SaaS application innovation using AI to improve employee productivity

Hello, welcome everybody. Welcome to Session Biz 109 SAS Application Innovation Using AI to Improve Employee Productivity. We're so glad that you're here and we really do hope that you've had a great week at Re:Invent.

I represent AWS at Fabric and AppFabric has had a large full week this week at Re:Invent. And we're excited to dive deeper uh on the topic of what AppFabric is doing uh at AWS and how we're solving some really great interesting problems.

The session specifically is for SAS application builders who are interested in developing new experiences within your application, within your end users. UI. We're specifically going to be talking about how AppFabric helps you use generative AI in your application. Let's begin.

My name is Holly Watson. I'm a senior product marketing manager on the AWS at Fabric team and I'm joined today by my friend and my colleague, Anant Patel.

Hey guys, excited to be here. Thanks for coming. Today. We're going to be discussing the how the rise in the adoption of Sass applications has created a handful of different challenges, challenges that AWS Fabric is taking uh head on and solving both from a security perspective, as well as from an employee productivity perspective. We'll dive into what that means for SAS application builders and developers interested in using and adopting AWS Fabric. And of course, we'll explain the details on how to get started.

Let's begin with the challenge. The number of Sass applications used by customers around the world has grown significantly over the last decade. In our discussions with customers in our interviews with end users, what we repeatedly hear is their preference is to use best of breed purpose built SAS applications. They wanna use these functional or these particular apps because they're feature rich, their capabilities are really full, they dive deep into the functionality, they need to perform their jobs best.

In fact, what we see is that users uh prefer them over monolithic platform suites and these patterns have accelerated the adoption of SAS applications. The average number of SaaS apps used at an organization, even among some of these end users, these customers of yours in the audience who are using these platforms or these suites, they're still supplementing those uh those suites with four or more additional applications to perform their job. And as I've mentioned, these, the vendors of the applications, they, they are, they're being loved and uh preferred over suites because they're so feature-rich their capabilities are deep and integral to performing end users jobs.

But the challenge presents itself that these applications are they don't always work well together. They don't always speak well together. They're not always integrated and that creates security challenges when managing the corporate data. What ends up happening is security and IT teams charged with keeping corporate data secure, have to begin building point to point integrations across the set of applications used at an organization that's challenging work that takes time and it's costly to manage and maintain over time.

Additionally, employees across the organization feel this tug and pull. They're having to jump between applications uh up to an hour every day as they try to complete their tasks and think about it. If you think about your computer that you might have open, you might have a, a few handful of tabs, your messaging app, your email, your calendar, right? Just to complete a task. And so that type of uh swivel chair jump between apps is, is really draining on your productivity.

These are some of the things that AWS AppFabric wants to take head on. And in June of this year, we introduced AWS AppFabric which connects multiple Sass applications together for better security, observable and end user productivity.

IT and security teams can use AWS Fabric to abstract away the need to manage point to point integrations across their Sass applications. AppFabric automatically normalizes the disparate data coming from each of these applications. We use a couple of different formats to do so the Open Cyber Security Framework or OCSF. We also offered uh the normalization in Parquet and Apache. Again, this is uh this is done through at Fabric. So all of that data can be fed into at Fabric and then sent to your security tool of choice.

And we work with some pretty amazing brands. Really excited to show over 14 different productivity applications that we're already integrated with. And we're compatible with uh five different security tools. With more coming every day, brands like Asana, Zoom, Miro, Atlassian and Jira. Security tools like Splunk. The vendors of that, we've again seen a large adoption of from our customers even.

And so understanding that App Fabric works in the background to connect these applications again, for better security helping IT teams normalize the disparate data coming from these applications. We knew we could do more, we knew we could use that premise to actually push the bounds and solve those challenges for end users using these applications.

And so again, as I mentioned, many of the SAS productivity applications we use are intended to make us more productive. But when they're not speaking to each other, when they're not integrated data silos form and work today, to be very honest, it doesn't happen in a singular team. You're oftentimes working across your organization collaborating on projects. Some of those tools are preferred by a marketer. Some of those tools are preferred by an engineer and some of those are preferred by your product team. Maybe finance data gets become siloed in their tools of choice and it breaks down your productivity, it breaks down your collaboration, ultimately creating that swivel chair and leading to what we're calling uh technology overload.

And I personally can admit leading launches, working with a ton of different teams. I find myself often questioning is that document in my message. Did I log it in my storage repository? Right. What is my leadership preference? Do they prefer a Slack message or an email? Oh, it's cognitive drain. It's tough. We want to do better. We feel like there's a way that we could collaborate more productively really.

And the challenge of technology overload. As I mentioned, you have these disconnected uh applications, it creates data silos the context switching results in uh this cognitive burden going back and forth where I leave that my document. How can I better collaborate with my product team, my finance manager, legal and ultimately, it's this cognitive burden that just drains productivity. It's a drag on productivity.

So again, we saw an opportunity to do more and enhance employee productivity using AWS AppFabric.

Awesome. So when we think about this technology overload that Holly is mentioning, we think about the applications that we all like to use today. We're all we're as AWS. It is super fortunate that they're choosing AWS to build these applications. They come to us for our reliability, our operational excellence, our security, our global footprint to bring their services to their users and all host of other services that help them run their applications.

But when we talk to these app developers and we ask them, what are the problems we're trying to solve for? What are they trying to do going forward? The most common thing we are hearing now is around generative AI which is the latest buzz. When they talk about generator AI, they see many different ways that they can apply this technology inside their applications to help their users get things done faster.

The first one is around collaboration. How do I take this synthesis of all this data that's happening in my application and make it really simple to understand, extract out intent, summarize information in a matter of seconds. It helps your users use that app much more efficiently.

The second one is around content, collaboration or generation, excuse me, how do I create emails, presentations, other artifacts that can be built on top of my data that can now extend my data in new ways. They're also thinking about automation. How do I take this chatbot experiences that has become very popular now and make it self served for customer service and troubleshooting.

And last, we're also talking about security. Sass security is also a high priority for a lot of organizations when it comes to adopting SaaS. So how do you think about using generative AI to help investigate security issues to create rules and alerts to help prevent that going forward.

But as we start to think about these problems and specifically around generative AI, we can start to think about how are apps starting to build generative AI experiences. Many of the apps that we showed on the previous screen have already released generative AI capabilities. And what the foundation of all of those capabilities is really around their data. And we heard in the keynotes earlier this week, the data that an application has ultimately is their differentiator. There's a lot of rich contexts in there that's super useful.

But for humans and us as users of the SaaS applications, all of our information is not scattered is scattered across multiple applications, not just one application. So really, we have to start thinking about how do we bring these applications better together in a more meaningful way because right now what we start to see is every application has their own generative AI capabilities, but it doesn't really change the core behavior of apps, switching and contact switching that Holly was talking about.

We still have to go to these each individual application to extract out the insights. And yes, the time to do that is now much faster with generative AI. But the challenge is we're still doing the same behaviors as before and without those missing context and missing data, it ultimately leads to these data silos will continue to form and the generative AI responses are going to be missing a lot of critical information that actually help us make better decisions.

So the second piece is around if I decide as an app developer to actually build integrations, what do I have to do now to actually get this data into my application? Well, first you have to build the integrations across each individual application. But the key challenge with that is every app has their own data models, their own APIs, their own schemas. So it requires a lot of tedious manual data transformations and data normalization in order to make that data super useful for me.

Now, on top of just building the integration is also managing it APIs change, new features are released.

And so it becomes this really tough decision with a finite amount of resources. Do I invest in innovation or do I invest in trying to build more integrations than my end users want?

And the last piece is around just trust with generative AI. As we know, this is a very new thing that's kind of came on the on the scene pretty recently. But there's things that are still we're still working on. There's hallucinations, there's security, there's responsible AI. All of these components are things that we still need to solve for in order to make generative AI useful and make end users actually adapt this technology.

So based on some of these problem statements, we're super excited to introduce you guys at Fabric for Productivity, which launched as a preview on Monday of this week.

App Fabric for Productivity is focused on helping app developers build more personalized app experiences or enriching existing AI capabilities with the context for multiple applications. This feature combines the goodness of generative AI as well as all of the user data to really focus on how do we make that user much more productive in the applications that they prefer?

So let's dive into a few of the the details of what this feature offers.

The first is we offer generative AI generative AI powered APIs to application developers to generate insights and actions from multiple applications. Insights is a very vague term. But think about it. How do I help prioritize my focus? How do I optimize my workday? Those are the types of insights that we're starting with, with these, with these APIs but the insights are just part of the problem. What do I do about these insights?

So we actually also enable cross app actions from my preferred application. I can actually complete actions across multiple different applications. So it allows me to stay within my workflows, allows me to execute workflows much more simpler now than having to navigate from application to application.

The second part is around how do we actually find the right data to generate these insights? So because AppFabric is now going to build and manage all these integrations across all the different data sources that are used to provide these insights, you don't have to do that anymore yourself.

And as far as building the integrations, we also go as far as normalizing the data. So that if I need to send an email, it doesn't matter if the email is in Outlook or Gmail, we're taking care of all that normalization complexity behind the scenes so that the objects that are common across multiple applications can be easily tracked and identified.

And that's right. Another core tenant of AWS AppFabric for Productivity is that we are not introducing a new UI but that sink in. We are not introducing a new UI you as the SaaS application developer choose how AppFabric manifests in your existing UI.

That means your end users are not having to pull away from your home page, your dashboard where you want them to be your end users stay in your application to complete their tasks. AppFabric works behind the scenes supporting your goals with generative AI in your application so that those end users stay in your app, they are surface recommendations. As Anant mentioned, they're recommended actions so that they all can perform their day of tasks, they can stay in their flow of work in their application of choice.

Additionally, we are very keen on making sure that we are developing trust, not only with our application developers, but also with those end users all the way through. And so a key component of AppFabric for Productivity is making sure that we are showing our work, we're showing the sources, the files, the information that's coming from the recommended actions and the uh the insights that we're driving.

So your end users can quickly click into the material that's being surfaced and verified. This makes sense. This is a good recommendation. This does help me draw some more insights to make a better decision.

End users control what apps they want to uh enable or disable. At any time, we've centralized that in a quick on boarding process that we will show in a moment. So again, your end users have control of the applications, they want AppFabric to be able to access and use to generate those personalized user insights for them.

And the best part is is that we are already working with a handful of different brands as sources. These brands include Asana, Atlassian, Jira, Miro, Slack, Smartsheet, Google and Microsoft 365 sources that are really used by many of us to complete everyday tasks.

So we're reimagining productivity with AppFabric for uh for Productivity, we're wanting to leverage all the goodness that can come from the large language models and surface them in your app of choice. So again, we're empowering your end users not only to uh choose where their center of gravity is, but also how you as a SaaS application developer want to actually surface those experiences.

Alright. So we're gonna do just a quick snapshot of the end to end service, what's happening here and then I'll do a bit of a deep dive on how, how this whole thing works behind the scenes.

So on the left hand side, you do see all of the some of the data sources that Holly just mentioned because we're building integrations with these applications. This is now we're bringing in and aggregating user content. We're doing the normalization and bring it into the middle here where AppFabric is helping combine that data and using generative AI to essentially bring out cross app insights and actions and surfacing it inside these applications that you'd like you'd love to use.

So how does this actually benefit the app developers? So now as an app developer on the right-hand side, what's happening now is you have a single source of cross app user data across all these different applications that you can now use to build and enrich new experiences in your application if you're an application on a left-hand side and you're more of a data source, well, now your data is actually being surfaced inside other applications and these other applications are thinking about how do i take that data and build net new experiences.

So the combined effect of this is now you're building net new experiences and people are building net new experiences on your data. And there's a lot of power that comes there from an end user perspective, when all these applications are truly connected together.

Now, when we think about the generative AI component of this at AWS, we're really focused on democratizing AI. And we're taking the exact same approach with generative AI. There's so many different large language models out there, each with their own characteristics, strengths around performance cost quality, et cetera. But really our goal is to focus on how do we bring the right insights to end users in a safe and secure manner.

So we're really excited to as our starting point to use Anthropic CLAD models to really generate insights for our, our features. And we we've been partnering closely with them to really understand how can we be responsible, how can we build that trust with users? They're founded on AI safety and that really resonates with us as an organization as well as us as a service team as well.

So we're really proud to continue to work with them and want to continue to work with them to expand these features in the future. But yes, we have Bedrock available in case you want to have the options for choice and adaptability as this thing evolves. But this is a great starting point for us and we're super excited about it.

Today, employees typically switch between 6 to 8 SaaS applications from different vendors to complete tasks, switching context, constantly application developers see an opportunity to use generative AI to reduce this context switching and provide a richer end user experience but are often limited to data within their own software without including cross app data, data silos form and generative AI results miss important context from other applications that help end users make better decisions.

AWS App Fabric for Productivity, reimagines productivity in SaaS applications by generating insights and actions with context from multiple applications. AppFabric fully manages the integrations with multiple SaaS applications and automatically normalizes data for use across apps. This means developers can enable AppFabric productivity features directly into their applications, user interface to maintain a consistent app experience for their users. Whilst surfacing relevant context from other applications with cross application context, developers provide a more personalized user experience that increases adoption and loyalty.

End users benefit from accessing insights they need without interrupting their workflow Embed AWS App Fabric in your application's new or existing generative AI assistant to enable your user to be more productive and more connected across their preferred apps.

Awesome. So that that video gives you a quick recap of what we've talked about today. But now we want to go a little bit deeper inside the specific APIs that are available in preview today that app developers can use to build new experiences.

The first one is around actionable insights. It's this idea of all this content and information that's happening across your applications that you may not be aware of how can we proactively surface that information in, in the flow of work so that you can now act and optimize your day better.

Let's take a very simple example. Sometimes we're stuck in a meeting. There might be a customer escalation email that comes in. Your team is already working on, on messages on what to do to solve this problem. And you've already created tasks to actually build the the next steps to mitigate this problem. You come back from a meeting and now you have a swirl of information. What do i have to do today to actually get this information? I got to go to each individual application. I have to figure out what the right information is. What happened. Why did it happen and how are we going to best proceed with it without Fabric actional insights?

We can do all of this work for you. We can look at all the context across the different applications and we can summarize it into a very succinct insight that allows you to act faster.

So let's break down the left-hand side of what you guys see there. The important updates side of the the the screen at the top. You see an insight title which tells you very, very quickly. What is this insight about? The second part below? That is a few sentences that says, give me more details about this insight.

What's really cool about this though? is that generative AI is not only providing a summary of what's happening across all your different applications, but we're actually surfacing the hyperlinks. So as Holly mentioned, these reference or sources show me your work. Now we can tell you which specific applications, what specific artifacts were used to help generate that insight. And it allows the user not to click into those individual applications if they want to. But now they don't have to if they don't want to.

The second half of that is like I mentioned actions. So I had this customer issue. What I want to do, I probably want to schedule a meeting with the customer or reply in an email. We're at a as a starting point. We're actually also enabling cross app actions that can be used and executed directly from those applications.

So if I have to send an email, if I want to schedule a meeting, if I want to create a task, these are starting actions that we support today. But what's really neat is the SaaSs are also building on top of the insights and thinking about what should these responses be? So they're going to prepopulate information around the different fields and say what is this email body? What is the email subject? Who should this be sent to all of that information now can be done for you.

Now, the users always have control though we never want to actually execute an action without user input or validation. That's super important to us.

So the user still has the option to edit this content and then execute the information as they see fit. Now, when it comes to actions as a starting point, we're actually providing a AppFabric UI that allows you to a small pop-up to actually execute the action. Now, as a starting point is I meant that intentionally because we do see a world where we can actually provide this natively, so you can embed this capability directly into your applications. And we'll work to build that in the future.

We're really excited that last screenshot I showed you was actually Asana who has, has built and embedded this capability into their application today. We're really proud to continue to work with them and collaborate with them and think about how can we build net new experiences into their application that home page that you saw there is where their users go to to ultimately start their day. What task, what actions do I have to do to actually get work done? Well. Now with AppFabric now you have a new source of cross a data that also extends and provides even more context on what I should be focused on and why we'll continue to work with them on, on this going forward. And we're excited for them to continue to roll this out to their employees and customers.

The second use case that we have is around meeting preparation. And the the scenario here is we all get added to meetings that we probably don't know what's going on or we're not on point to present, but family emergencies happen, conflicts happen. How do you best prepare for that meeting in a matter of minutes? Now, the meeting preparation API essentially focuses on what are your upcoming meetings on your calendar and what are relevant sources and artifacts across your application that could be used to help prepare for that meeting.

So let's take this example here with Miro. So Miro is this visual workspace, this collaborative white boarding application, their users today use this application to dump in a lot of information, help organize their thoughts and then use it to organize and come up with their presentation or meeting content. What's cool about AppFabric and some of the ideas that we're talking about here is now they can use the meeting preparation API to actually help bring in that content in a matter of seconds. What's also cool about Mero Mero is that they have their AI capabilities Miro Assist and they're thinking about ways to extend this even further. How do I take the cross a data that's now on my visual whiteboard and actually use it to build a presentation for my for my meeting or an agenda for my meeting.

So now it really is this end to end flow of you have data coming in all the data that you use as a user, but you're also able to extend your app, your specific apps capabilities and build on top of that data. So it's really powerful stuff when we think about that.

So a little bit of more information now about what's happening behind the scenes with these features and how this all works. As Holly mentioned earlier, this all starts with the end users themselves providing their consent, their authorization for AppFabric to actually generate insights and actions on their data. So what we've done actually to start now is we have this user portal, we want to centralize all of this authorization flow in a simple, easy to use manner.

So we have this end user portal which individuals can go to and connect whichever applications that they see fit whatever apps they want. They go through a very simple authorization flow that we're all super familiar with at this point in time where you kind of click allow or accept and allow AppFabric to generate those insights. What's important now about the this this process is we're only going to start bringing in data based on what that individual has access to. We don't want to bring in all of the potential data. We're only focused again on user level insights and what that user has has access to.

So as we bring in this information, after the after we got the authorization, what we're doing is normalizing it and we start generating embeddings embeddings are essentially numerical representations of this data. And allows us to essentially do semantic searches to find similarities across all of these different artifacts. After we generate embeddings, we create a user level index. And I'll say that against a user specific index, not a teams specific, not an organizations, but a user level, we can do that intentionally because we want to keep securely the types of data and artifacts from each individual user and what they have access to separate.

So even if someone has access, 10 people have access to the same document, we'll create 10 different versions of that in order to keep that segregation of data strong. So what's next? What's the actual step in generating insights? Well, the first step is to find relevant information. AppFabric uses a common process known as retrieval augmented generation. Our goal is to help augment the AI's response and drive more personalized app experiences by providing relevant context that's specific to that user.

So when it comes to the actual insights or the meeting preparation API let's take the meeting preparation API for example, based on my upcoming meeting on my calendar, I now do a search, a semantic search around similarity of of content to say what is relevant for that particular meeting. And I extract that out from the user level index that we created previously. What I'm doing after that is actually bringing then all of that relevant content and context and feeding that into the LM. And the, the great thing about this step right here is we're bringing in the right context to make it more personalized. But in terms of the prompt itself, we're also actually abstracting away that complexity.

We're providing this all encompassing use case API that gives you both the data as well as the prompt and the insights. So that when you the response of that is ultimately just a JSON response, it allows you to render and update that however you see fit inside your application. It's a really succinct way to just bring in the very key details that your end users want. It's going to be the inside title. It's going to be an inside summary, those embedded links and the references to the right artifacts that were used to generate that insight as well as all the relevant actions that were there.

So this is super powerful now because we think about that single API call, does authorization, it does data management, it does insights generation, it does actions, all these things are now highly abstracted away and you can now build those generative AI experiences in your applications that much faster.

We think about where we're going to go with this, this capability. It's great as a starting point to be able to provide all this abstracted complexity away from you. But in the interest of time, we'll continue to think about ways to build more flexibility, more customization in this. But right now when we talk to customers this is the easiest way for them to start building net new experiences inside their applications.

That's right. And a key aspect of uh what we are trying to build here is of course, first working with our customers getting feedback from them, hearing understandings of what's going to work with. uh the the ideas that we have with AppFabric for productivity, we work very closely with some of those key brands you saw. And so these are just a few quotes of what those applications are saying.

Zoom, for example, is really focused in on sticking true to Zoom and still enabling that close collaboration that they offer through their platform Miro just the same collaboration is key for them. They want to drive that through their entire UI and with their end users expanding from, you know, enterprise organizations all the way through uh to you know, brand new start ups collaboration is key for Miro and they see AppFabric for productivity as a way to to enhance their existing UI.

Smartsheet. For example, another really prominent application that's used for project management. Uh they're actually looking to use AppFabric for productivity to enhance their existing AI experiences, their Smartsheet AI platform as they call it. The biggest takeaway for this is not only are we trying to work with some of the brands that we recognize that end users prefer, but we also want to make it adaptable if your SaaS application does not yet have a generative AI assistant at fabric for productivity could be your starting ground for that.

If you already do have a generative AI assistant, you're wanting to enhance it, you're wanting to expand it uh beyond what it's currently doing at fabric can also help you as Anna mentioned. These are API calls that you have choice on how it actually manifests in your UI, how it ultimately has surfaced for your end users. And it really gives you the um the control on what that means.

End users again, have full authorization, they control what apps are connecting through those steps that Anna mentioned. Uh but those are the key aspects that we are really trying to drive. And as I mentioned earlier to a core tenant for AppFabric because we're not wanting to introduce a new UI, all right.

So just to kind of slowly wrap this up here, some of the key takeaways from our discussion today, we've talked a lot about how do we bring data from other applications into your app? What does that actually mean for app developers? Well, now you have a much more stickier app experience and users are going to want to engage with that app a lot more. If all of the data that they actually care about is at their fingertips, that engagement leads to better satisfaction, that leads to better retention that leads to ultimately impacting your top line.

The second core core piece of this is that we don't require the end users or end customers to actually have an AWS account. So after you build this capability into your application, you can sell it to 100% of your customer base. This allows you again to really think about impacting your top line while leveraging generative AI in a more meaningful way for your users.

The second thing around bringing a more richer app experience as, as we've mentioned today, you have full control of your application and how you want to embed this capability. It could be a Asana’s widget, it could be inside an existing chatbot. It could be part of a dashboarding virtual workspace. There's a ton of different ways that this can be embedded and really the goal of that is, you know, your users the best, you know your app, you know how they want to use your app. So we want to empower app developers to be able to build and integrate this in a way that they see fit.

Third because AppFabric takes care of the integration workloads because we do the building and managing of integrations, we do the normalization work on your behalf. It really does allow you to focus your resources now on the right types of work, focus on the innovations that are going to drive net new opportunities for your business.

And lastly, when we think about just the quality of AI more broadly, I'll say it again, your data is your own differentiator. But end users have their data across all these different applications bringing in their data into a single spot. Now allows you to have higher quality, more relevant results for your end users that are going to drive better trust. It's going to have a more transparency on what type of information is used to generate that response that ultimately is going to allow them to start having more confidence in using these types of features.

So let's talk about how you as an app developer can get started. There's a few handful of steps and that only includes you first reaching out and registering your application with AppFabric. We'll go through a little bit of a verification process, then you ultimately get access to the APIs as Anna mentioned, those APIs would allow you again to call them, put them in your app and design the UI experiences for your end users.

That of course is then on your launch schedule and your launch time frame, right? You would be able to choose how you want to deploy this in your application and ensure that AppFabric is surfacing the right insights and the right actions for for your end users.

If you're wanting to learn more about AppFabric for productivity, please do scan this QR code. It will take you to our website, you'll learn more, not only about the applications we're already integrated with and supporting from our security aspect. But you will be able to click in and dive deeper into our productivity functionality. There is some getting started pages there where again registering your app, gaining access to these APIs is all available on this site.

We do really appreciate the time today and your guys' attention. And we look forward to taking a handful of questions during the peer talk that immediately follows our discussion, Anna and I will be around to answer a few questions if you are wanting to stay in the room. But please, we encourage you try out AppFabric for productivity. It is in preview today starting on Monday.