AWS re:Invent 2025 - The agent-enabled workplace: Transforming businesses with AI (INV203)

🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.

Overview

📖 AWS re:Invent 2025 - The agent-enabled workplace: Transforming businesses with AI (INV203)

In this video, Pasquale DeMayo introduces Amazon Quick Suite and Amazon Connect's AI capabilities for transforming workplaces and customer service. He addresses the 42% failure rate of AI projects and demonstrates how human-centered AI can achieve 70% improvement in work completion. Amazon Quick Suite is presented as a unified workspace that eliminates AI sprawl by integrating enterprise data, metrics, and agents for insights, research, and automation. Customer examples include AstraZeneca using Quick for clinical research automation, BMW leveraging it for Neuerklasse development workflows, and 3M improving sales effectiveness across 100,000 meetings. Amazon Connect showcases 29 new agentic AI capabilities, with Priceline demonstrating 50-second savings per call through automated summarization. The session reveals Amazon internally has hundreds of thousands of Quick users with 50,000+ research reports created and Connect powering 100,000 customer service agents processing millions of daily conversations.

; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.

Main Part

The Challenge and Promise of AI Integration in Business

Please welcome to the stage Vice President and General Manager of Amazon Connect at AWS, Pasquale DeMayo. It's wonderful to have you all here at re:Invent on the first day. We have so much to talk about. I'm going to dive right in and talk about how the agent-enabled workplace is changing and how we're transforming that business with AI today.

Has anyone here done work to try and enable agentic or generative AI in their business? If so, you're probably already aware that this is directionally correct. What we're hearing is that 42% of these projects are failing. That's a pretty high failure rate. I've heard as high as 95%, which is probably a little exaggerated, but the reality is that people are really challenged by what's happening here, and it's been very difficult for people to get the success they expect.

However, you've all also seen what the potential is. Some of the data shows that when these human plus AI integrations are successful, when people really make them work together, they're seeing a 70% boost in work completion. When you see a result like that, it tells you very quickly that you have a real opportunity here. Anything where you can achieve a 70% improvement is almost always certainly interesting. I'm sure some of you are in businesses where a 1 or 2% improvement is a massive win.

What we've seen here at Amazon, and what we're here to talk about today, is that we're moving from a world where AI is just a tool to a world where we need to make AI human-centered. We're bringing AI to meet you where you're at and meet you where your workforce is at. We need to get out of that space where people are having to transition back and forth, where you see some point maximum potentially and sometimes no advantage. Instead, we're moving into a world where every aspect of what you do is being enhanced in a way that is actually helping that person and making them better at the things that make you great as a human, while taking away the things that are mundane, boring, and don't bring real value.

From Tool-Based AI to Human-Centered AI: A Marketing Example

Rather than building AI as just a sophisticated tool, build it as something that's amplifying the human capabilities within the business and within the workflows that exist. Let me start with an example that many of you may relate to. Imagine you're a marketer and you're asking yourself, how come our most recent campaigns didn't hit the targets we had for them? You're probably going to go and try to collect a bunch of data. You might get some CRM data, you'll get some campaign data, you'll try to patch that together, maybe you'll get some customer feedback if you're lucky, and pull some reports, maybe financial reports. All of that is going to give you some insight.

If you are even more lucky, you might have some business intelligence support. Maybe it's a half a person, maybe you even have a BI team. But in any case, when you're asking people to come bring you stuff and then it doesn't quite give you what you wanted, so you ask them again. By the fifth time you've asked, they're annoyed with you. No one's having a good time. And at the end, do you really end up with the data and the insights you need? Probably not. I've certainly struggled with that myself.

In a different world, one where AI is actually supporting you, you can actually ask the AI directly: "Can I get some insights on why our Q4 campaign is underperforming?" Those answers are then tailored to your business. The AI can understand the things going across all those data sets, and the AI never gets bored. It never gets tired of being asked to reformulate or think about it a different way. It can even start to think about pulling in data from other places, maybe external sources like news or events that happen in the real world.

This is so much of an improvement over where you are today, where you're asking someone to help you while trying to communicate in ways that they're understanding. You get frustrated, you wait for it, you get frustrated moving back and forth. Instead, enhance this into a way where you're not dependent on a human's ability to slice massive sets of data. Use AI for what it's great for: helping you understand these things and then bringing it to you in a way that you as a human can consume it.

Transforming Customer Service with Amazon Connect

Another area that I'm incredibly personally passionate about is customer service. I'm the VP of Amazon Connect. Are there any Amazon Connect customers here today? Well, thank you so much. Amazon Connect is a customer service solution that folks use to run contact centers and other aspects of their business to engage with their customers more effectively. In a customer service example, think about Maria in this instance. She's reaching out about a problem with billing, and it's a confusing problem. She talks first to a bot, but the bot doesn't really understand it. Maybe then she calls and talks to some form of interactive voice system, but it doesn't really get it. Then she talks to an agent and repeats herself all over again, having the same conversation three times, not really getting satisfaction, and wasting a ton of both folks' time.

It's not saving the company money, it's not helping Maria solve her problem, and most of all, it's not building a relationship between Maria and the company by any means. Starting over all those times is no fun. But you can imagine a world where instead, when she calls in, she's able to actually talk to a generative AI operative, an agentic AI capable system.

That is understanding her problem but also pulling data about everything in her history that understands that she did make a change to her billing address and it didn't work. She did try, and it did have an upgrade recently to a new system. It says, "Hey, you know, it looks like these are the challenges you're having. First, let me help you fix your billing address. That's not something where a person helps you do any better. I can fix that. I can double check and make sure the address is right. I can then write that to the system for you."

While I'm doing that, I'm going to hand you off to a billing expert, and that person is going to have all the context of what happened in this earlier part of the conversation and all the fact that we've verified you as Maria. So I know I can help you, and I'm going to then take what was a complex and convoluted conversation with Maria and turn it into actionable steps and then actually help that agent take actions while they're able to focus on Maria and think about what makes that experience human or even superhuman given the fact that they are getting support the entire time through this AI concierge that is helping both Maria and the agent at the same time, making them better together in ways that they never were before.

We see folks getting the advantage of this. I quoted some numbers before, but I think these ones are really compelling. First off, I'll say 25% improvement in performance to complete the task is great. That's a direct savings in effort and time, which is always wonderful. But even more important, I think, is that the human beings are getting a 40% higher quality of work outcome. This is when it's done well, where you're using this integrated AI approach. It's not having the AI take over and write work slop that you're hearing so much about. Instead, the AI is working in a cohesive way with the human beings to make them superhuman.

The Problem of Application and AI Sprawl in Modern Workplaces

So at this point, I'd like to dive in and talk about how we're going to think about two things today: the AI-enabled workforce and the AI-powered customer experience. And so to help me with that, I'd actually like to welcome Jose Kunnackal John from AWS in the Quick Suite. Thank you, Pascale. So let's dive right in. Let's think about your typical workday. What does that look like? How much time do you spend hunting for information that is hidden across multiple systems? How many times has that approval been stuck in email chains and you have had to dive in to sort that out? How about the number of times where you have to copy data from one system to another?

Reality is that business processes today have been built taking into account the multiple systems that people have to navigate. These involve multiple decision trees, multiple data entry points, all of which need to be managed by users. Teams are working through multiple applications every day, and they live a fragmented experience. The good news is that AI has shown promise here. Over the last few years, you've seen multiple AI solutions, multiple point solutions that jump in to save the day. Whether it is a better email writing assistant in your email system, or whether it's a better financial modeling software within your finance team, or maybe just something that lets you find the latest information for the next customer meeting, what you'll find is that these solutions are making impact in terms of saving time.

However, they've created a new kind of problem: that of application and AI sprawl. So what you're finding now is that organizations have different AI solutions for different needs. They're managing dozens of these disconnected AI point solutions. Finance teams, marketing teams, sales teams, operations teams, everybody has a solution. But each of these solutions comes with a separate experience. It comes with separate licensing, separate user and change management that is required to onboard these, separate procurement, and of course the need to manage data and governance across all of these systems. So what turned out to be an effort to simplify makes things more complex.

Your employees are now context-switching across both applications and these assistants and AI agents, none of which can see the full picture or work seamlessly with each other. This is where we believe that a fundamentally different approach is required. One where AI agents are intelligent orchestrators, intelligent orchestrators that can stitch together the context, understand that, and then take action as needed. AI agents are not just point solutions, but form a unified system that can then break down these silos that you have and make sure that they can bring together workflows. AI anticipates your needs in this case.

Introducing Amazon Quick Suite: A Unified Workspace for Knowledge Workers

It knows what you would have needed to know and then makes sure that it can take the action on your behalf. Recognizing these challenges, we started thinking about who we build this for and what we are doing here. The thing we saw is that we are building this for the end user—the end user who does not code, who does not build automations, who does not build dashboards.

All a knowledge worker wants to do is to do their job easier, faster, and smarter. They want to make sure that they are not doing tedious, tiring work day in and day out, but get to more strategic items that they can do while some of this grant work is just done for them. When you look at what people want to do, it's typically a few different things in a sequence.

Number one, people want to find information. Remember we talked about emails and how to find information across those emails. Sometimes you're hunting for that particular email which gets you the right insight. Once you find that information, you are taking that information and analyzing it. Sometimes you need to dive in a little, sometimes you need to dive in a lot. Sometimes you even have to go talk to a coworker, whether it's a finance person or a supply chain expert—an expert in some domain who's going to tell you more, who's going to tell you whether the way you're thinking is going to be the way to go or not.

After all of that, you take some action because you have to convert whatever you found into an email, a presentation, or maybe you're kicking off a process. All of that together makes the end-to-end cycle work. Different users have different proportions of this, but they all do some form of the sequence. Taking all of this into account, our fundamental belief is that the way we need to work here is to have AI agents work as orchestrators that can do the work for you.

This is where this vision of eliminating agent sprawl, of bringing together context that is required, and empowering end users and non-technical users with the tools they need to help them go faster and smarter is why we built Amazon Quick Suite. What you're seeing here is a unified workspace that brings together context across all of these information sources that you work with, as well as agents that can then use that context in order to do things for you or do things with you.

When I say context, I mean information that exists in your enterprise systems. Think systems like SharePoint and Confluence, where you have enterprise information that is stored. You also have metrics, so think about information that is in your databases, data warehouses, and data lakes—numbers and dashboards that convey the information on a daily, monthly, and weekly basis. You also have team information because not all information is centralized or with IT. Every team has a way of working. It might be in documents, it might be in team SOPs, and they need to be brought together because that's important context.

And then, of course, there's internet information, so everyone looks at what's going on before they make that final determination. These different pieces of context, along with agents—agents in the Quick Suite include agents for insights, agents for research, agents for automation—so that you can then take the insights and the context and then work with it. What's more important is that all of this comes in one place, but of course, it can't be that you're always working in the Quick web application.

You can take Quick with you, whether it is in your Microsoft Office 365 applications like Word or Outlook, whether it's in your favorite messaging app, or even in your browser or on the go on mobile. Quick brings all of this context with you so that you have that with you as you go and need to make decisions. Quick represents a fundamental shift in how information workers work with their context and with each other. Think of this as having an incredibly intelligent teammate that is working with you and that has all the context and that is also able to reach out to get external information as required.

Core Capabilities and Interconnectivity of Amazon Quick Suite

With Quick Suite, you can ask questions and get clear, concise, and comprehensive answers across all of that context that we just talked about. You also have teammates—agentic teammates—so for research, for insights, for automation. Once you get that piece of information you wanted, you probably want to deep dive a little bit. You probably want to go and research more into that, and Quick lets you do that. You can visualize and analyze data all from Quick.

Once you have that insight and research, you can take action, whether to send it out as an email or kick off an automation. Remember, knowledge workers don't want to do tedious work. That's where Amazon Quick Suite lets you write those automations really easily and then get going. What's more, all of this is in a way that can be easily rolled out to users. Amazon Quick Suite comes with permissions and governance and other controls that are required when you think about rolling this out in an enterprise. So it doesn't matter if you want to roll this out to a team of 10 users or you want to roll this out to hundreds of thousands of users, you have the controls and the capabilities that you need in order to roll this out.

Now one important piece here is interconnectivity, because the context you need to make decisions is not all in one place. Across a suite of native connectors, as well as technologies like MCP and OpenAPI, Amazon Quick Suite allows you to connect with your data and information. It doesn't matter if it's in a third-party system, such as an application like Box where you have your information, or if it's in a custom system that you have created. But apart from the context, you also need to connect to the actions and invoke agents as required. Not everything is going to be built here, so you might have systems where you have something as simple as sending an email via Outlook. You might have custom agents that you've built in your company and you want to invoke that via MCP. All of that becomes possible because of the breadth of interconnectivity.

Let me give you a simple example. Let's say you have your information in your inbox. You can go and ask Amazon Quick Suite to retrieve that information, reason through it against the information from your Outlook calendar, compile a set of notes, maybe create a presentation in Canva, and you can do that end-to-end flow within Amazon Quick Suite. Then you can ask Amazon Quick Suite to send out the links of that presentation to your co-worker. See how we've brought all of those systems together? Even though information resides in all those systems, you've brought it together in one place.

Amazon Quick Suite in Action: From Search to Automation

Now let's take a look at a few examples to put this all into perspective. First, let's imagine that you're asking a question of Amazon Quick Suite in order to prepare for a meeting where you're going to look at your Q3 performance, and then you want to understand the areas of opportunity and what might you want to look at. When you ask this question, the question is going against all of your enterprise context. Remember, your enterprise data in SharePoint, your files that you've probably uploaded yourself or your team has uploaded, metrics and dashboards, so bringing together all of that in one place. It's not just a search because it's agentic search. What Amazon Quick Suite is doing is putting all of that together, reasoning through it, and then coming back to you. What it does is actual numerical analysis and sometimes when it's needed, it comes back to you with a summary that you can then use. All of that is also cited across the sources, so you know where all of this information is coming from.

Now, once you have something like this, it's great, but you might want to deep dive. The next natural step is, hey, I've got my summary. You could send this out as an email to the team, that's very well possible. You just have to mention that. But let's say you wanted to deep dive and you want to look at where your tickets are and what the feedback is from different teams that work with you, such as open support tickets in this case. When you do that, you can go against the systems that you operate in.

It doesn't matter if it's Asana, Jira, or in this case, you're seeing an example from ServiceNow. It is using credentials that you've provided, so your permissions are being applied, your permissions are being respected, and that information comes back. Amazon Quick Suite is able to reason through that and then again, get you the information in a way that you can action. Again, if you want to shoot off an email to the team, you can do that right here. You don't have to switch context. All of that is available for you right there.

Now, remember we said there are tedious tasks that sometimes you have to repeat. Let's say you did this well for that customer for that weekly or monthly meeting. Great. Now, next time you have to do this, you might have to run through the same process. This is literally the story that is being repeated across organizations. Amazon Quick Suite offers a product called Flows that allows you to do this just by simple natural language. You can describe what you want to see automated and then ask Amazon Quick Suite to go automate that. You don't need to be an automation developer in order to do this. All you had to do was simply specify what you wanted to see. Once you do that, you will find that Amazon Quick Suite is generating that flow. A flow is a sequence of steps.

That can be anything from searching something like ServiceNow that you saw, to searching the internet and reasoning across that, so you can have reasoning steps in there, and then you can ask it to take an action. These flows can be scheduled so that next month when you have to do this, you don't have to go do this all over again. It just does the work for you.

We spoke about speaking with experts. One of the things that Quick offers is also research. With research, you have the ability to invoke what is a PhD level researcher. So you can go and say, compare our results to industry benchmarks. When it does this, Quick is going to go against the web, it can go against specific files you've uploaded, Quick Suite assets, or now you can also have trusted sources like S&P Global, Factset, IDC all included in there.

When it comes back with the results, it also shows you why it made a statement. So you can go to a specific statement and understand why that claim was made, why that statement is in that document in the first place. Quick also includes charts and graphs so that you can visualize some of this all without having to do that second-level deep dive yourself. All of this is available and you can even iterate on this. So if you have a specific comment in there, you can go mark it out, and then Quick will go and iterate on that for you.

All of this is changing how we work. You'll see that now you no longer have to be trapped in that sea of emails in order to get to an outcome. You don't have to go hunting for that information. All you need to do is ask the question and get comprehensive insights across all of your context, because it's all interconnected, it's all available in this one place, and it's available on the go as you go. Quick is transforming the way that we work, but let's hear it from a few customers who are now starting to live this transformation in their own lives, in their own organizations.

AstraZeneca's Journey: Accelerating Medical Research with Quick Suite

Well, first off, thank you all for being here. I appreciate it. I'm sure the audience will love to hear from you. So let's start off with some quick introductions. I'm Celine Laurent-Winter, Vice President Connected Vehicle Platforms at the BMW Group. I'm Vaishali Goyal. I'm a Senior Director of R&D IT at AstraZeneca. Hi everyone, I'm Nithin Ramachandran. I lead the global data and AI practice as the Vice President of Data and AI at 3M.

Well, first off, thank you all for being here. I think this is great to hear firsthand from you about your experiences with AI and transformation. Vaishali, why don't we start with you? Staying current with medical research is incredibly important in the pharmaceutical sector. So why don't you tell us a little bit about how Quick Suite is delivering real value in the trials and MVP that you're running so far?

Yeah, perfect. With Quick, we actually seen in the clinical life cycle a heavy amount of our practitioners spending copious amounts of time finding data, finding research. Often they're looking at market research and insights, but they can't find it so easily. They would sign up for alerts and you get email notifications, but you know how that gets, it's very taxing. What happened after we engaged with AWS and Quick, we were actively able to automate all of this in a much faster way. So now we were able to take all of this information that they would be sourcing from various external information and actually bring it to life in a concise and consistent format that was actually tailored to give our leadership and even the TAs and the different practitioners that we have there the quick insights.

One thing that we also noticed with our hematology therapeutic area, which we started with, they've been actively able to now add insights to this as well to augment the agents, and that's benefited us greatly. For the future, we're looking to roll this out to all the therapeutic areas within AstraZeneca in 2026.

Can you maybe speak a little bit about the specific capabilities and how you're using them? Sure, so we actually are utilizing three of the Quick features, specifically

we are using Flow as well as Research and Automate. With Automate, we are actively able to scrape across all those different external data sources and bring that to life. We are also able to do it in a real-time fashion, so it synthesizes that data. With Research, we're actually able to map it against our internal needs, and we're able to do a compare and contrast. Then with Flow, we're actually able to distribute it. As you mentioned before about sending out a concise email, we're able to actually send that out to leadership, but also give this to our practitioners.

One thing we're very focused on is we're not taking away any capabilities, but we're actively enhancing their ability to find information much faster. I think I love how those three capabilities come together for your users. That's really good to hear. Well, Celine, why don't we switch to you, and first of all, thank you for being here. Why don't you tell us a little bit about the Neuerklasse initiative and why that's so important for BMW?

BMW's Neuerklasse Initiative: Scaling Software-Defined Vehicles

Absolutely. Well, the Neuerklasse stands for electric, digital, and circular. Regarding digital, we have a lot of great new features. The panoramic iDrive is probably one of the highlights with the enhanced assistant functions and the BMW intelligent personal assistant that anticipates our customers' needs. We also improved our over-the-air software updates, making them more convenient, and our customers can even personalize their BMW via the BMW app even before taking delivery.

The Neuerklasse is a truly software-defined vehicle with a new electronic and software architecture, a state-of-the-art foundation for mobile services, real-time interaction, deeper access to functions, with a broader range of remote services and personalization as well as new connectivity options inside the car. With a lot more Neuerklasse cars connected in the world with new features sending and receiving more data, it has become absolutely crucial to scale in an efficient way. It's about choosing the right architecture and also about automating our workflows. It's really important to accelerate our time to market and reduce our costs.

That's really cool to hear. That's a lot of data flowing around. Why don't you tell us a little bit about how Amazon Quick Suite is helping you think about those Neuerklasse processes and such? We are developing software inside the car and in the cloud in a distributed global context. For instance, the connected drive backend offers premium digital experiences to more than 24 million connected vehicles. We process and interpret 184 terabytes of data traffic per day and manage billions of requests.

The BMW Group software community has more than 12,000 software developers working on 500 million lines of code with daily builds. It shows how complex it is to develop a premium digital car. We have been working a lot on automating our workflows with AI and with generative AI, not only to reduce costs, but also to enable our software developers. Successful examples are, for instance, our cloud bot analyzing and screening our cloud accounts to make optimizations for operations or to reduce costs. To visualize the results and the findings, we use Amazon QuickSight.

We also use AI to chat and visualize our data that are stored in the cloud data hub, the BMW central data lake. For the last two years, we have had a QuickSight-first strategy for our consolidated data assets stored in our AWS central data lake. In the last two years alone, the number of BMW QuickSight users increased fivefold. Since the launch of the new AI features in Amazon Quick Suite, we have been actively evaluating the new capabilities.

My department has been one of the first to evaluate Amazon Quick Suite. What we have done, for example, with an AWS team is to evaluate how to automate all the tests, the full test cascade from requirements to test specification, generation, execution, and then evaluation. A service like Amazon Quick Suite can help us there. Since we can use it to trigger events and as it's serverless, we only pay for when we use it. What I also like a lot is that it's low code, so it makes it easy to use and fast to prototype.

However, we still have to gain more insights on the cost and benefits and we'll adjust our strategy accordingly. That's great to hear, and one of the things you pointed out is that it's not just the time, but it's also the fact that consistency of results is important. Being able to repeat this is very important. Thank you for that, Nathan.

3M's Sales Effectiveness Transformation and Future Outlook

Let's switch to you. So we heard about research, we heard about development. I know you're tackling a whole different type of problem: sales effectiveness. I was amazed to hear that you have over 100,000 one-on-one meetings that you're managing across the team. Let's jump into some of the insights from you.

That 100,000 number honestly was a head scratcher for me too. But when you understand the context of 3M, I'm sure that when you think about 3M, the first thing you think about is Post-it notes and Scotch tape and so on. However, a large part of our business actually is selling very complex material science solutions to customers across the globe. You have a complex product, you have a complex process across a global footprint, and that's why you have so many of these sales conversations that happen because it's a long sales cycle.

For us, as we were looking at it, we were trying to solve two problems. The first problem was one of content. When you have a large organization that has a storied history like 3M, you have application sprawl. Our sales team has to go through multiple applications to collect data and collect insights from CRM systems to databases and multiple applications. They were spending so much time trying to pull all of this information together. In fact, more time was being spent on pulling information rather than actually analyzing information.

The second problem was context. We have the sales team dealing with so many customers over a long life cycle, and there are behavioral interventions like sales coaching that are necessary. For sales leaders, it was more about first finding this information, and on top of that, they have to analyze the information. So the second part of what we wanted was how can we provide intelligent context so we can elevate the quality of these coaching conversations.

Across those 100,000 one-on-ones, we have to do both, and that was essentially what we were trying to solve. That's a tough problem. Can you share a little bit about how Amazon Quick Suite helps with that whole process?

Initially, we did a lot of experiments with a lot of AI tools. I know you talk a little bit about the first wave of AI and the agentic sprawl that was happening, and we have a lot of that. We have disparate tools which we're trying to be intelligent but within the context of the system. So that's when we started working with the AWS team early in the Amazon Quick Suite journey. For us, the biggest part was unified context. By leveraging the browser plug-in feature, we could switch context very easily. We could use quick actions to go to a database or to the CRM platform, gain insights, and bring that back, but we could do that as a plug-in within the context of whatever web page the team was browsing on.

If they were on a website looking at complex product data and trying to compare products, if they were on the CRM system trying to get insights about their sales plan and map sales effectiveness, they could do that. If they wanted a summary of their sales targeting plan and how that would work, that could be done as well. All of this was filtered into a sales coaching application which was across the sales team. This was another bespoke experience for the sales leader so they could do it.

Just by leveraging a few of the capabilities of Amazon Quick Suite, we were able to create a unified experience that essentially was able to take away the application sprawl problem but also bring in coherent intelligence that helped both the sales team as well as the leaders in driving a better plan forward. I also love the fact that you're using the browser plug-in, so you're actually taking it to where the user is without them having to navigate to a certain place.

Exactly. That's great. So I know you've all shared where you are right now in your journeys, but I'd love to get maybe a couple of lines about what's next, and maybe starting with you, Celine.

Well, in just a few weeks from now we will introduce the new enhanced intelligent personal assistant at the CES, setting new standards for in-car intelligence. Yes, it is exciting. Language model technology inside the vehicle is setting new standards for voice interaction, making it more intuitive and more natural conversation. It shows the combined innovative strengths from BMW together with AWS and Amazon, uniting cutting-edge automotive expertise with state-of-the-art cloud and voice AI capabilities. At the organizational level, we need a unified open platform to automate our workflows at scale. That's why we keep working on our potentials and we'll give you feedback on a Quick Suite, and I can't wait to see what my teams will build in the next week. I'm excited. Thank you for having me.

Well, Shelly, how about you? Yeah, so I've been reflecting on this a bit. Really, truthfully, to be pioneers in science, you need to be pioneers in technology. We know that technology is going to be the force, so we're really excited to see the emerging tech that's coming. We continue to partner also with AWS because I think this is a really good strength that we have in trying to actually accelerate a lot of the R&D pipeline as well. So I'm really excited for that. Thank you.

Nathan, I know you're delving into other areas. Yes, you know, for us it's about vertical expansion first of all within the product suite, right? So within Quick Suite as we look at it, we're only using a subset of the features. So as we think about our sales and marketing functions, it's about how do we leverage research, how do we leverage automations to kind of grow within that domain. But a company as large as ours, there's huge opportunities for automation across the spectrum, whether that's our complex supply chain, our finance and enterprise functions. There's huge opportunities for us to leverage the automation capabilities to drive efficiencies into our business process. So there's loads of opportunities.

AI-Powered Customer Experience: Four Pillars of Amazon Connect

Well, thank you all. I appreciate it, and a big round of applause for our panelists please. Cool. Well, hopefully that was insightful and you got to see some real customers using the tools and getting improvements. I think that's always far more compelling than just hearing about the products, but there's always a good balance of both. So I'm back up here to talk about the area that's near and dear to my heart, which is AI-powered customer experience and just customer experience in general.

When I think about the experience you probably still have sometimes, hopefully less than you ever did in the past, it's one where you call in and at this point have grown frustrated with the experiences and you were saying things like, "Hey, operator, operator, can I talk to a person, a human being?" and "Can I talk to your manager?" You may have learned words like, "Can I talk to a retention specialist?" Anyone out there know or heard the term retention specialist? That's not something a human being should know, right? And when you got there every time it was, "Let me tell you my problem all over again." So while things have improved a lot, we're still facing some of these challenges in the world where customers are today. Though, they really are building solutions that bring all the context from the beginning to the end, and with the launches we have here at Reinvent, that's becoming even more powerful.

So if you think about a customer and an agent really being on the same team, treating each other like friends, as they would treat a friend, I should say, is probably the right way to think about it. That is fundamentally different, and the AI can help enable that because it can bring context for all of that and let the two people focus on the human aspect of it. Particularly the agent getting away from that feeling that they're using tools and heavy lifting and trying to get things in the right order and back into an experience where they're passionately engaged with the customer, showing empathy, showing understanding, because the AI is making the solution technology recede and bringing forward the things they need to solve the problem.

Let me give you a quick example. Hopefully this didn't happen to any of you on your way here, but I'm confident at least at some point in your life you had a flight delay. And in that experience, you probably found out about it when it was already too late to do much about it.

And at that point, you then try to maybe use the app, maybe you try to use the chat, and the chat maybe tried to sell you something instead of help you. And then of course, you went through and eventually you probably ended up calling somebody to try and get them to help you. And when you did that, you probably started all over again. At that point, what happened was maybe they solved your problem.

Now, imagine a different world where when you call in, immediately you get the question: I see your flight's been delayed. Would you like me to try to book you on the next one? And then all of a sudden that changes fundamentally your understanding of what's happening in this interaction and your willingness to engage because you can see this thing is not just giving you an answer that's rote and useful to no one because it's designed for everybody, but instead it's personalized to your needs and it's predictive of why you're trying to call. The technology to do this exists today, but unfortunately up until now it hasn't really been brought to bear to help you achieve those outcomes. And so we're working with our customers to do that right now.

Now, the next step beyond that is honestly to instead of saying, hey, I see you called, how can I help you? Oh, is it because your flight was delayed? But instead move to a world where I know you're a business traveler, you probably need to get there earlier. I can predict that there's likely a weather problem. Maybe I actually want to pre-book you or offer you the option to pre-book on a flight earlier than the one you were going to take. So instead of getting you there six hours late when you're not going to make your meeting, I get you there six hours earlier and maybe you actually get to make your meeting. And of course you can always say no, I don't want that, but it gives you the opportunity to then be ahead of the problem instead of behind it.

And so this is a world where you are not just getting a better outcome for the customer, it's also a faster outcome, it's a cheaper outcome, and it's a customer sensation that feels like they're understood and cared about by your brand. And so these are the things we are delivering a product to enable you to do. And this solution that we deliver is based on four key things that we are really focused on at this launch. And those four things are being powered by literally 29 new capabilities we launched this week. It's an insane number. I had to memorize it like the states, you know, Alabama, Alaska, agentic AI agent assistance. It just rolls right off the tongue.

So with that in mind, the first area I want to talk about is action with AI. And if you think about it, that is the big promise of agentic AI: that it can actually act on behalf of the customer, on behalf of the company. To do this, Amazon Connect bravely moves into an agentic world, and you can actually do a 100 percent agentic implementation on Connect and never have any human beings or agents. We don't think that's typically the right solution for someone, but it's capable of doing that. You can also do a fully human one and we probably think that's probably not the best either. The nice blend of those two is really where you want to be.

But with this new capability, you can orchestrate agents and we are bringing the capability to bring third party agents in as well. And so we're giving you choice, which is something I don't see anyone else really doing in the industry the way we are. And so this really fundamentally changed the way you implement agentic AI. It means you can enter that space at your own pace and move forward. And the great thing about this is these agentic upfront self-service capabilities feed directly into the experience of the agent and even after the agent has completed it.

So you think about an agent sitting there trying to help a customer. You don't want them scribbling on Post-it notes or trying to take notes in some other thing or going and opening up a task to follow up. What you want them doing is listening to the customer and helping them. And so in the background, we can actually take all the context from before plus what's happening in the conversation right at that moment and even data from previous to that, and give that right to the agent and the right things to do, the right things to say. And also meanwhile in the background, we're creating tasks to follow up.

We're listening and say someone says something like, I want to change my password. Actually, the only reason I want to do that is just so I can cancel my subscription. An agent who's well trained might still say, OK, let me get that password, and I know you're counting me on average handle time. How fast can I do my job? Well, obviously that's the worst case scenario. Instead, what you want to do there is you want to say, OK, I'm going to help you, but I'm also going to try and figure out why you want to cancel. I'm going to try and show you the value of my product.

And with AI we can even get ahead of that and say, hey, we should look at what happened well before we got to this point, well before the person got to the point where they took all that effort to contact you so they could cancel and show them the value of the product at the time when it makes a difference to keep them as a customer. And so this is key to how we think about elevating the workforce: every step of the way we're helping them be more human, do the things humans are great at, putting them in that place. So maybe an agent, maybe even should have contacted them a while ago and help step them through what they need to get the value out of it. Oh, I noticed you haven't been using this, all the capabilities we have. Let me help you with that.

OK, and so in order to do this, you need data. Folks here, I'm sure if you've done any of this work, you've discovered that getting data out of your systems is really hard. Almost nobody I talked to says, oh, we've got the data problem solved. Well, in Connect, what we've done is we've created a set of capabilities built right into it where we can pull data from all over your enterprise. We break down those silos, we bring together order data, claims data, and customer data into one place.

We can store it there for you and be the source of truth, or it can just be a clearing house for you to have incredibly high scalability and performance. This gets that information into the hands of your agents and into the hands of your AI to make much better decisions and take action. This is a game changer for folks because this is how you actually build relationships.

We have built merging capabilities that are leading, and we have enriching capabilities to help you make it better. We're also building predictive capabilities, which is brand new to Connect, that allows you to understand what the customer is likely to want and what you should do to help them get it. This can be something as simple as upsell and cross-sell, which obviously some people care a lot about, but it's also about understanding what the customer is going to do. Are they likely to want to cancel? Are they likely to want to do more? What should you be thinking about with that customer? This is an awesome area for us to build together and an awesome place for you to take advantage of.

Finally, all of the outcomes we're trying to deliver are things we want to get faster at delivering. Every decision point and every aspect of what you do within Connect is being enhanced by AI. You don't have to do any integration work, you don't have to do any work to try and get it, and you don't have to go negotiate new pricing because the pricing is built into Connect's core price. You can use Connect AI everywhere at one low price, and you don't have to do these cost-based trade-offs that have been killing you.

You get easy integration, easy implementation, and you don't have to have an argument with procurement over getting access to it. You just use it and get great results. Those are the key things to how we think about AI in customer service. Last year on a stage a lot like this one with fewer chairs, I gave a talk and told people how excited I was to see that six billion minutes had been processed in the previous year using Connect AI. I thought that was awesome.

Priceline's Real-World Results: Live Transcription to Quality Assurance Revolution

Well, just a year later, now it's been twelve billion, so we've doubled the amount of AI usage in just a year. This is not slowing down. Folks are getting real-world results with this, and that's the whole thing with it. It's great to talk about AI, and I'm sure sometimes you've been pressured to say, "What are you doing with AI?" and your boss might be asking you, "Have you done anything with Gen AI this week?" I think that the real answer is what a customer is doing with the product, and that's why at this point I'd like to bring up one of our great customers, Priceline. If you don't mind, I will welcome to the stage Sean Huberty. Thank you.

Thank you. At Priceline, we've been heavily leaned into AI for many years now. Brett Keller, our CEO, has been speaking about this frequently in the news. I'm not going to spend time today talking about Penny, our agentic agent that might help you book a flight or help you with a post-booking experience. Instead, I'm going to talk about some experiences where we're using AI to help our customers, our agents, and our business perform better.

Let's talk about the telephony space and where we're using Amazon Connect. Live transcription is a service we've been using, and a fantastic example is when a customer is calling in to book a flight or a hotel and they say, "I'm looking for a hotel, I'm looking for the Tulalipin in Puyallup, Washington." We have sales agents that are overseas, and English is probably a second language for them. For them to try and understand and spell that correctly is a challenge. When we have live transcription that pops up on screen, the agent can see it instantly. It builds rapport with the customer and helps avoid frustrating customer questions like, "Can you repeat that?" or "Can you spell that for me?"

There's fantastic technology coming in accent neutralization, and this is tuned to specific geographies and traditional accents that maybe some of our associates have. If it helps a customer understand our agents better, and maybe it can deal with some noise cancellation, that's really cool and powerful technology that we're leaning into more and more. Call summarization is a place where we record the transcript, and at the end, we summarize exactly what's going on for the customer. I'll give you an example of that in a second.

Smart dispositioning is a way to take that recording, summarize exactly what's going on, and avoid the agents having to choose dropdown scenarios where likely they just choose the top example anyway. It gives you far better insights and actionable data. It makes it much more real time, so you can search and understand exactly what emerging issues are happening with a far better level of fidelity.

Finally, there's the concept of guided workflows. Imagine a step-by-step solution built into your CRM where the agent knows, because transcription is happening in real time, exactly what the next best solution could be. The system then guides them into that scenario, so they're getting the right answer at a far more frequent rate.

Let's take a look at a typical agent summarization. If you see the box in the middle right there, that's a real-life example of agent notes that they've drafted post-call. Sometimes agents are doing this and multitasking while a call is happening, or they're spending time after the call drafting these notes. You can see the real-life example on the right of what happens when we use automated call summarization. The clarity is far better. Imagine you're a follow-up agent who has to read and understand what's happening. It frees up that agent so they can focus on the customer, build rapport, and truly listen to the situation the customer is going through, resulting in a far better level of empathy.

There's also a huge benefit to Priceline. We save 50 seconds on average by implementing this solution. Those are real dollars to the bottom line. It's a far better agent experience and a far better customer experience, so we're heavily leaning into this.

The trick is how you change agents' behavior. People have been recording and writing notes for years and years. We actually had to use some forced behavior change and limit the amount of after-call work time available just to get them to trust the AI and see it in action. Within about a week, the behavior changed, and we're realizing that benefit.

Let's talk about the quality assurance revolution and what's happening. Today, if you go to a traditional contact center and think about QA, agents receive about a three percent sample rate. Maybe they're going to get around three to five coachings in a regular week if they're really lucky. What happens is a supervisor takes a recording of a call, reviews it, scores it, and eventually passes that back to the agent. The agent has to review it, maybe a week later, so they may not remember exactly what's going on. They have to re-familiarize themselves, and then there's going to be a debate around the subjectivity of the content matter. Did I get the right solution? Did I get the wrong solution? Did I display empathy?

What tends to happen is you score things like whether the agent said, "Hello, my name is Sean, how can I help you today?" Those very binary tasks tend to be the basis of the scoring. With AI solutions, you can create whatever questions you want to be scored on, and you can capture insights on three levels. First, there's compliance. Maybe you're a financial institution and you have to read your license number at the beginning of every call. You'd ideally like to have one hundred percent compliance in that scenario. Second, there's process and procedure adherence to make sure you're getting the right solution. Third, there's a level of soft skill evaluation. Are you building rapport? Are you displaying empathy? Are you truly understanding what the customer is talking about?

By using AI, you get a far more unbiased, data-driven approach, and you can customize the level of questions you're asking and the components you're evaluating. We're not quite there yet, but imagine a time very soon where you want to go up to one hundred percent coverage of all your agent interactions. Maybe you want to provide real-time coaching, so you provide little micro-learning snippets to your agents five to ten seconds after a call, so they can see what they could improve on and what they did really well. Not all coaching is corrective, but a lot of it is positive reinforcement. When you can provide that in the moment, it helps people get better and you save all kinds of time with agents not having to review things or your traditional QA staff not having to review things a week later.

I think there's an incredible amount of exciting things happening in this space that are going to change the way QA happens. It's going to change the jobs of QA staff. They're going to become far different coaches than they have been today, and it's just an exciting space that I'm looking forward to seeing the developments that are coming across the industry.

I believe that AI is a tool that enhances human capabilities and compensates for human limitations. Together, if used well, you're going to elevate the experience for your customers, for your associates, and for your business. Thank you.

Amazon's Internal Adoption and Call to Action: Start Your AI Journey Today

Now I would like to welcome Pasquale and Joe back to the stage. One of the questions we get a lot has been about how we use these tools internally at Amazon, and specifically Amazon Connect. People are typically familiar with Amazon customers. Well, thank you very much. We definitely appreciate it. One of the things I hope you've seen is that we strive to be the Earth's most customer-centric company, and in doing so, we knew we had to have great customer service. Connect is actually powering that customer service for Amazon.com, and we call it Amazon CS. Of course, we have many more businesses than that, whether that be Ring or Audible, maybe you're using those as well. We love to give great customer service there. We actually partner closely with those folks as we do with our external customers.

When you think about it, we have 100,000 customer service agents at Amazon CS alone, and those folks are trying to deliver great outcomes for people like you every day. They're spending every waking hour thinking about how they can do better, and that creates a partnership that is incredible for us to learn from. Processing millions of customer conversations every day, we've learned a tremendous amount that we then share. We also learn from our external customers and partner with them as well, but this is a very deep symbiotic relationship, and it's what allowed us to move so quickly at first with Amazon Connect to deliver great value.

Every day, what you see if you call in, hopefully, is you feel like someone who would want to be your friend is trying to help you honestly to get every aspect of your problem solved as fast as possible in a way that makes you feel great about it. We even do things like connect folks in Audible with agents who might read the same books, so you might end up in a nice conversation in the middle of this that creates a different experience. These are the kinds of things people do with Amazon Connect, and these are the kinds of things we learn with Amazon.com and our other internal teams.

Let me switch that over to Quick and how teams are using Quick. We launched Quick to Amazon internally a few months back, and what we've seen, the response has been tremendous. Today we have hundreds of thousands of users who are using Quick internally. We've had over 50,000 research reports created and over 300,000 flows run. Teams are using this for all sorts of domains, whether you think about sales teams and account teams that are preparing for meetings coming up using flows and research, finance teams that are partnering and using Quick to get briefings for their weekly business reviews, supply chain teams that are digging into their systems using Quick. The list goes on. Over 40,000 agents were created for disparate purposes within the company, all within Quick. It's great to see this adoption and these numbers.

One of the big questions that comes up as you think about this, and I'm sure this is the question on your mind, is what about safety, security, and trust? Remember that both Quick and Connect are built on the same enterprise-grade security framework that AWS has, which means that your data stays your data. There are controls for encryption, and permission frameworks are respected. All of this makes sure that you are able to deploy this in the ways you want with your data being safe and secure.

I also mentioned context earlier, and Pasquale mentioned the same thing. Connecting with your integrated systems, whether that is your CRM, whether that's your knowledge bases, your business analytics, your collaboration tools, all of that is available within both of these products. With Quick, we are delivering better equipped employees. And with Connect, we are delivering better customer outcomes.

If you don't mind, I'd love to leave you with three thoughts. If you only took away three things from today's talk, the first one is we truly believe in and are seeing proof that the future of work is AI plus human teammates working together to build a better environment for both and to create an incredible outcome for you, your customers, and your workforce. The next thing I'd like to say is you can start today. The things you saw today are real. There are things you can actually do with your business right now.

The last thing I would say is don't wait. Do it. Pick something really good, some really good opportunity you have. Maybe it's a small workload, maybe it's something where you want to bite off a part of something and just get going. You're going to see incredible results right off the bat, and the learnings you have right now will accelerate your outcomes moving forward.

At this point, I want to say that one of the reasons why I'm very confident in that is the folks who got up here today talked about their own lived experiences. You heard from 3M, BMW, AstraZeneca, and Priceline, and you saw they're making real results right now. Come on with us and let's go do it together. We'd love to do it with you too. So thank you so much for having us.

---

; This article is entirely auto-generated using Amazon Bedrock.