Great and welcome to Monday, a free invent and welcome to our session, Boost Agent Productivity with Real Time Transcription and Insights. Thank you so much for being here today. I'm glad you could all join us to discuss how gene or ga I is revolutionizing customer service. My name is Arvin Sundar Raman. I'm a senior manager of product management responsible for a suite of a i services. We call them the Language A I Services here at AWS.

Today, I'm joined by my colleague, Vivek Singh, senior manager of product management, Amazon Transcribe and my good friend and AWS customer, Harry Saban, director of solution a i delivery at walters clo.

So why are we here today in this new era of gene? We have an incredible opportunity to transform customer experience in the contact center space. Now, for many years, contact center agents spent much of their time in admin administrative tasks such as not taking, you know, taking summary notes as well as putting customers on hold to look up information. But thanks to the new advances in ga i agents can kick back and let the machines do all the grunt work which frees up time for the fun stuff, which is any guesses helping customers.

In today's session, we'll talk about how customers that have a contact center can boost agent productivity with aws geni powered real time transcription and insights with that context. Let's dive into the agenda for the session, you know, to start, i'll talk a little bit about the common challenges customers face in managing contact centers.

We will then explain how aws is gen i powered contact center intelligence solutions. We call them the cc solutions here at aws can address those challenges. After that, we will take the stage and dive into some of the key capabilities and features of our gen i powered real time analytics and agent assist solution.

Harry will then walk you through walter clover's journey partnering with aws to implement these gen a i based solutions in their contact center. And we'll also show a cool demo of the solution that they have deployed at walters. And before we wrap up, we'll arm you with a quick how to get started guide to make sure you walk away with some practical and actionable ideas that you can implement to further improve your contact center operations.

We'll have about 5, 10 minutes of q and a at the end. So hold on to your burning questions for the end. You know, back in my undergrad days, this was a long, long time ago. Ok. A couple of my friends, my batch mates want to make some extra money. So they took an evening job at a call center, you know, a week later they were all like, oh, i love this job. I'm talking to so many customers and i'm learning a lot and it's good money.

I was thinking, hey, should i give it a shot? You know, a couple of months go by and we were sitting at this cafe, you know, drinking coffee and this guy goes, well, i can't do it anymore. All i'm doing is typing and typing all day and it's killing me. And the other guy goes and on top of that, i have to memorize a ton of things and they give me a new script every day. They both quit their jobs and then went back to their dads for extra pocket money.

You know, fast forward almost two decades later, things haven't changed much now while making a lot of our product decisions, we spoke to a ton of contact center leaders. Most of them reported a significant increase in call volumes in the last couple of years as demand soldiers. What happens? Your costs grow as well and your agents struggle with high volumes and multiple systems and adds that they need to navigate to get their job done leading to agent burnout and customer tolerance has changed.

Now, today's customers simply want faster and more personalized resolutions and of course not having the right analytics on customer conversations and agent performance doesn't really help, right? Makes it super challenging to identify areas for improvement and make data driven decisions.

So contact center leaders must adapt to this new normal while addressing key customer and agent challenges. So the application of a i specifically taking advantage of the evolution of gene i can not only address these challenges but can also help you lower your cost. Transforming your contact center from a cost center into a profit center.

What are the business benefits of implementing a i in contact centers? You spoke to a lot of customers and I'll give you five based on my conversations, no one, your agents can provide faster and more effective service that will lead to increased customer satisfaction. And with real time agent assistance, your agents can receive suggestions and answers to call us questions right in front of them on the screen in real time which reduces time to resolution.

And number three, you can offload repetitive issues to a i which improves agent satisfaction. And four, you can reduce your operational cost through self service technologies for redundant responses, making the best use of your agent's time and lowering turnover.

And then five super important. You can identify business implement opportunities by capturing more and better data on customer satisfaction, potential product issues, compliance, script adherence or training gaps. And you can mine all of your customer conversations and gain insights on not only product implement opportunities but also cross sell and upsell opportunities.

Let's see how at aws we are all about meeting customers where they are and providing them with options they can choose from to meet their goals. We provide lots of innovations for contact centers all the way from building chat bots, too smart, recording your calls to real time agent assist to post call analytics. We offer flexible options for contact centers.

One amazon connect just for me to understand, show of hands how many of you are currently on amazon connect or in the process of migrating your contact centers to amazon connect or plan to use amazon connect sometime in the future. Ok. Awesome. Excellent.

Now for context, amazon connectors are single unified cloud contact center for voice chat and task management. It includes a capability called contact lens for amazon connect, which provides a i powered real time and post call analytics.

Now show of hands how many of you are on contact centers other than amazon connect your genesis cisco abayat eight by eight. Ok. Excellent. So we have made a lot of the a i capabilities that we'll talk. We'll talk about today available natively as part of amazon connect and for customers like many of you that have other contact centers in today's session.

We'll talk about how you can augment your existing contact centers with gen a i powered cci solutions that combine a i powered voice text and chat capabilities. These solutions are available as open source solutions for you and as partner solutions as well. If you choose to work with your preferred partner for implementation.

Now, what are these solutions? Right. Aw scc i solutions use a combination of aws ml powered services for text to speech, translation, intelligence, search, conversational a i and transcription comprehension capabilities. And many of you who already have a third party contact center with the likes of genesis cisco avaya talk desk eight by eight can benefit from the cci solutions to provide automated self service capabilities, analyze your calls in real time and to assist your agents and also learn from your contact center conversations using post call analytics.

So these issues empower you to improve your customer experience, boost agent productivity, improve operational efficiencies and gain insights by adding a i into your preferred contact center without any ml expertise.

So this is one of my favorite slides. As you can see customers across many industries use aws contact center solutions to enhance their customer and agent experience. So whether an organization has a small contact center or is one that handles millions of calls every year. Aws has the tools, technologies, skills, expertise and experience to help meet their goals.

So we aim to be a strategic partner in providing innovative scalable and cost effective contact center solutions for all types of customer service organizations.

Now let's dive into the key use cases for cci right? You can apply cc i to three main use cases.

Number one self service, virtual agents that includes conversational ivr and chat bots that deflect calls by enabling users to quickly find answers and complete transactions on their own without the assistance of a live agent. And these solutions can also identify the intent of the caller and direct them to the proper agent thereby reducing the number of call transfers.

Number two, post call analytics, now post call analytics cc i solutions enables you to process 100% of your call volumes. With that you can gain insights such as emerging trends, customer sentiment training gaps for your agents and business implement opportunities.

Number three, the fun stuff, the main focus of today's session, real time, call analytics and agent assist. So this really is the holy grail of understanding your customer in real time that with cci we are focused on relieving the burden on the agent who is on the front line, communicating with customers.

So with real time analytics, you can process these live conversations and understand the context of the call and with agent assist, reduce agent fatigue from information overload and provide them with jim powered real time suggestions and answers generated from your own knowledge base or other data sources. Customer sentiment can be monitored and supervisors can be alerted to jump in and help their agents with calls that may not be going as expected.

Let me now welcome Vivek onto the stage to go deeper on some of the key features and capabilities of our real time call analytics and agent as a solution.

Ot thank you. Hello, everyone. I'm awk singh. I lead the product management team for amazon transcribe today. I'm excited to share with you who you got here. Hey, jonathan is, can you turn the mic on? Hello, mike turner. Hello, can you, is it better? Ok, awesome. Thank you. Thanks for pointing that out.

All right. So as i was saying, i lead the product management team for amazon transcribe today. I'm excited to share with you an overview of the aws real time call analytics, an agent, a solution. We developed this solution to help businesses create better customer agent and supervisor experiences by addressing the key challenges which the contact center customers face today challenges such as long call resolution times, inability of agents to find answers to customers questions quickly and the significant time spent by agents on after call work like creating call summaries, distribution leverages and optimizes a variety of aws services that have been specifically fine tuned for contact center conversations.

For example, it uses amazon transcribe and automatic speech recognition server to produce highly accurate real time transcriptions with sub second latency. Amazon transcribe has been optimized for telephony, contact center conversations and is designed to handle a wide range of speech and acoustic characteristics including multiple accents, noisy backgrounds as well as variation in volume pitch and speaking rate.

We have also developed custom natural language processing models for conversation analytics such as detecting sentiment, identifying call drivers and also identifying acoustic characteristics like interruptions and loud speech. The solution leverages fine tuned amazon bedrock large language models for powering generative a i capabilities like creating call summaries and also automatically generating responses for agents to address customer questions.

All these capabilities together help agents solve customer problems more effectively and respond faster than before lowering, both both the call handle time as well as the after call time.

Now let's dive into the specific capabilities of the solution. We start by looking into real time insights that help agents save time and reduce their fatigue from information overload. Here is a sample agent, customer conversation. You can find turn by turn transcripts that are broken down into customer and agent turn this makes it easy to read and understand the transcript. And it also removes the need for agents to take notes during the call.

You can also leverage customization capabilities that transcribe provides like custom language model and custom vocabulary to boost the accuracy of domain specific terms like brand names and product names that are unique to you.

Issue detection allows you to know the reason why a customer is calling without having to listen to the entire call, recording or reading the call transcript using the issue detection capability. You can find the shortest set of words in a conversation to that represents the reason why a customer is calling agent assist, helps identify the intent of a customer during conversation terms and it automatically surfaces generative a i powered responses to agents to address customer questions.

In addition to boosting agent efficiency, cci real time insights can also provide supervisors more nuanced visibility into the customer conversations. We provide the transcript along with the sentiment for both the agent and the customer. You can find turn by turn sentiment. We also provide overall sentiment and sentiment by quarters so that supervisors could easily look for sentiment shifts within the calls to identify what's causing those changes.

We also provide acoustic based conversation characteristics such as non talk time, loudness, interruption and talk speed. Using these characteristics

Supervisors can identify various scenarios such as abnormal duration of silence in a call, loud speech, frequent interruptions, if an agent is talking over the customer or comprehension issues. If an agent is talking too fast or too slow, you can also leverage CCI capabilities to notify supervisors of urgent issues in real time.

Both supervisors and managers want to be notified as soon as an issue occurs so that they can address the customer issue before it gets escalated further. This is critical to improve customer satisfaction and also increase the first call resolution rates with real time alerts. Supervisors are proactively notified if a call needs their immediate attention.

When certain rules that you have defined are met on a customer call, automated call categorization capability enables you to create these rules to identify specific scenarios such as manager escalations, account cancellations or competitive mentions. For example, you can create a category for escalation requests to tag all live calls that contain keywords, like "talk to your manager" or "speak with your supervisor."

You can also create these categories based on acoustic characteristics. For example, you can tag all live calls that have a duration of silence of greater than 15 seconds. You can essentially create categories by looking for presence or absence of keywords, phrases or acoustic characteristics at any point during the call.

When the categories that you have defined are detected, supervisors receive an alert or a notification in their real time dashboard. Based on these alerts, supervisors can provide agents guidance on resolving the customer issue via a whisper coaching session or supervisors can determine that a call transfer is needed with real time call transfers, you can pass the in progress call transcript from one agent to another agent or a supervisor. This reduces customer frustration because customers don't have to repeat themselves during call transfers and their issues are resolved much faster.

Supervisors can refer to the real time transcript that contains both the customer and agent sentiment. And they can also see the call summary generated up to that point using the transcript, the customer sentiment trend and the call summary. Supervisors can quickly get up to speed and assess the appropriate action needed.

CCI real time solution also makes it easy for you to enhance the security and compliance of your contact center. You can redact sensitive personal identifiable information, PII entities such as names, addresses or credit card details. And you have the option to redact all the PII entities that we support all at once. Or you can pick and choose specific PII entities that you may want to redact, such as just the caller's name.

This redaction removes the sensitive information from both the call transcript as well as the call recording. And you can also control permissions within your organizations for access to the redacted and unredacted version of the recording as well as the call transcript.

So far, we have covered how you can leverage real time transcription insights and redaction to enhance the agent and supervisor experience in a secure and compliant manner. Next, we will look into how you can leverage CCI generative capabilities to automate after call work for agents.

Agents report that after call work like noting down call summaries or follow up action items can take up to a third of the total call. So they sometimes skip it or fill in complete information. Failure to accurately capture and act upon important action items discussed during the call can erode customer trust. Additionally, incomplete call summaries can lead to supervisors spending significant time listening to the call recording or reading the call transcript to get the gist of a customer conversation when they are investigating a customer issue or evaluating an agent's performance. This also makes it really hard to scale quality management within the contact center.

Hence, we launched a new generative AI powered call summarization capability powered by Amazon CodeWhisperer. This call summary captures the customer's problem, outcome of the steps taken during the call, and the next steps. Supervisors and agents can access the summary of a conversation just few seconds after the call using call summarization. Agents can save anywhere between 45 seconds to two minutes per call depending on the overall call length from reduced note taking needs at scale. These time savings can help businesses save millions of dollars annually.

To further reduce the after call work time for agents, you can leverage generative AI capabilities to also simplify the follow up workflows. Follow up captures the promises and commitments made to the customer that need to be kept. Common examples include a follow up email that needs to be sent to a customer regarding specific instructions, such as sharing more details about a particular policy or a follow up phone call that an agent has promised to make to a customer.

CCI's solution enables you to explicitly capture the follow up action items for a customer call. And as you can see, it can also auto create generative AI powered email responses to help you close the loop with the customer much faster.

All in all these gen capabilities can not only significantly boost agent productivity but also impact customer experience because agents can now move to the next caller in the queue much faster, reducing call wait times.

Now let's look into a few examples of how companies are leveraging AWS CCI capabilities to create business value and enhance contact center experience.

First, we have Magellan Health, one of the largest managed behavioral healthcare companies in the US. The company manages employee assistance, mental health plan and work life programs through its nationwide third party provider network. Magellan Health wanted to boost the productivity and efficiency of its call center agents who had to look for information across 13 disparate data sources to answer customer questions.

Hence, Magellan Health used AWS Real-Time Agent Assist solution to surface contextually aware knowledge articles pulled from these disparate data sources for quick access by agents using CCI. Magellan Health was able to reduce the average training time for agents by 3 to 5 days and they were also able to reduce call handle times by 9 to 15 seconds.

Second, we have Parcel Pending by Quadient, one of the leading providers of smart locker solutions for residential, commercial and retail properties in North America. The company wanted its agents to not use scripts but take ownership over processes to give better service to customers.

Parcel Pending uses Talkdesk CX Cloud contact center and its Interactions Analytics solution which leverages AWS CCI real time capabilities to transcribe, detect and identify customer issues in real time with the help of CCI. Parcel Pending was able to increase first call resolution rate by 29% and reduce average call handle times by 45 to 90 seconds.

Last we have a large multinational financial bank that started off by replacing the incumbent transcription service provider with Amazon Transcribe because Transcribe had 12% higher accuracy compared to the incumbent with more accurate real time transcriptions. Along with the automated PII redaction capability that Transcribe provides, the bank was able to more accurately categorize all their calls for future processing while adhering to their data privacy and security requirements.

So that completes our overview of the capabilities of the Agent Assist and Real-Time Call Analytics solution. I do want to emphasize though that the performance of the speech analytics capabilities that we just talked about is highly reliant on the performance of the underlying speech to text or automatic speech recognition service models.

We are also seeing that in order to effectively leverage gen capabilities, customers are increasingly looking to transcribe their audio data with the highest possible quality across a variety of languages. Hence, at AWS, we're continuing to double down on innovation on the ASR front.

To that end, I'm excited to announce the launch of a new next generation speech foundation model powered ASR system that expands Amazon Transcribe language coverage to over 100 languages. This multibillion parameter speech foundation model was trained using best in class self supervision algorithms to learn the inherent universal pattern of human speech across various languages and accents.

This model has been trained on millions of hours of unlabeled audio data across 100 plus languages. Powered by foundation model, Amazon Transcribe can now deliver up to 50% in relative accuracy improvements across most locals on telephony speech which is more pertinent to contact center conversations and is also generally a challenging and data scarce domain.

Amazon Transcribe now delivers up to 70% in relative accuracy improvements across most locals. Additionally, this model also enhances the readability of its output with more accurate punctuation and capitalization. In addition to these substantial performance improvements, this new ASR system also delivers several differentiating features across all 100 plus languages related to ease of use, customization, security and privacy. These include features such as custom vocabulary, automatic punctuation, automatic language identification, and speaker diarization and many more.

Now before I move forward, I would love to invite Harry Sabani from Wolters Kluwer on stage. Harry will share how Wolters Kluwer is leveraging AWS CCI capabilities to boost agent productivity and enhance the contact center experience.

The integration with Genesys, the ability for their system to integrate with our own existing AI, the overall extensibility of the solution. And finally, the scalability which is very high also on our list of priorities.

Another set of reasons why Amazon won was because of the infrastructure jumpstart. So this was really incredible. Um so the team of uh Arvind actually helped us with this and they, so someone named Bob Strahan and, and his team was an architect. Uh the CCI architect gave my team uh a set of CloudFormation templates. We got a system up and running within 2 to 3 months, we were able to demo it to Ami and her team. And that's when we knew that this was viable, it was real and that we could actually touch it.

Um we also got help with, with building out the custom language and vocabulary. So we got someone from Arvin's team to join us and handhold two members of my team. One's a computational linguist and and one's a machine learning engineer just to kind of help them handle a bit to help them get going.

There was also a Call Analytics dashboard that the CCI team had built. So we took that used it as a template and that's how we were able to do our demos to a team. And finally, um you know, with, with the help from Brian Spiro, who's our account manager from AWS for Walters Cleer, we got AWS credits to, to help us out, right, to help uh progress this um quickly.

Final set of things, a very strong collaboration. This, this one is just mind blowing, right? Um we met with Vivek's team and Arm Fin's team members every week, right? This is every single week and it was a cool experimentation process. What that meant was we would um come with our requirements and, you know, several times we would see a turnaround of like a week, right? The, the CCI team would come back and say, hey, we did it, you know, we got this done. Can you try it out? So this is very interactive weekly co experimentation that went on for a few months.

Again, this is a complex overall system, right? So it took a little bit of time. Um we got, so we got um also expertise as there was people in, in uh in the team who had actually worked with Genesis very deeply and audio codes. They came to these sessions. They, they helped us significantly to integrate um the IVR with with AC C I solution.

Um we had interactive troubleshooting when there were issues and most of the the the issues are actually, you know, configuration related. And then finally, um you know, we also had collaborative uh load testing. So we ran our own load test at the same time. Um you know, the, the CCI team also very graciously, you know, said we want to help you with this and, and did that. So overall, the high level of initial ongoing support was really powerful.

Um the solution is, is the high level solution is here on the screen. Um you know, uh behind the scenes, there's a lot going on. But in general, the journey starts with a call being made by, by a customer. The call goes to our Genesis IVR. It reaches um Amazon Chime Voice Connector and the audio stream makes its way to um Amazon Kines uh video streams and the that stream goes to uh AWS Transcribe and there, there's a um speech to track, uh there's a translation of speech to text through an ASR system. There is sentiment analysis taking place there and the P redaction is also taking place around there.

When the text is created through the transcription system, it makes its way to our LLM which does the generation of summary and also topical and intent tagging. And then ultimately, at some point, the traffic makes its way to our Real Time Agent Assist which is finding relevant answers uh in real time to queries made by the customer. And it's also a way of context and, and things like that.

So the business impact has been pretty high. So phase one is in production, we have real time transcription P I redaction and sentiment analysis in place. This has allowed us to allow actually a MIS team to gain deep insight into the content of our calls, insight into the customer's sentiment, provide enhanced B I capabilities also allow us to collect samples for LL and fine tuning.

Phase two. Also in production has generative AI driven summarization and intent and topic tagging. The business impact has been really high time saves on after call work because our agents don't have to summarize anymore. They review the summary that's automatically generated and then the intent tagging topic ta tagging is also enhanced. Uh B I capabilities.

Phase three is gonna go out in Q4, sorry, Q2 2024 that includes real time agent assist via our homegrown advanced virtual agent that's gonna save agents a significant time on finding answers to customer queries. And we can also uh at that point avail of the things that ha that we didn't phase one and two to acquire um samples for, for L and M fine tuning, not going to go through a demo of a system. This is the system that's out there in production.

Um on the left hand, side of the screen, you're gonna see a um a sales force screen and our solution is embedded in a widget in that screen. Yeah, and the right hand will show the real time, you know, transcription happening.

Hello, I'm Eric. How can I help you today? Hello, Eric. Uh I wanna pay my invoice. Ok. Can you please help me with your account number or your name? I am Cynthia Weber. Give me a moment. Cynthia. Yes, I can see your account. Ok. Ok, I can see an outstanding amount of $69. Do you want to proceed with the payment? Yes. Ok. To process the payment, Cynthia, I'll need some information first. Can you please confirm your email address? Cynthia at cw.com. And what is the billing address on your credit card? 2950 Birch Street. Please share the name on the credit card. Cynthia Weber. And can I have the credit card number? 4589365274125369. And can I have the expiration date? And the CVV code? It's 26 2025 and the CVV is 3845. Cynthia. You are authorizing Walters Kluwer to charge your credit card for the total amount of $69. Yes, I am. Ok. The payment is submitted and you'll receive an email confirmation with case number 16523458. Ok. And is there anything else I can help you with Cynthia. Uh no, but thank you for asking. You. Have a wonderful day. Bye. Ok, thank you. Bye.

After the call ends, we use generative AI to summarize the call and also create topical tags for us saving a significant amount of time for agents. Another important set of analytics CCI provides us with is information on the ongoing and overall sentiment of both the customer and the agent during the call. This will help us provide valuable insight through our BI system and will also be used to determine if manager intervention is required in real time.

In our third production release just after tax season. In 2024 we will integrate our existing advanced virtual aging a chat solution highly tuned for finding precise answers with our CCI system to provide agents with real time answers to customer queries. The agent assistant segment on the video shows a sample response from the virtual agent. This will significantly empower our agents who will get answers in real time. Instead of having to spend time on a search engine to find answers. This chat solution will be integrated with generative AI to automatically summarize the answer to a user's query from a set of relevant matching documents.

Um before I hand it over to Arvin, just a big shout out to the people who made this happen. Just a few folks, I want to give a shout out to first this Arvind. Uh thank you so much. His team was a huge part of this. Vivek also very big part of uh our solution here. Um shout out to Brian Spiro, our account manager and Bob Strahan, the CCI architect and also of course, uh Ami uh Darlene Alexander for, for help, you know, being such a strong partner for us and William Flannery, our head of advanced technology at Walters Clu. On that note, I'll hand it over to you, Arvin.

Thank you, Harry. Can you all still hear me? Ok, awesome. That was a great presentation and thank you for all the kind words, Harry, appreciate it. See our customers get tangible business value from our TI services. Always exciting and fulfilling for me and I hope you all are as excited as I am about our Foundation model powered new ASI model. We're super happy to get that out for your consumption.

Well, if you liked what you saw today and want to get started, how do you get started? Right with the broadest and deepest EI and ML services, we can help guide you on your ML journey. Now with the, you know, we can work with you to identify the right business challenges that you're facing today and that you want to solve with AI and help you against those with our CCI solutions.

We have trainings and acceleration programs to help you get started. We offer both in person and virtual training for your business and technical stakeholders through our AWS ML Embark program, as well as the full free online training and certification program for all of your developers. And lastly, we have more than 40 CCI partners that really understand the space very well and can work backwards from your challenges and business requirements and help you add the right AI capabilities into your preferred contact center.

Here are some additional resources including blog posts and solution overviews for all of the use cases we discussed today to help you get started.

Thank you. Thank you so much for taking the time to be here today.