Duration 15:26
16+
Play
Video

Applying Conversational AI in the Enterprise | Rasa Summit

Mady Mantha
Senior Technical Evangelist at Rasa
  • Video
  • Table of contents
  • Video
Rasa Summit 2021
February 12, 2021, Online, USA
Rasa Summit 2021
Request Q&A
Video
Applying Conversational AI in the Enterprise | Rasa Summit
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Add to favorites
259
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speaker

Mady Mantha
Senior Technical Evangelist at Rasa

About the talk

Launching conversational AI in the enterprise depends on many things that are essential for successful adoption and delivering value. This talk will walk through why enterprises need conversational teams, real user data and insights, and a unique approach to launch assistants that can handle mission-critical tasks.

Presented by Rasa Senior Evangelist Mady Mantha at the 2021 Rasa Summit https://rasa.com/summit/

- Learn more about Rasa: [https://rasa.com​​](https://www.youtube.com/redirect?even...​)

- Rasa documentation: [http://rasa.com/docs​​](https://www.youtube.com/redirect?even...​)

- Join the Rasa Community: [https://forum.rasa.com​​](https://www.youtube.com/redirect?even...​)

- Twitter: [https://twitter.com/Rasa_HQ​​](https://www.youtube.com/redirect?even...​)

- Facebook: [https://www.facebook.com/RasaHQ​​](https://www.youtube.com/redirect?even...​)

- Linkedin: [https://www.linkedin.com/company/rasa​​](https://www.youtube.com/redirect?even...​)

#conversationalAI​​ #enterpriseai #aichatbot​​

Share

Everyone, welcome and Maddie. And I did product evangelism in a row. So you know, the last time that I spoke at the rise of some it was actually in September of 2019 so it feels unreal. That was almost a year-and-a-half ago and you know a lot has happened since then, right? I actually wasn't it, rather back then than a few months after the summit I started working at rasa but really A lot has happened since then, our community continues to grow exponentially and contributes to our capacity to innovate and

In just the past year, we've had millions of downloads worldwide, right? And your Enterprise companies that are embracing digital transformation across different verticals and industries continue to use. Any conversation has also grown and more and more Enterprises are building a assistants to work on these Mission critical tasks. And as you heard Ben Baker from from Gardner earlier this week, the majority of Enterprises about 67% will be using conversation and Dropbox will soon be your most important development

activity. And so we want to build interfaces that speak our customers language so that we can meet our customers where they are and allow him to Simply have a conversation to get this done on their own terms. And that's the entire premise of conversational AI, right? Is that people can just Express in their own terms, what they want and it's the assistant's job to translate understand and help until natural language interfaces have the ability to be truly transformational. So today we'll revisit some of what we talked about at the beginning of the summit, which is

scale. One of the biggest challenges was getting value. Some conversational AI is being able to supply its scale. And, you know, when you're in the process of doing that, he often started the POC Stage production before you grow your user base before, you know, you grow your teams that work on these applications and have the ability to share data between application. And today, I want to talk about some of the most common challenges that we've encountered and we seen people in Kemmer

We've talked with our customers with market research analyst experience some of this ourselves and some of the most common challenges receipt are, you know, for conversation, has to be in production, in an Enterprise, getting their required to meet Enterprise, security standards as a model or, you know, the entire software and Hardware stack. I'm getting your assistant to execute actions on your users behalf. By the way is one of the most rewarding things to do when you're designing and building conversation software to integrate with several third-party services.

And based on your assistants domains, you can start to get really complicated but it doesn't have to be We have to talk a lot about how standard software application best practices. Don't go away, right? If you're building conversation software. In fact, most of these best practices and principles lend themselves. Really well to conversational AI development and Enterprises might have varying devops expertise. You might not know how, when, or if you need to have sex with you, but it's always helpful, early in the process.

And your application will scale and complexity and scope and I will scale with users as well. And so as your user base grows the number of conversations that people have and that introduces a whole set of other challenges. And, you know, while you're trying to scale, it's really important to ensure a great customer experience throughout. So we'll talk about some of that today. Now, let's talk about how we can either plan ahead or be aware of these problems of friends. Or, you know, I try to solve them with security. Mandy, if you need to think about a few

things you need to take a Secure Storage data, privacy access controls and Enterprise security standards, of course, you know, depends on the type of infrastructure or all your system. and when it comes to Secure Storage, some platforms, like kubernetes, which out-of-the-box support for injecting credentials and environment variables, and experimenter envy, I might need customization Summoner. Prices might even mandate using the service like both too many sensitive to decorations and credentials. It's really good to

plan ahead and be aware of these up ground And with regards to data privacy rights, security mandates will tell you how to secure data and motion. And this is typically done using transport layer security and Minnie cake with each other and with the outside. So if PLS is a mandatory requirement, certificate management will be something that you'll need to plan. And, and if you're using kubernetes, you might want to use a package like certain manager to make certificate management more sleep less.

Another common strategy used in kubernetes environments is Jake supposed to service via Ingress zip ties. Out-of-the-box support for TLS termination and service loan balance. So we talked about Jaden motion but Enterprises may also require that the conversation data be secured at the rest and since is typically delegated to use the built-in encryption capabilities of these databases to secure the conversation data. And you know when it comes to it, authentic. Asian an authorization, single, sign-on or SSI. Was a de facto standard, most Enterprises and you farting already. You know, now

exercise implemented using Elle Decor Samuel and you know, Rosa Enterprise supports demo basis. So so you might want to make sure to enable this to prevent unauthorized access And you know, if you're already using rounds at Enterprise, it's important to know that the principles of least privilege when it comes to access control center on the form of our back. And you can manage your roles and permission to ensure that users are given appropriate access privileges. So you can control how your team works and which team works on what and so forth.

You can start to add way more value to your assistant, when you can get it to do the mission critical tasks in a reliable way. And for you to be able to do that, if you need to integrate your assistant with third-party services, and this process can get complicated and especially require some planning if you plan to scale and integrate your assistance with more and more thorough cleaning out your integration Point based on your assistance domain start narrow always a good practice to start small with maybe one or two absolutely needed integration. And in order to figure out

what kind of vinegar do you need really study user behavior and user engagement signals, right? Like what are users looking at or clicking on and learn about their needs that way? But you know, obviously follow privacy ethics to learn what we can and cannot do your users privacy and following ethical considerations when building a system is really Extremely important. So as long as you're meeting those ethics, standards and privacy standards and look at what he's in addition to looking at what they do. So for example, is 40% of your

users. Ask your assistant to make maybe a credit card payment on their behalf, then you might want to add this killer action. So try to come up with a data-driven approach. When you're building your integration plan and building skills his first and then based on traction implementing integration as an actual user goal, which is something that the user wants to do. If you optimize University, staff strategy based on what users need and asked for you end of building, something that will actually help them and something that they will actually use

when you come up with these Integrations right? Establishing connections to be Services, can depend on a few things like a firewall and network will be sure to not hard code or store configuration and connection details as part of your source code securely store. So you might want to do services like vultures store and retrieve these details. It's also a good plan to follow the single responsibility principle when you create custom actions and custom actions or how you kind of make it Universal points will happen. That

means single responsibility. Principle means that it's best to not perform, too many things under a single custom action, for example, if you want your assistance, but they were building a help desk assistant and you want that help desk assistant to open follow up on and clothes into management ticket. Then a single custom action should execute help me stop. You want to create a separate customer. So, you know, devops is a pretty important part of software development in general. And for those of us who aren't familiar

with Cher and tools that help you deliver software application to scale the devops tools help automate this process for example, is a tool sets that you can that can help you deploy and improve AI systems in production environments but sometimes there's a wide variation in the level of devops expertise and companies and this can cause loss of problems. Among one of those being, it will ultimately delay getting consistent production and throw all of the best practices when it comes to typical software

development. Still apply here right now, having a conversational team with Derrick representation that says service ESC, pipeline runs, automated test, that virgins, both code and content can really encourage them to newest and scalable delivery. So what are the biggest challenges with building resilient applications is that it's impossible to anticipate? All of the things that users can say to it and this is, especially true when a System Scope on user base expand or your team starts to support

additional capabilities and additional conversation has the chances that your assistant, mishandled user requests will obviously, like understandably become higher. So you still need to ensure that your assistant and behavior as opposed to forcing the user to adapt a pre-programmed conversation. Enter the best way to scale. This is not to come up with synthetic or developer generated data to follow conversation for Urban Development where you're listening to you and using those insights make your system, more resilient to these conditions. And as

your starting to do CD, which, you know, you should do as soon as possible as soon as your assistant is in the Prototype phase as your starting to do this, right? It's understandable when your assistant doesn't always, in fact, it was a lot of the time, but when this happens in a production setting stakeholders that are unfamiliar with conversational, AI Market concern. So there is a place off this problem. First ensure that your system provides a fallback response. If the user provides an immediate response then maybe try to disambiguation by prompting the user to

say, Second try to ensure that the assistant return to specific fall back responses quite off-putting. So for example, if the assistant can answer questions about credit card payments, instead of returning, a generic response Friday. Maybe a specific one know something like asking $0.02 with easier to someone on the credit card. So all of these things that we talked about are some of the most common challenges that we see at the Enterprise level and some of the solutions that

we discussed or at least raise awareness, so that we can both conversation software, that works reliably ultimately provide, the best customer experience to get things done. They don't have to try to act within the confines of the assistant. Way of understanding things that weigh natural language. Interface is truly have the potential to change the way that we do work to change the way that our customers interact with us and our businesses So we started this damn it off with, you know, couple days ago with talking about the many faces of scale and then our enter penile yesterday when you

looked at how other product leaders and themes of solving, some of these challenges its tail. And then later in the day yesterday, we know we do pin to Brazos unique approach of making assistance. Really great. And you know how you have to use your own data to accomplish that. And then earlier, today we still have T-Mobile. One of the largest Enterprise is out. There is using a set of processes that we just talked about. Now just yell reliably all the while providing a great customer experience. I don't know what that I want to end this talks and

thank you for being here. We hope you enjoyed the summit and now let's open it up to some questions.

Cackle comments for the website

Buy this talk

Access to the talk “Applying Conversational AI in the Enterprise | Rasa Summit”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “Rasa Summit 2021”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “Artificial Intelligence and Machine Learning”?

You might be interested in videos from this event

October 7 - 20, 2020
Online, Mountain View
19
5.41 K
google, googledevs, it, machinelearning, mlsummit, network, platform, tensorflow, tfx

Similar talks

Shikib Mehri
PhD Student at Carnegie Mellon University
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Rachael Tatman
Senior Developer Advocate at Rasa
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Dominique Boucher
Chief Solutions Architect at Banque Nationale du Canada
+ 1 speaker
Eric Charton
Senior Director, AI Science at National Bank of Canada
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free

Buy this video

Video

Access to the talk “Applying Conversational AI in the Enterprise | Rasa Summit”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
735 conferences
30224 speakers
11293 hours of content