EPISODE #10: Key Findings From Researching the AI Market & How They Impact IT

Original article can be found here (source): Artificial Intelligence on Medium

Guy Nadivi: Welcome everyone! My name is Guy Nadivi and I’m the host of Intelligent Automation Radio. Our guest on today’s episode is Ritu Jyoti from IDC. For those of you who are unfamiliar, IDC is a global provider of intelligence and advisory services for the IT, Telecom and Consumer Technology markets and they employ over 1100 analysts worldwide to figure out trends and opportunities in over 110 countries. Now, Ritu is a program Vice President at IDC and she recently co-authored a report which caught our attention because it examines the use of artificial intelligence in digital transformations.

So we’ve invited Ritu to come on the show and share with our audience some of the very interesting insights she and her colleagues published in that report. Ritu, welcome to Intelligent Automation Radio.

Ritu Jyoti: Thank you Guy, it’s my utmost pleasure to join the session today.

Guy Nadivi: Ritu, you’ve talked about the speed at which artificial intelligence functions and that it’s faster than the blink of an eye. And you actually point out that the time it typically takes to blink your eye, on average, is 300 milliseconds. While the time it takes for AI to detect fraudulent credit card activity is only 40 to 60 milliseconds. I think this might be a good example of what you’ve described as AI’s ability to “accelerate time to insight.”

Can you please elaborate about accelerating time to insight and why that’s something that should make IT executives and business leaders want to deploy AI?

Ritu Jyoti: Yeah, absolutely. It’s a fun trivia in terms of the difference between 300 milliseconds of the blink of an eye and a 40 to 60. I just enjoy that data point. So, it’s basically a guide of a digital transformation.

Digital destruction is real. If you look at the average company life span on the S&P 500 index, in the third platform which we characterize by cloud, social, mobile, and digital technologies, starting in 2005 is 18 years. But when you compare it to the first platform which is characterized by the mainframe, the average life span was 61 years. So, you can see the contrast and the differences there.

IDC has done the research and it shows that organizations across every industry are under threat, and the average percentage of traditional revenues that are at the risk of destruction and digital transformation it varies from about 11% for hospitality to about 29% for utilities, and as we all know data underpins digital transformation and people are using the digital transformation to balance the business objectives between tactical and strategic objectives, whether it is improvement in operational efficiencies, reducing risk & penalties, increasing existing product revenue, and if you look into all of this, timely access to insights is crucial to enable all these business objectives.

In today’s world, if you see, data is being increasingly distributed. It’s stored at the edge. It’s on- premises. It’s on the cloud. It’s very dynamic in nature and also diverse. Gone are the days when we just had structured data. Now we have a lot of unstructured content. We have semi-structured data, and with such huge volumes and distributed data sets, it’s humanly impossible for us to go through files of data to gain timely insight. In leveraging AI, one can address the time to insights need.

Also, data quality is important for AI algorithms’ trustworthiness, and AI algorithms are being used to dynamically create data validation checks and improve the quality. I’ll give you a very small example, we [that was] just recently — Microsoft Azure capability was announced. If you see there are 800 million people who use mobile applications today, and they have an increasing amount of reliance on the mobile banking app. And with this rate of mobile banking adoption, mobile device fraud has inevitably increased as well. To detect the fraudulent transactions have become all the more important, and just two weeks back Microsoft developed a mobile banking fraud detection architecture, which actually uses artificial intelligence to spot fraudulent transactions. And it is using a combination of Azure Cloud Services. So, it’s a long and very important juncture that we are at, where getting timely insight can help us in meeting all the business objectives.

Guy Nadivi: We live in the age of the algorithm and thanks to sophisticated algorithms, pervasiveness of data, GPUs and accelerators, and of course the infinite scalability of cloud computing, you’ve predicted that by 2019, just a few months away, 40% of digital transformation initiatives will use AI services and that by 2021, 75% of commercial enterprise apps will use AI. That’s a pretty dramatic uptake for this technology. So Ritu what should CIO’s, CTO’s, and other IT executives do right now to start preparing for this landmark shift in the way that IT will be providing its services?

Ritu Jyoti: Yeah that’s a great question, Guy. We get this question from everyone. So, our advice to the leading executives, CIO’s, CTO’s, and all is to have — establish a change management organization — to proactively address any human concerns and retaliations. [Retaliations] is a strong word, but real. To embrace AI-infused infrastructure or technology to help them gain and meet the business objectives.

Look for ways for retooling their IT staff to support those strategic initiatives and also fill in the skills gap. They could look for augmenting their in-house staff with external consultants who have already done this, and have some experience. They should look into, you know, a lot of public cloud service providers. I just gave the example of Microsoft Azure. They help us to jump start their adoption of AI. They make it very easy to use with some pre-installed, pre-configured templates so, you know, I would suggest that they should start their journey by embracing public cloud services, and then look for evolving an approach to adopt AI technology and AI-enabled infrastructure.

As a first step, they could use it for predictive analytics and once they have gained some valuable insight, then they can slowly save in [integrate] the automations once the trustworthiness and the quality of data is established, and the comfort level is also improved.

Guy Nadivi: At IDC, you conducted a survey and found that IT automation is the top use case for AI and machine learning with 65% of your respondents saying they currently use AI & Machine Learning for IT automation. Ritu, what do you think makes AI and machine learning so compelling for IT automation as opposed to some of the other use cases like workforce management, or CRM, or supply chain and logistics?

Ritu Jyoti: Yeah, you know, I mean in fact we were a little bit surprised. We were thinking that it’s mainly used in the front office functions. But when you kind of sit through and think through as to why this is so important — because remember I just talked about how the digital transformation’s business objective is to improve operational efficiencies. If you think about, you know that there’s significant amount of operational inefficiencies in the IT organizations right now. So when we looked into — in addition to that the skills gap, there are folks trying to figure out how to use the AI algorithms and kind of bring efficiency in delivering the need, or the IT needs to support the business objectives.

So, you know, when we kind of talked through, it’s basically because if you start from the back office functions, there is already a lot of computer to computer interactions in the IT and finance and accounting in the back office function. Especially there the IT automation can help. So there is much more easier [to get the] buy-in. I’m not saying that the front office functions like CRM, supply chain, and all cannot benefit, but the reason folks are starting from the back office is because it’s easier. The return on operational efficiency is crucial to make even the front office functions succeed and again value all the AI algorithms.

Guy Nadivi: In that same survey that IDC conducted, you found that improving operational efficiencies was the number one business objective for AI & machine learning projects. Improving operational efficiencies was even more important to your survey respondents than increasing existing product revenue, acquiring new customers, or reducing costs. What do you think makes artificial intelligence and machine learning so compelling as a tool for improving operational efficiencies?

Ritu Jyoti: Yeah, as I just answered in the previous question there’s a close tie up in the two situations, and if you look into it when they ask them what are your big challenges in adopting AI, skills shortage was rated to be one of the top challenge. Organizations have tight budgets between investing in innovation versus investing in OpEx spend. Any routine repetitive task can easily be automated. You know if you think about organizations are spending significant amounts just to keep the lights on. Let’s take the example of the infrastructure. There’s a lot of telemetry data, there’s logs, there’s stats available, where ML algorithms can easily be used. For example, IT organizations can reduce or eliminate call centers by using ML to process and integrate incoming help desk calls routed to the right person for problem resolution within a single call. They can, you know, look for using machine learning algorithms to optimize software deployment strategies and reduce the failure rate.

So these are very simple examples that not a lot of extra overhead or intelligence that is needed and manually it’s impossible for folks to kind of go through heaps of log files and predict server failures or outages. So it becomes very, very simple, and there’s not a lot of dispute about the guidance or the results that they get from the ML algorithms usage. It’s a simple starter, easy to use, decent trustworthiness here unlike the other examples that we talked about where there could be some ethical biases concerns and stuff. Improvement in operational efficiency is the number one place where the ML algorithms can be used and the IT automation — very, very direct correlation with IT automation being the number one case and with the business objective being operational efficiency and how the two tie together to support that objective.

Guy Nadivi: Ritu do you think there’s any specific industries that stand to benefit the most from deploying artificial intelligence & machine learning for IT automation?

Ritu Jyoti: Actually I think it benefits all the industries, but if you have to think, you know, a few examples. Retail and health care jump a little bit because you know these are the places where a lot of edge processing will be key. For example, in the case of retail at the point of sale areas, AI could be used to do automated out-of-stock detection on the aisles. In the health care, AI algorithms could be used to route the inquiry to the right physician or there could be virtual radiologist checks done, and in all these edge locations, you don’t have the IT staff. You don’t have that much of bandwidth or IT hand holding sort of thing, and in those cases, you switch to AI algorithms to scale to support the need is all the more highlighted, but at the meta level it can help across all different industries.

Guy Nadivi: IDC interacts with thousands of organizations around the world. Can you share with us an example organization or two that made a successful digital transformation thanks to artificial intelligence?

Ritu Jyoti: Yeah, I mean if I have to take an example — payment services you know the company PayPal? They are using GPU-accelerated deep learning algorithms for fraud protection. There’s another company consulting firm Accenture’s R&D arm & other businesses, they are using to detect internet security threats. Drive.ai is using the AI algorithms preparing to offer a self-driving car service for public use very soon. It spans different industries and currently you see most of the people are using it in the fraud detection or large scale, web-scale examples, internet security threats, but the key examples are growing rapidly.

Guy Nadivi: If I’m a CIO, CTO, or other IT executive, what are the critical KPI’s or success metrics I should focus on at different stages of the digital transformation journey?

Ritu Jyoti: Yeah that’s again a very good question, Guy. This is something that we think in and out in all our interactions with the C-level executives. You know I think transformation is integral. Without the IT transformation, without the improvement in operational efficiency in IT automation, there’s no way that the digital transformation can succeed. So, when we think about in the past, most of the IT organizations just focused on total cost of ownership, or cost metrics, or ROI investment. But today people are looking into a little bit more business specific-KPI’s. They’re looking into — to make the judgment call that how much of IT spend was on newer business initiatives, how much of that is used to grow the business, transform the business. They are using the metrics to see present stage of their spend that is spent on customer-facing initiatives. They are tracking their customer satisfaction scores for business-facing initiatives, and it’s all about looking from a business eye than the traditional IT efficiency metric. There’s a huge transformation happening in terms of the KPIs.

One very interesting thing that we see more and more executives doing these days is that there’s a lot of projects that are funded by IT and there’s more direct linkage of how these IT projects line up with the business objectives. Now there is a concept of program governance where every IT initiative is linked to a business objective and there are tools which are supporting this tie up. So that’s another very upcoming metric that a lot of customers are trying to tap into.

Guy Nadivi: What do you think are some of the biggest misconceptions in IT departments about artificial intelligence and machine learning?

Ritu Jyoti: Yeah this is interesting because when we talk to some folks they think, you know, I mean it’s become like a buzz word right. Everybody thinks that people use this term and for everything very loosely. It’s become like AI is a silver bullet and that it will solve everything. So, there’s some players who actually believe that, and there’s some players who actually are very — the nay-sayers and they think that, it’s too premature, it cannot be used, there are ethical biases, and skill shortage. So, I would say that it being a silver bullet is a big misconception and also that it’s too early and I have the time and the bandwidth to get on to this. The reason I say this is that because the rate at which we kind of say that this is a slow-motion explosion, the way it is kind of evolving every day, the innovations are happening at a rapid pace. So they don’t really have the time. They really have to kind of get on to it and embrace it.

Guy Nadivi: You stated that the top three challenges for adopting intelligent infrastructure are data volume and quality, advanced data management, and the skills gap. Can you please elaborate a bit on how leading organizations are addressing these challenges today?

Ritu Jyoti: Yeah sure. So, poor data quality has a direct correlation to bias and inaccurate model build-out and ensuring data quality with large volumes of time makes the most distributed data set. It’s very difficult, and it is hard for developers to know, correct, and accurately check for validation. You know in the past people just had structured data sets. They had it within their own boundaries and within their data centers. The diversity of the data wasn’t there and there was not a lot of unknown factors. Today, the data is distributed, it’s dynamic, it’s varied sources, internal and external, and it’s humanly impossible for coders to develop all the checks and validations. And to address these challenges, enterprises desire an autonomous data quality and validation solution. So such a solution could automatically learn data’s expected behavior, create thousands of data validation checks on the fly without the need for coding. You can update and maintain the checks over time and eliminate both anticipated and unanticipated data quality errors so that you can make the data more trustable and usable.

So, when we talk about that enterprises have — what are their top three challenges, data volume & quality is important, but also skills gap. There are, talents meaning both AI engineers and data scientists. They’ll be needed to support the growing segment of AI-dependent digital transformation initiatives. But when you think about it, the IT organizations are still grappling with a shortage of these professionals.

For example, it’s not just also having enough data scientists or data engineers, but even the data scientist they haven’t done something like this. Leave aside that they are fewer in number, they haven’t done something like this and they have a high learning curve in building, and optimizing, and training model skillsets a lot of data scientists don’t have. They really need to improve their productivity and some of these, if you look into the life cycle of building-out of an AI model, a training stage is very iterative. It can take millions of hours, days, weeks — and the toolset supporting that requirement is not there. There are some organizations who are helping out with that.

So, you know, how the training stage is happening, and it’s still in the middle of the process, you cannot sit & wait for the entire process to complete over 36 hours and then revisit it. People need to have an instant support of something has failed and we can stop there and restart it. So those tools are evolving and essentially and I’m talking about if you notice, AI itself is coming to the rescue of solving these challenges that they are having. It’s an interdependent kind of relationship. They are leading to AI problems because they don’t have the skillset, or the data is distributed and dynamic, but then again, they have to use AI to solve those problems and it becomes a very, very integrated problem that folks are — there’s no way. They have to embrace AI and continue to use it to solve the challenges and escape.

Guy Nadivi: Ritu if you can offer one piece of advice to CIO’s, CTO’s, and other IT executives that are thinking about taking the plunge with AI for IT automation, what would it be?

Ritu Jyoti: Embrace it. It’s not optional. As I was just saying at the start of the conversation that there are companies who are thinking in the IT industry that it is a magic bullet or that we have to stop and wait. Companies that are not embracing strategic innovations with artificial intelligence and analytics, they’ll fall behind and they’ll lose their competitive advantage. So embrace it, set up your change management organization, speak to your people. Augment it with external help but, embrace it fully and run with it.

Guy Nadivi: Prudent advice. Alright, looks like that’s all the time we have for on this episode of Intelligent Automation Radio. Ritu thank you so much for joining us today and providing some unique insights about the role of AI in IT automation. It’s been a real pleasure having you as our guest.

Ritu Jyoti: Same here, likewise. Thank you Guy such a pleasure to share the insights that we gather from the broad industry and look forward to future collaboration. Thank you.

Guy Nadivi: Ritu Jyoti — Program Vice President at IDC. Thank you for listening everyone and remember don’t hesitate, automate.