no front line agent or specialist is going to trust a tool that confuses the words chase and cheese.

So making sure that these things are relevant and they could actually see and feel the improvement of the engine as it goes, creates a level of confidence in the tool if it understands and can see these things and it, it stops getting simple things wrong. I'm gonna have a better, better faith that it is going to give me the right advice, trusting a tool to tell you what to do, take some time.

Um the other thing we learned is that i i said in the beginning, we are very security focused here at chase. Not only do we have regulatory compliance, things that we want to make sure we're, we're handling, but personally identifying identifiable information is key. And i don't want to be the person on the front page of the wall street journal having messed something like that up.

So, um we had a very manual three step process to extract any pi i from a transcript before we could store it or use it or push it to downstream systems, it came out of the box and we were able to test it and hit our accuracy requirements that we have internal to the bank.

Um the other thing that i i as a an engineering lead loved was that because this was a managed service, although we did work through optimization, but i didn't have to have the same level of manual data wrangling. We didn't have the same kind of level of effort to try and pull all of these things together with our pipelines, our constant maintenance and support.

Um i was able to save three ft now that doesn't sound like a ton, but that does help me advance my u i. The other features i've got to do and work through and not just i was going to use the word baby head, but really focus on just one piece of the puzzle. So having a solid foundation that our agents believe in that we see improving that we have the ability to optimize, really started to unlock our capability here.

So i talked before that we wanted to be able to use these models um for inclusion or the the output of the transcription capability to go into modeling. But let's talk a little bit more of the why call summary, the way we use it today is uh we have an accom tool, takes all of our transcriptions and categorizes everything with the main call reason.

We also have some agent tools depending on the line of business where they select the main reason. And if it was handled or not, 40% of those calls are labeled general inquiry, that's not insightful. That doesn't help me figure out what i have to fix in a digital workflow. It doesn't help me figure out how i improve our knowledge management structure.

If it's about a payment, what was wrong with the payment if it's about a transaction, what do you need to know? How do i improve the customer experience when all i know is that 40% of my 33 million interactions are labeled general inquiry and that's 33 million a month.

So speed to insights. That's what i'm after. How can i change and improve the customer experience? How do i improve the support that i'm giving my employees? That's one of our keys. Now, paul summary is also an interesting beast because if i can figure it out from a call reason i can get more predictive.

I can catch them and chat. I can email them before it happens and i can learn more about next call. Avoidance. How do i get to a suite of capabilities that says while i have you, why don't we walk through and set up autopay? Why don't i help you set up some alerts? So you'll get a text message the next time your bill is due.

Why don't i, you know, show you how to use the app. Let's do this together. Every interaction with a customer at chase is an opportunity to deepen the relationship that we have. We strive to be the bank for all. I can't be the bank for all. If i don't understand the needs and goals and help you deliver it as a partner, that's what we're after here and that's what drives us every day.

So i talk all the time about data and needing data for modeling. These are the kinds of things that i can get through and justice features into our model and change the way we work. And lastly that more seamless. Um omni channel customer experience.

Has anyone ever been talking to a bot and then really needed a person and then ended up having them call and then you get put into the ivr and then you have to authenticate 700 times and then the person who finally answers the phone says, what are you calling about after you already told them on the chat and then told them in the ivr and then told the person a third time and then they say hang on, i can't help you. Let me transfer the call.

I never want that to be my chase experience. So that's what we're after. How do i seamlessly pick up a workflow and help you finish that transaction? So let's talk about knowledge management a little bit. This is exciting for us because our knowledge management um answers repository, i don't know how many other people feel this way. But my opinion is it reads kind of dry like an encyclopedia.

So when you're trying to get the right answers and you're sifting through all of these things, it's a lot easier to try to use a tool that allows you to get to sentence level accuracy of what is the answer? If i just want to know what is the apr on that card, i can get there using a tool like this.

So the curation of uh doing the elastic search took us a while um making sure that we have the ability to kind of get through and have relevant answers. Uh was a van a very manual task. We also needed to have uh the right set of training data and uh representative sample of all those intents that aren't frequently asked questions.

How do i answer those? Because that's what the agents need help on not the calls that they see every single day, they know how to do that. They can process the payment with their eyes closed. But what happens in those niche cases? And how do i get them to the right answer as fast as possible?

So we did a proof of concept with kendra and the experience was great. Um i gave the team and, and this is important. I gave my team a week, you have one week to hack this together. And let's just see what it can do without us over engineering and trying to make sure it's absolutely perfect. Let's just see how good it can be and what comes out of the box.

So in that week, they were able to plug it in, get the piping all, all done. Um but this is what i saw that made me excited. I had the ability to out of the box incorporate as feedback. We have that capability in our existing tool, but it's at the overall experience level like every call, did this tool help you? Yes or no.

This allows me to get just about the knowledge article. Is it relevant or not? And how do i make sure that i listen and learn over the course of time? Um it also had an out of the box filtering capability so a customer might have more than one account with us. How do i make sure i can toggle between them, i don't just make an assumption. They have one question.

They very often say while i have you on the phone, you need to be able to toggle that capability came out of the box. We also have the ability to add fa qs from other sources of, you know, any kind of um alert system. So instead of them going through an email, we could create something that said, hey, there was an event that happened yesterday and there's gonna be some people who are affected by it. Here's what happened. Here's the people who are affected and allow that to just be a part of the the the initial experience.

And like i said before, i gave them one week to get this going and the accuracy was really comparable and the team was really excited about the out of the box capability, but also the ability to optimize. You know, what if we enhanced our intent modeling? What if we combined this and kendra? What if we, you know, how could we take lex comprehend and kendra and make something really impressive. So more to come on this.

And then lastly any engineering team knows that your end user does not use your product the way you initially thought. Um so one of the things we're able to do is create champions

We have a model office. Now we sit side by side, we listen to their calls, we watch the tools that they click on, we see their struggles through knowledge, um management searches and what's gonna be relevant or not. But more importantly, they've taught us about the business. So I sit down and have them grade my papers. How, how are the intent models doing? What level of granularity are we getting? We can go through the ontology and the parent child relationships of the entities that we have built out and they will tell me you are great up to here. But there's a slight nuance between a stolen card and a lost card or the way we execute this versus this or why they can't get a new card right away because they changed their address or they did whatever. And when you go through the conversations with them as an engineer and as a data scientist, you can start to see the nuances and the things you need to pick up on to be able to optimize it.

The other benefit you get is that these model office users become advocates and champions. They're part of your development process. So now, now, not only are they thinking with you and saying, hey, i want new features that do this. They're saying to their friends part of this project and it's actually helping me in my day to day job. So let me teach you how to use this tool, which is half the battle in creating any new employee tools. Behavior change is the biggest hurdle that we have to do to adopt some of the capabilities in a imo so this is where we are today.

We still have the audio streaming in real time, but now we're using transcribe, we've had the ability to paralyze some of the the audio stream to not only hit transcribe but to be able to get those n lp models straight away and then have kendra kick off based on the entities and intents that it's extracting and retrieve knowledge articles for us. Alright, with that, i'm gonna hand it back to buffer.

Thanks amy. So this is the architecture that at the moment uh ev program is running on. So the calls are coming from the psdn. It goes to the uh obviously, I only put two boxes here, but you can imagine how complex is contact center of chase. It goes through a number of um fire firewalls and number of services, homegrown services as well as other third party solutions. And then we have got a server that we call it uh secure conversational gateway. What it does actually you got two options when you want to stream the idea more than two options. But one of the options that we we chose for ev, which i'm not going through that. Why? Because we had lots of experiment, there was to use the open source protocol called g pr c. Uh that at the moment, we are transferring the voice to aws another team actually at chase that after that, we've done that, they are using chime sdk and using kinesis to transfer the voice there. And we are trying to compare these two projects at chase to um to see that which one has got a better performance. Because as you know, the performance is the key you want to have this in the real time, you want to stream the audio in the real time, you want to get that answer back, cannot if, if five seconds gone, then the moment of helping the agent is gone. Um so the key every millisecond for us was matter when we were designing this and then it goes through the aws do connect, obviously go through the transit gateway. So the latency of the network was for us was um it was the issue that we were doing lots of load testing there and the stress testing there.

And then you got two a ks cluster. So the first one that the audio is getting a stream to is responsible to uh call the transcribe, get the transcription and uh we're using ms k because it needs to send it for lots of um auditing purposes and, and other uh places that needs to get story goes there. You call the t transcribe, you get back, you got the intent on everything that you already done machine learning before. Uh that was happening on prem nonage maker that you know, ok, what are the intents? Oh customer says i have lost my credit card and then you'll find the answers to that dynamic db or kendra, which is not at the moment because we are going live with the kendra uh at the moment the poc is done. Um and then uh there is a u i on, on eks as well that the agent can real time while they are talking to the customer. The things pops up here to help the customer. I said i've lost my credit card. These are the three links that will help you with, with, with customer. So that's the chase. But obviously it was lots of challenges. And for us, the main challenge was to do the low testing for the number of calls, about 33 million calls, 10 thousands of agents. So we needed to make sure that all of this infrastructure can scale up and down wherever they need it. When you got 1000 agent or 10,000 or 15,000. This, this should, this should do exactly the same thing. So most of our time was spent on the low testing.

So maybe this was complicated at the beginning when, when we started and and doing it. But uh going back to cc i solutions as i was talking through my talk. Um these are the things that we already built for you. Everybody in this room, you can just go do a couple of click and you'll have this up and running in your account and i'm going to show you how and there are two things we always say at amazon, we always say that everything in it goes wrong and very wise person told me that don't do a live demo, but guess i'm gonna go do a live demo. So let's see.

Right. Let me jump on my aws account or one of the aws accounts, right? You can just go on google and just search for live call analytics. You'll see this demo uh or we call it lc a or you can, i think the first item that comes on top is this this blog that you have got for live colonized. That's one of the solutions that we built for cc i goes through all the configuration. You can actually watch this, this, this really interesting presentation that we've done in the past. It goes through all the setups, everything that you've got and it gives you a cloud formation there. So you can actually run this cloud formation. And as i said, it's few clicks. We call it lc a re invent if i can smoke 22. And these are all the options that gives you based on your need. For example, you can uh have a demo uh or, or you are using choice you, you would like to use chime voice connector or you got genesis cloud, for example, or amazon connect whatever it is. So obviously, this is a starter kit, this solution is going to give you a starter with just a few clicks. So you'll have your live agents run live agent as is up and running and then you start customizing it for your business needs. I'm not gonna do any of that.

So it goes through kendra. Do you wanna do call recording on three? How you want the transcription configuration? All of your transcript configuration is here. Which, what language do you want to take the pi i out or not? So, on that blog that i've mentioned, it's going to explain all of this configuration to you and all i need to do. Say that, yeah, i'm happy with everything and then create this stack for me. Wait for a 10 minute. You've got a live agencies. I'm not gonna keep you waiting. I've created it before. Let's jump on another account. So that's the information that i built. And if you look at the output of what it gives you, we actually give you an ui so let me just go back to that architecture for a few seconds to explain a bit more what's happening.

So basically what i've created here is going to put a demo uh or, or you are using choice you, you would like to use chime voice connector or you got genesis cloud, for example, or amazon connect whatever it is. So obviously, this is a starter kit, this solution is going to give you a starter with just a few clicks. So you'll have your live agents run live agent as is up and running and then you start customizing it for your business needs. I'm not gonna do any of that.

So it goes through kendra. Do you wanna do call recording on three? How you want the transcription configuration? All of your transcript configuration is here. Which, what language do you want to take the pi i out or not? So, on that blog that i've mentioned, it's going to explain all of this configuration to you and all i need to do. Say that, yeah, i'm happy with everything and then create this stack for me. Wait for a 10 minute. You've got a live agencies. I'm not gonna keep you waiting. I've created it before. Let's jump on another account. So that's the information that i built. And if you look at the output of what it gives you, we actually give you an ui so let me just go back to that architecture for a few seconds to explain a bit more what's happening.

So basically what i've created here is going to put a demo for me on ac two, which i'm gonna call it now. Uh just basically a sip trunk. And uh we are using chime connector stream it. Uh we're going to stream that call over here. We are using lambda. We call the transcribe. We actually created one which i think we went live yesterday or day before we created one with tc a as well. So this is using transcribe and comprehend. We created one with tc a as well. So you can, you can search for that and, and you can use it and uh we put stuff in the three bucket, obviously, again, another kinesis stream that you do your uh comprehend to find out what's happening with the call. And we obviously, it's a very simple u i. Then later on you can replace that, that api that it creates your u i is very simple u i that is on the three bucket. And we are using appsync to update that u i for you. And the agent can, can see that. And as you can see here, things that this cloud formation is created for me is already created. This u i obviously, you can create this, you either way that you like and let's see how it works if it works.

Let me check that he already give you by the way that demo, it already create a phone number for you. I'm sure not calling my manager with a live demo. Thank you for calling mechanic michelle seattle's only all female auto shop. This is maria. How can i help you today? You can see the call there. We already got a call in real time on talking obviously, that's a demo. Nobody is there. Don't worry. So uh and i can call to my phone and say that hello. Um, um, let's go through this call and see what's happening. Uh i'm very happy with your service. Um, i'm grateful and you'll see that life is happening there and you can see it's kind of happy, neutral and then maybe we are talking in the middle. I'll wait for the demo to say something, maybe less try, say what? I'm not happy with your service. I need to talk to your manager. It went red. You'll see that. Um, would you like me to try over the phone with you or would you like to bring it in and have? No, i just don't like anything about your company. I want to cancel my subscription and i wanna go. Ok, great. Um are you when you come? Right? So you'll see, it's already actually picked up an issue, an issue detected there. We've got the issue detected there. You can already alarm it. You can already send a supervisor to go and, and you can see all the online. Let me just disconnect the call as soon as it was disconnect the call. You got the full transcription there as well for you. You can see your call. You can see what happened during the call. Obviously, it is a demo. So it, it was only 15 seconds, but it can go longer. And, um, and you got all the inside of the call that you can see exactly what happened? What time it went? Good. What time it, it didn't. You got all the transcription that you wanted on the call. If you remember, we've talked about the call categories in tc a, you can have those call issues detected and do something about it in real time.

A very quick demo, but it just create two clicks. Everybody in this room can do it. Please go try it and we have just launched one with tc a as well. Not with transcribe. Let's go back to some of the resources that i'm going to share with you.

So we just went through this, how we can help. Obviously, there are many workshops. We've got the cc i team, we've got the business development team. We are more than happy to reach out to your solution architect uh that they are happy with you. You can go to that website. You see all of our solutions on cc i that we can help. There are many sessions although uh some of them are yesterday, but you got today, more sessions and tomorrow we got more sessions as well. Um and these resources like to take a picture. There are or if you just search cc i or live call analysis, you'll find them on the google as well. Of all the blogs that we've written and all the solution that we've created that you can use it as a starter kit and then you can evolve it for your customer need the way that we have done for chase. And thank you very much. We're gonna go to a q a.