Rise of the Machines: AI Innovations for Service Management and Beyond

May 21, 2019

Ian Aitchison | Director, Product Management | Ivanti

David Martinez | Sr. Product Marketing Manager | Ivanti

We cannot overstate the impact of AI on our lives; machine learning, deep learning, natural language processing, automation, and bots, are set to transform IT and more. All these technologies are starting to make significant contributions to IT, particularly in service management.

Join us to explore the past, present, and future of what we broadly refer to as “artificial intelligence.” We'll consider the impact on our lives as well as human society as a whole. Learn about exciting innovations and dig into what AI will mean for service management in 2019 and beyond.

Transcript:

Dave: Well, I wanna thank you, everyone. Good morning, good afternoon, or good evening, depending where you are when you're listening to the webinar either live or on the recording. I wanna thank you for joining our webinar today from Ivanti. Rise of the Machines, Artificial Intelligence in IT and Beyond. I have one of our esteemed specialist on the subject, Ian Aitchison, he's a noted expert and speaker on this topic of artificial intelligence and what does that mean, in general, and also in IT specifically. So I'm looking forward to hearing what Ian has to say and learning what's going on in this field.

Before we get started, though, a couple of quick housekeeping items. Everybody's on mute right now. WebEx mutes you automatically but sometimes there's some issues. So if you could mute yourself on your side, that'd be most appreciated. Also, we're asking you to submit your questions. So we have a Q&A chat window available in WebEx so you can open that one up in there and just submit your question in there. We'll get to it at the end of the session.

And yes, this session is being recorded and will be distributed later. So look for that. And last thing is go ahead and follow us on Twitter, #ivantiwebinars, so you can see what's coming next from Ivanti, and some news updates, and also some interesting things you may wanna stay abreast of. So I'm remiss, I should introduce myself at the beginning. My name is Dave Martinez. I'm one of the Product Marketing Manager here at Ivanti. And I have the pleasure of working with Ian.

And with that, let me turn it over to Ian. Ian, if you wouldn't mind giving yourself a quick introduction. And then take us through the wild and woolly world AI and tell us about machine learning, heuristics and neural networks and everything else that's going on in the landscape. So thank you for joining us, Ian.

Ian: Thanks, Dave. Yeah, hi, thank you for that intro that was great. Hello, to everybody. Good morning, good afternoon, good evening to you whether attending this live or watching the on-demand recording. As Dave said my name is Ian Aitchison. I work here at Ivanti and I head up our IT service, and assets, and identity, and automation product creation business. So we're the guys that build the roadmaps, and deliver the features and ship the products out to our customers in cloud or on premises.

As you can see on the slide there a little bit about me. I have been in the service management industry for possibly too many years depending on who you ask. Approaching 25 years now working around service management and working towards unified IT. When I'm not doing that I'm spending my time either playing music badly and loudly or falling off boats into water.

And the purpose of today's session is to really share with you some of the interesting changes and evolutions that are happening around the very broad topic of artificial intelligence. Those that

attended our recent Interchange Customer Conference would have seen a session very similar to this that was talking about how our world is being transformed. And new technologies are coming into our personal lives as well as our work lives. And this is a cut-down, but still very focused session based around the same area of how everything is being turned upside down and is changing very, very rapidly.

And what that really means in opportunities we see for ourselves, we all work in organizations and how we can embrace and benefit from these new technologies coming in. So first of all, let's start with what is artificial intelligence? Well, as you can see here, this is my view of it, right? You bring the data in, you can interact with it in different ways, often through sentences and conversations. You can learn from the data that can lead to conclusions or recommendations. And you can drive from that learning activity across technology.

Now that contains within it many, many, many possibilities, and many things we see in our ordinary daily lives now. So there's another way of looking at it, you know, what we call artificial intelligence, this is back actually in 1956, I think. It's just "whatever hasn't been done yet," this quote came out. Oh, yeah, nearly 1965, that was the one "As soon as it works, no one calls it AI." It's only AI when it's not quite here yet, when it's on the edge of being normal. And then we no longer call it AI. I actually think that's changing slightly. I think we're now very comfortable with some things that we would describe as artificial intelligence or certainly machine learning, and there's a very close relationship between those.

So this session, this presentation, there's quite a lot of slides, I'm gonna be going pretty quick as we step through this. Those of you that have attended any IT industry presentation around the topic of artificial intelligence can be fairly sure at some point, you're going to find somebody talking about exponential growth, and talking about the history of artificial intelligence and artificial machine life. And I've boiled it all down into one slide to save us time. You will see these in many other presentations too of course.

There has been a long timeline, which is accelerating fast, you can go all the way back to Hephaestus, I can't even say the name. The maker of automatons for the Greek gods in mythology, the idea that machines could be close to sentient. I mean, that's nothing new and, indeed with Frankenstein, in fiction, we love the idea of bringing life to the inanimate. As we introduce actual real machines, Charles Babbage and his difference engine. Alan Turing and the ability to interact with machines in a way that you cannot distinguish it's not a machine, but it's actually a human, the Turing test.

And then as we go into the '70s and the '80s, we start to find popular culture. Whether it's 2001 of Space Odyssey, or war games, where artificial intelligence takes control over what we all remember war games in the '80s. That really takes control of a massive military situation, which will get very scary, very entertaining. In more recent times, the application of machine learning into game playing, into chess, into go, the delivery of machine learning and artificial intelligence into the palm of our hand, into mobile devices famously beating humans not only at board games but at TV game shows and self-learning.

Now, all of this pretty much condenses what you'll hear in many other AI presentations. So we are on that path but we're gonna stray quite a long way from this as well. We're gonna start by popping a few myths artificial intelligence. These are all not true. There is no single definition. There's no one definitive description, you can go to many different locations, you'll get different definitions.

However, it also is not all hype. It is a myth to say AI is hype. We are every day in our working and personal lives interacting with machines in a learnt or understanding fashion that definitely falls within what some people would refer to as the simple artificial intelligence model. We're not getting close to creating a general artificial intelligence in the AGI despite many scare stories. We're a long way from machines taking over the world.

But equally, we're also a long way from AI solving anything at all. In fact, AI, as we know it now, is a very, very precise and very difficult technical exercise. And I think this is the last myth there, that AI can do anything with any data. We'll talk a little bit about this. But there's a great deal of preparation of very large data sets to enable technology to apparently learn from that data set and use that to reach conclusions that then guide us through everything else.

That said, in the real world, in our lives, you'll find all sorts of examples of AI that are really, really interesting. And there's hundreds of them out there that we're familiar with or not familiar with. If you're not familiar with Talk to Books, Google Talk to Books, just Google it. It's a great place to go, it's really interesting. You can ask questions in natural language and the results are presented back to you that are from an understanding of a massive library of many tens or hundreds of thousands of books. It's a good example of natural language and apparent intelligence, understanding and presenting results back to you.

So for example, if you ask the question, will AI take over the world? It comes back and recommends a book that tells you that AI will work in concert with humans in human-machine teams. Then it also recommends a book, saying that "AI could fall into the wrong hands." And then finally, it recommends a book that says, "AI could take a treacherous turn against its creators after it takes over the world." So Talk to Book, it's great fun. If you ever research any topic, I highly recommend that you have a play with that.

I'm only thinking about our acceptance of enhanced human interaction with machines. Some of you may be familiar with Google Duplex, which was launched... Gosh, when was it? Around about this time last year at Google's big launch conference. They announced the availability of Google Duplex.

I don't have the video with me now but you can look it up on YouTube. And it's great because it's really interesting.

Google Duplex phones out and talks to a human being and the human being doesn't know that it's a machine. And the machine Google Duplex phones a hairdressing salon and makes an appointment for an individual in a series of questions and answers and conversations. And the recording is amazing. Because there's no way you can tell from that conversation that it was a completely artificially generated conversation.

And when it came out, there was lots of negative backlash around how scary it is that machines can phone us up, talk to us and we don't know they're machines. They can gather information, make appointments, answer questions and we don't know that they're machines. And there was that big backlash. And then this report recently, this is just March. The backlash didn't stop it, it's rolled out to 43 U.S. states now and it's gone beyond Android phones to iPhone.

So yeah, the Google Duplex reservation system is a great example of actually what Alan Turing would recognize as potentially passing the Turing test. You don't know it's a machine, you can't tell that you're interacting with a machine.

So what makes all that possible? Well, much of it comes down to good old neural networks. And I do a session sometimes where I get quite deep into what's a neural network is. I'm not a data scientist

I'll just go through a couple of very high-level pieces as we go forward on this session. How does a neural network work? Let's take an example. Let's imagine there are two types of flower that exist. And we're taking photographs of those flowers. And we need a machine to be able to tell us which type of flower it is.

Here we've got a yellow flower and a red flower. And we can look at examples of yellow flowers, and we can see that you know what, they're all typically quite short, and they have quite a low height, they have some length, maybe length typically is three. And the red flower has a height of two and a half and a length of five. That's data we know. That's not something machine has learned here. But if we know that, we can extrapolate it and measure lots of flowers and say, yeah, there's a pattern here.

When we look at an image, if we see petals, and they have a low height and a low length, then they might appear in the cluster that tells us it's a yellow flower. Or if we look and we have what appears to be petals in our image, and they have a higher height and a higher length, they probably appear in the red cluster. So we can conclude they are red. So if you are able to measure the heights and the length from an image, those are two inputs that go into the most basic building block of a neural network.

We feed those two inputs in with a little calculation which is called the sigmoid function, and we'll see that in a second. You're able to say, yeah, merging these two values together, my result comes out near a zero or my result comes out near a one. And you adjust the weighting that's used in that calculation so that consistently when I feed in the size of yellow petals, it tells me this is zero. And when I feed in the size of red petals, it tells me this is a red. Sounds very simple, you feed in some data and we know what the result is gonna be, and we just confirm by adjusting the weighting that the result is what we expect it to be. That's done.

As I mentioned, with a little formula called the sigmoid function, I'm not a mathematician, so we're not gonna spend time on that one, but that's how you take two values and turn them into a result between zero and one. Now, when you have that very basic take two values, have a consistent and reliable conclusion, and you've got you're weighing right for the data that you feed in, then you combine that. And you combine it and you combine it, and you combine it with many, many, many, many other pairs of inputs that are fed from the output. The previous pairs that are all looking for and measuring data across many, many factors. We've used just the most basic to help explain this in this case.

And you feed that structure with tons and tons and tons of data, knowing what the end result would be. And you adjust the weightings until the end result is consistent for the data that you're feeding in. And then you can feed in data that you don't know the results of, and be fairly confident that the results is going to be an accurate conclusion because you've proven it time and time again with very large data sets.

And that in a very, very simple fashion is how...very simplifies how a neural network works. Now there's probably data scientists on this session that are holding the head in their hands and saying "It's really much more complicated than that." And we know it is. But hopefully, that gives us an idea that it's about data. And it's about training and some machine learning, is about feeding that back into adversarial neural networks that feed into each other and correct each other and some is manually maintained, then it's about the data.

Recently, IBM, recognizing the importance of the data introduced datasets of over 100 million human facial images that was provided to enable artificial intelligence, to enable machine learnt conclusions to accurately and consistently be able to recognize human faces. And this is shared fully openly, fully released large and diverse data set called diversity in faces for facial recognition technology.

Because if you don't have that, you introduce a little element of risk. Recently, Amazon's machine learning experts had to turn off an AI-based recruiting tool after they discovered that it was preferring to recruit men over women. Because the training data that it has been fed was not a broad and diverse enough one to enable it to engage correctly with a broad and diverse human species. So it is only as good as the data you feed into this.

And it's also the data that comes out. You may be familiar with the term deepfakes. Deepfakes is using neural networks almost in reverse to generate an image based on a number of criteria, so you almost turn the model upside down. Usually, it's images, recently, the idea of deepfakes, the text came out where you artificially are generating text. And this is another one that was turned off. This is a great story.

There was an AI text generation system called GPT-2, which was created to generate example press articles and blog articles. And it generated such good thoroughly artificial stories that the research individual is working on it, switched it off and didn't distribute the code under the usual open source general sharing rules. Because they were concerned that in the wrong hands, that a large number of fake news stories could be very, very easily generated at volume.

So it gets a little bit scary. And we take that idea of fake news and we talked a little bit about fake

visuals, fake images. That's where deepfakes and that concept increasingly comes in. Do you know now there is a female robot anchor presenting news in China? That's totally artificial. The story she reads out is presented into a machine and the machine generates the image and generates the audio reading out that story. Now, we don't know, at this point, whether an AI is generating the story as well. But that would be an interesting situation.

So we're still in that area of complex sets of data. And if you're in an IT or a service organization, you may be thinking, well, how does that relate to the job I do? Images are often used when we talk about AI because images are lots of data. Every image is lots of data. And you can use an image set of data to come to a conclusion, or indeed, you can generate an image set of data by starting with a requirement. And we'll see in a minute how that translates into a business model.

But it's interesting that that area of generating reality, artificial reality, through reversing and

reverse positioning AI machine learning model is really interesting. All of the above, by the way, were artificially generated, none of the above images ever was a real photographed piece. They're all using a thing called a GAN, Generational Adversarial Network, which is pairs of neural networks feeding back into each other. And usually, they can come up with some good results. Sometimes they come up with some crazy ones like we see on the right-hand side of the screen, they're generative adversarial networks, they're real.

And then you get to things like this, and you start to say, "Well, who are these people?" And you have to ask the question, Well, you know, are these people real? Well no, this person doesn't exist. If you have a few minutes after this session, why not Google thispersondoesnotexist.com. It's great. You can go there and every time you refresh the page, a completely artificial image of an individual that never existed as a human being is built and presented to you.

But it's expensive, building some of those images, some of the first ones I showed you there, it takes…at first it was taking between 24 and 48 hours to generate the first images, it was using a considerable number of kilowatt hours. The electricity the average American household uses in just under six months to produce the results of an experiment initially, which created a bit of a backlash, okay, it's possible, but it's expensive, roughly enough energy to power Cleveland for the afternoon, possibly a joke, but it's changing.

So here's the interesting thing, that ability to recognize and to communicate is becoming easier and easier, and easier, because increasingly that power that's not paid for and held in the hands of one individual it's part of what we think of as the cloud. So imagine if you could take very simple components. Imagine if you had a Raspberry Pi and a loudspeaker, and a video camera, and you put them together, but you connected them to image recognition, into audio recognition, into speaking technology. There's a great piece called...a little project out there called "What is THAT" put together by a research student. It looks like a cardboard box. That cardboard box has in it a camera, and it has a loudspeaker. It has a processing unit, and it's connected by Wi-Fi to the relevant machine services in the cloud.

Let's see what it does. So I've got a little recording here, and this is just showing what's detected by the cardboard box. And there's some audio that comes with this as well. And we'll just see how we interact with this.

Woman 1: What is that?

Machine voice: That's a mother duck.

Woman 1: What is that?

Machine voice: That's a Boston Terrier.

Woman 1: Who's logo is that?

Machine voice: That's the Starbucks logo.

Woman 1: What does that say?

Machine voice: That says, "Not everything that can be counted counts, and not everything that counts can be counted." That's a quote by Albert Einstein.

Woman 1: Great, thank you.

Machine voice: My pleasure.

Ian: So that shows just how increasingly easy it is to bring some quite extraordinary capabilities into the common space where we live and where we work. And some an interesting diversion from that if you take may be a step or two further forward. I mean, if you do have equipment, which is able to look around and make decisions and understand what it sees and communicates, then we start moving down the whole topic of autonomous weapon systems. Now, that's not the purpose of this session here, this webinar, so I won't be going down that path.

But there's an interesting evolution down there, it's getting easier and easier and easier to sense, to detect, to decide, and to take action. But what I'm gonna dig into now is understanding some of those kind of light-hearted and playful pieces. Where are we in terms of our industry, the world in which we work? My background, particularly service management, often very focused around IT. How are these changes impacting into us. So we'll start with a good few little survey results and quotes from commentators and analysts in our industry.

Forbes did a survey recently 80% of U.S. executives recognize their organization needs to begin mastering the art of human-machine collaboration. Eighty percent said placing powerful AI field applications in the hands of knowledge workers is critical to productivity and performance. Forbes study highlighted that AI is not replacing people. It's enabling collaboration, quality of life, better-focused value-added work, more frequent communication. It's not replacing employees, it's enhancing productivity and making people more effective.

Because what is the sexiest job in the 21st Century? And credit to a colleague of mine that put these slides sites together for an internal Ivanti machine learning briefing a while back, and I'm reusing some of these because they're so good. The sexiest job of the 21st Century, data scientist. What is a data scientist? Well, they're better at statistics than a software engineer, and they're better at software engineering than a statistician. But they do need to know about numbers and large number sets and statistical modeling. And they need to know about code. They need to know about the latest scripts, models, and TensorFlow, and all of these great things that enable us to build neural networks and deliver AI capabilities of the back of large data sets.

So of course, we're Ivanti, we're very focused on delivering the value of unified IT to our customers all over the world and already we're making strides. Ad we know others in our industry are doing the same as well, we're seeing it all over the place. We're making strides down that path of artificial

intelligence.

So let's talk a bit about that. Some of our, let's say, assumptions and principles around this is just like Alexa on your desktop, just like Siri on your phone, indeed, just like that cardboard box. The real power of AI comes when you use the cloud. When you're able to have not only very large data sets, often from multiple sources, with very strong processing power, but also very importantly, backed up by people. Backed up by people that are tuning, and improving, and watching how the AI capabilities work and making them better. The cloud makes all of this possible. The cloud delivers recommendations from GPS apps, like the waves into the palm of your hand or Siri, or all of these great things that we see increasingly, that come from the cloud.

Now we launched Ivanti clouds not that long ago. Some of you may have seen some of our announcements around Ivanti cloud. Our new cloud platform which delivers a range of capabilities to enable unified IT. I'm not gonna dig into this in great detail. But it's an interesting example of whereby feeding data up into a shared platform, we're able to understand very large patterns of data, draw conclusions, make recommendations, interact, and automate. And this is possible through a combination of AI concepts and cloud technologies.

Going back to the abstract slightly, what does it all really mean in a work context? And I think we can boil it down primarily into the fact that you have conversations and you have data. Increasing your interaction with machines is becoming conversational. It's becoming sentence based. We prefer to use sentences, they might be spoken, they might be typed. Typing is fine, bots are fine. Sometimes we prefer to speak. And we have lots of data.

And those conversations let us interact with chat bots, and virtual agents, and understand recommendations and lead towards the creation of knowledge. The data helps us augment the ability of an individual to work and make advice to automate. And all of that enhances the experience ultimately for not only the support admin IT admin function, but for the individual in the business, the employee or anybody interacting with an organization.

So the first of those is about conversation. The future, when we look at AI in the workplace, is increasingly conversational. We talked about Ivanti cloud is one of the examples, you can see many other technologies out there now where the points of interface is a question. A question that you type all that you speak. And just like when you ask a question to Alexa, you either get a fantastic answer. Or if you don't get a good answer, that question goes away into the cloud and it tunes the clouds learning, intent recognition, and people are involved and machines are involved. Next time you ask that question, you finally get an answer.

But of course, if it's a shared platform, which all good cloud AI is, it doesn't rely on you asking the question, so you get the right answer next time. All it relies on is anybody anywhere asking that question once, and that will help feed into the next time if it's you asking it, there may be an answer ready, already on that one.

So that's one thing that we're seeing increasingly as an interface into technology. The other of course, in the self-service context we're seeing user bot, and a bot interaction with the end-user for support is becoming increasingly a demanded expected capability that we look for in support and service tools. Not everybody likes bots, not everybody wants to chat with a machine, that's okay. But some people do, that increasingly that some people percentage is growing.

Goodness knows I am sick and tired of phoning up telephone numbers and having to spell my last name out time and time again to somebody who's maybe of a different culture, or a different continent, or a different time zone. I'd like the technology to know who I am and greet me with a "Hello, Ian" and ask me what it can do to help.

There are lots of service-related bot capabilities out there. What we're particularly recommending at the moment is remember Alexa, don't become a bot developer. If you're looking at a bot at the moment, if it requires you to build complex conversation workflow, if you just require design capabilities, then that's not the simplicity of Siri or Alexa or any of those other capabilities. That's the heavyweight that requires you to continually self-maintain the sentences and the utterances that come through those. So do watch out for that.

And then the other part, of course, is data. You remember that diagram this conversation, there was data, and this takes us to data lakes. I don't know if those of you on this call have a data lake, maybe you do. Organizations that implement a data lake outperform similar companies in this particular survey by AWS that was outperformed by 9%. What's a data lake? A data lake takes your production databases, merges all the data, and enables that data to be accessed in an unstructured form through a business intelligence layer, which lets you deliver valuable metrics and reporting.

But very importantly, it also lets you learn from that very large data set. Remember the neural networking model? Remember the feeding data, and then being able to process data and draw conclusions. Data lakes make machine learning possible.

So let's look at some examples of AI. I'm gonna run through, I think there's about eight of these. And the first case we'll talk about is a generic use. And then we'll focus into an IT or a service example of how this could be used as we see the industry, and the technology evolving.

So the first example of concepts in machine learning, and here we're in the area of classification particularly. You take large sets of data and for example, you need to know from a large set of data whether there are differences between those data sets. And you're able to perform a hypothesis and say yes, there is a conclusion from that. Now, that's something that... This is kind of a slightly light-hearted first example maybe it's something that vendors such as Ivanti can use, it's a great benefit.

Where we can look at large datasets and very quickly ask the question, for example, do different industries get greater value from the solutions we provide? Are there some industries where our solutions don't add value? Do we want to do something about that? And there are some industries where we are fantastically doing well. And that's something that comes from being able to question large data sets.

We mentioned classification, what kind of a thing is this, we've all got spam filters. There's a great example of machine-driven automatic classification. An email comes in and often you never see it because thankfully the spam filters are really very clever at recognizing the phrasing, the combinations of words, the meaning in emails that are deliberately constructed to try and bypass

them.

So why stop at something as simple as an email trying to sell me something I don't need, why not turn that around and say an internal communication? Is that telling me that individual is productive? Are they happy? Are they getting the service or the support they need? Can they use the technology they need? If I'm gonna go and visit them, do they seem to be happy or am I gonna have a really hard conversation?

Here's another example, ranking. And this is about something you'll be familiar with. And actually, there's Roger's name, he did some of these slides, so credit to Roger. When you Google something, you don't just want 10 million results with the same words in, you need something which indicates these are the best matches for what we think you're looking for.

Now, those of us in service management particularly are very familiar with searching and knowledge management, and large complex descriptions of the right thing to do. And what better context for these sorts of capabilities than delivering knowledge searching, which is AI-powered? And there's some understanding of what you're trying to look for. And talking about what you might be trying to look for, what about recommendations based on what you've been doing, what will you want to watch next on your favorite streaming service.

So here are some top picks, well, if I was in IT, and I needed to provide support or service or if I was an employee, a business consumer of service, and I needed somebody to help me or assist me, wouldn't it be great if technology could tell me, "Yes, you need some help with Excel, why don't you contact Ian, he's an expert at that. He's only three desks along from you."

Then you get to numerical calculations and trending. This is about, for example, predicted house pricing over time. And let's turn that into an IT workplace context with Ivanti. What if we were able to predict user satisfaction, or productivity, or future service availability? What if we're able to look at large data sets and identify there is a problem that exists that needs something done to stop future incidents occurring in a service management context there?

Let's look at the patterns. Let's learn from those patterns. Let's look at the patterns and look at the clustering. We're all very familiar with while you bought a pair of shoes, other people that bought those shoes also bought these things, there's a common model. And that's, again, just like all the other examples, that's looking at data sets, drawing conclusions, providing recommendations or advice. The clustering of data translates perfectly into, well, you know what, let's group together all the people that might be impacted by this issue. Let's communicate with everybody that uses Microsoft teams, and let them know that we've identified an issue that currently we're looking into what the problem is.

And then there's anomaly detection, what are the odd things? Why is there suddenly a spike? Where has something changed? What we see is, you know, a spike of trends, for example, we're often seeing these coming through. And again, many of these we recognize in consumer, personal world. And if we look in the business context, we move maybe beyond service management into security and being able to recognize there are repeated trends that weren't there before.

Why suddenly are we getting all sorts of activity through a particular port? Why suddenly are we seeing a large number of failed logins happening in a particular location, or from a particular desk or from a particular external IP address? And just alerting based on simple rules these days isn't sufficient when you're dealing with very large data sets.

So many of these concepts come together as we look at ideas like recommendation, we look at ideas like searching. Finding what you need, and being recommended to take the right or the correct action. We're doing a lot of talk at the moment, a lot of work around augmenting the work of a member of service and support teams to make the right choices, to understand the sentiment, the happiness of an individual, to understand who best to pass work to. To understand, for example, what are some other examples of this type of work that have been successfully resolved in the past. What knowledge articles have been proven to be successful? What are the sort of actions that other people take with this?

Wow, right, how are we doing? I mentioned actions. Now a big part of the AI diagram that I showed you earlier, conversations and data, if you went through that diagram, you'll have seen automation. And automation is a key part of the AI vision within IT and within a business context. Because the ability not only to recommend but to then act is essential. As we are able to start seeing the time it takes to do things go from hours, or days, or even weeks, now to minutes or seconds. When it doesn't have to wait for a human being to do it.

A very typical example we're all often these days very familiar with is going to a request catalog in a self-service interface and asking for something which is delivered automatically to the individual that asked for it. And in IT context, something like access to a cloud service or a new logo, or maybe some software that automatic delivery might, in the past, have taken weeks, it can now take minutes, possibly even seconds. Automation is a key part of taking recommendations that come from analyzing large data sets to taking action with caution.

It is important also, as you implement automation, as you plug AI recommendation into automation, just be careful that you don't accidentally let decisions be made that apply on a large scale and can impact the business in an unexpected way. So under the concept called human in the loop. Human in the loop is one where are deliberately designed into your connection between AI and automation or machine learning and automation. You build in something that says, "Yeah, this one's unusual, we're not gonna automate this, we're gonna put this in front of a human."

It's a kind of defined concept and it sounds very simple. But it's very strongly recommended as people are increasingly automating. And then not only automating action but automation decision making, too.

So what should we do? This wave of technology is coming into our workplace, in many cases, it's already there. What's the right thing to do? Is this a tidal wave coming in? It's gonna wash us all away? Or is this an opportunity for us to surf and go faster, and get to places much quicker than we wouldn't otherwise have been able to do?

Well, extracting a few quotes from other organizations here. Here are some great quotes from a Gartner document, actually a couple of years ago, but still very valid. They were saying then looking into the future, "Many AI projects in industry context will fail, but those that don't start on projects will be left behind soon. And those initial failures may lead to people being doubtful saying, 'These technologies, this won't work. Why are we doing this?'"

We introduced a new sentient super bots, and all it did was it unlocks and locked the rear door 300 times a second and didn't do anything else. Therefore, the whole idea is crazy. Well, no, because if you stop there, you will be left behind by organizations that are trying the next project, and the next one, because success comes incrementally. And bear in mind this advice is back in 2017. So this is talking to organizations two years ago, saying, guys get started, get started now and make mistakes because you will find you lose control, and you will be left behind while other organizations that take the first move they are ones that benefit the most from this.

Start planning is another recommendation a year later, actually, nine months ago. Start planning intelligent virtual support. We talked about bots powered by automation. Minimize development, try and avoid the builder bot concept. Try and avoid writing your own AI. Even the largest vendors in specialized spaces such as unified IT and service management will use the best, most proven, most established backend technologies to power components of what we do. We don't build everything ourselves. And you should take that same approach in your organizations, take best to breed from vendors that do the best possible things, don't develop from scratch.

And adding one here for myself, automate the heck out of everything. The ability to automate actions that people would otherwise do is what makes a transformational difference to the productivity of an individual and the success of an organization no matter what it's focused on. Putting automation in use cases that can be identified time and time again, is really valuable. But don't wait, except that some of those may not be as you need. Every time you have an experiment, every time you learn, you're making it easier to be successful the next time. It's a long game.

Now when people say, "Hang on a minute, I'm scared, you're introducing machine learning and AI into my business, and this is gonna be bad for me." And people then say, "No, it's not because it's gonna take all of that repetitive work away. It's gonna take all the repetitive work... We can automate those simple password reset requests. We can automate that delivery of software and services to an end user, we can automate things that's great."

So all of that easy, repetitive work goes away. So great engagements, then they say "Well, what's left when you take easy away?" What's left is hard work. There's two ways of looking at this. Do we want to do jobs that are all just hard work so we don't have the joy of an easy job to do? I'm gonna do that tough one after lunch, I'm gonna do three of the easy ones. Oh no, there are no easy ones for me to do, I'm gonna have to do the tough one, and then the tough one, and then the tough one.

There's another view on this which is...and there's a great analogy for this. We don't wanna be serving Big Mac and fries in the jobs we do. We wanna be creating filet mignon with different [inaudible 00:45:29] potatoes. We wanna be doing fulfilling work and sometimes yes, that is hard work. So there's kind of two ways of viewing that. So one thing that is clear as new technologies come into the workplace, those of us in service and in IT, particularly, unified IT context, we more than anybody else we will be okay. We will be more than okay because we live in a world of change. Fundamental to service management is change management and principles of change are fundamental to IT, is the fact that operating systems change, and technology changes, and viruses change, and ransomware changes, and software license models, everything keeps changing.

And this survey done a little while back would say, well, one of the things that are identified as most valuable to be successful is a future where those machine learning capabilities come into the workplace and start to change the jobs we do. And you know, no surprise technical skills isn't the most valuable. Increasingly, we look to vendors like Ivanti to provide the technical capability. What matters most is an ability to change. And that's something that those of us in IT, in a service model are very comfortable with because that's what we do. And we solve problems, and we think creatively and with fundamentally about communication and collaboration. So these are all skills we should be very comfortable with as increasingly changing workplace heads towards us.

So yes, things will change dramatically. Some will take a little longer, some will take considerably less time. Those of you who watch the market will know Ivanti and other vendors, we're all focusing hard on what these new technologies can bring into the workplace from outside of work context. You know, there's some scary things out there, too. We look at autonomous weapons systems and all sorts of possibilities where machines might do things other than we wanted with the deepfakes and the fake news. But very much of it will be for the good. Very much of this new technology will be enhancing our personal lives, it already is. Is enhancing our abilities to detect medical conditions, it's advancing our abilities to get help faster, to locate people, to educate people better.

But all of that comes down to individuals that work in organizations and do jobs increasingly around working with IT and unifying IT capabilities into that next level. That means if you work in an IT role, a service management role, you're part of that future. So we would urge everybody to be putting AI project actively onto their plans for 2019 and 2020. Not everyone will be super successful, but unless you start, you'll find it harder and harder and harder to take advantage of that.

And there we go. That takes us through an interesting view, maybe on the AI landscape. I hope that was interesting, and hopefully sort of inspired a few thoughts and ideas. What I might do now is say, hey, Dave, are you there on the line?

Dave: Ian, I am. And thank you for that session. And I think several attendees were inspired. We have a few questions coming in. But before I start with them, let me just thank you again for the session. And I learned definitely quite a few things, and the image of AI in a box really stick to my mind right now. So let me get the questions, but remind people, we still have time, so submit your questions in the Q&A window, we'll get to them.

Ian, the first questions. Let me just combine them together, we have two questions coming in. You're telling people not to be a bot developer, so I thought that was a good question coming in, that said, okay, are there APIs or standards available for AI in these technologies?

Ian: Yeah, I mean, a lot of the time... The thing about a bot is it's not the bot itself, it's what the bot interacts with. You think about that automation topic. It's not the ability to chat with a machine, that's no big deal, so what? It's what the machine then does. It's what it leads to and that's about the bot interacting with other technologies. Now, increasingly, those are workflow, my background service management. So a lot of our bot conversations are all around service management. And the ability to connect to a service management tool that has workflow that does many, many, many clever things is where a bot goes from being a pointless chat experience to really becoming a virtual assistant.

So there are standards around APIs. Absolutely. There are standards around interacting with tools. I mean, the rest model for APIs is very common. But also look for vendors in that area that offer bot solutions that are pre-integrated into their tools as well. We have a bot that's directly connected into our service management workflows. So there's no concern about writing custom API scripting. And increasingly, we're seeing people moving away from scripting because you shouldn't need to do that sort of thing anymore. I hope that answers that question

Dave: Okay, that's good. Thank you. That relates to another question, let me get into it. So you mentioned service management, where else is Ivanti using AI? Is it only in Ivanti Cloud? And a related question, is the hub only available in the Ivanti Cloud?

Ian: Oh, good questions. Excellent. So we're using AI in many areas, some of which I described in that session. Augmenting the ability of a support analyst work. But we're also using AI in our analysis of data inside the Ivanti Cloud to enable what are called Smart advisors, recommendation based model across large data sets. To identify, for example, where we've detected a number of endpoint computers that are running old operating systems are out of warranty. And you know what, it's not worth trying to fix this, it will be easier for you to just replace this.

I think it's called the re-image or replace advisor, which looks the data sets. So you know, don't waste time working on this, looking at all of its life, and its age, and its situation. It's just gonna be cheaper for you to simply swap it out. Those sorts of conclusions come further. So yeah, all the way across that into security, we're doing a lot of work around security patterns and recommendation. And yes, it's almost exclusively a part of Ivanti Cloud or Ivanti Cloud capabilities. Sorry the last part of that question, Dave, something about the hub?

Dave: Yeah, this person thinks that the hub is part of the about the Ivanti Cloud. Do you wanna just clarify that?

Ian: Yeah. Possibly an Ivanti customer. Yeah, the hub is part of Ivanti Service Manager in the cloud,

which is subtly different to the new Ivanti Cloud set of combined unified IT technologies, but tightly integrated. So if you're a service manager cloud customer, you have the hub that does integrate with Ivanti Cloud, but it is regarded as a part of the service manager feature set.

Dave: Okay, great. That's a good lead into this other question. Let me just jump to it. I think this is an Ivanti customer asking, how do I get access to the Ivanti Cloud? Is there a trial available that I can use or access?

Ian: Yeah, I believe there certainly is. I think if you go to Ivanti's websites, you will find a lot of information about Ivanti Cloud up on there, including a click here for more information. Speak to your Ivanti rep, they will be very keen to talk to you about Ivanti Cloud. If you attended any of our conferences, our customer conferences recently, they were also given direct access into trying the Ivanti Cloud. So that should be really easy for you. Please let us know if there's any challenges there because you should be able to get close pretty quick.

Dave: Okay, great. Here's a follow-up question that somebody's asking, a little bit what's underneath the hood, what AI technology is Ivanti using?

Ian: No, I'm not gonna give specific technology names partly because I'm not a technical expert, and I'll probably get some of the names wrong. What I would say is we deliberately and with forethought, use a combination of technologies, those that have proven best of breed in the data science and machine learning side of things with a mind that, you know what, if we need to, we would swap one for another as the market and the technology evolves. So we have components, some of our natural languages, Microsoft LUIS capabilities, but we also have some natural language, which is working with some Google capabilities.

And that's a deliberate approach to this. It's a rapidly evolving area, it's kind of the same as don't build your own on this. We're deliberately ensuring that we always use best to breed as we go forward and the market evolves.

Dave: Great. Somebody asked kind of follow up question on that, I think you answered it. The markets going at rapid pace, how do you plan to keep up? And how do you plan to be unique or differentiating? So I think you touched on that, do you want to just elaborate a little bit more?

Ian: Yeah, I mean, the unique or differentiated ultimately, I think you need to reference unified IT and Ivanti strategy around how we see the future as a seamless blending of service management and assets, and security, and endpoint management, and even identity governance as well. Rather than separate integrated technologies that just pass data backwards and forwards, we see a growing future, which is AI-driven, and presents from the cloud a different way to approach these which is really transformational, both in the job you do and the way your business operates.

Within that, of course, there are specific use cases around the application of AI. One of the most exciting at the moment is, I've passed over pretty quick, but the conversational interface where you can question using a sentence, the life state of connected technology in real time, it's called Ivanti Real-time. And you can ask a question like, show me the state of devices without this, or show me devices without their firewall enabled. And live, it talks to a sensor running globally on any device as part of the defined scope. And in real-time comes back and shows you that information.

And that's an example of the move to both conversational and also there's no data stored as we do that. There's no database holding discovered data values. It's using the Internet of Things MQTT, MQQT protocol, whichever of those two, to live query two devices and to get immediate responses back in response to a question. That's one of the most innovative pieces at the moment, but there's a lot more coming down the line. Okay, Dave, I've lost you, are you there? Okay. Right at the end of the session, I seem to have lost Dave Martinez. I don't know if you're hearing me [crosstalk 00:57:43].

Woman 2: I'm still hearing you, Ian.

Ian: Hey, you're still hearing me. Well, I think we were just about there actually. I think we're just at conclusion point. So Dave fell off his chair just at the right minute. So I guess we'll probably wrap up there. Were there any closing comments from you guys before we end the session?

Woman 2: I think we're all good. Thank you so much, Ian.

Ian: Okay, thank you very much. Thanks to everybody that attended. I really appreciate your time. Have a great rest of the day wherever you are in the world. And I look forward to speaking with you again sometime soon. Thanks a lot.

Woman 2: All right. Bye-bye.

Ian: Bye-bye.