AI/Automation

Artificial Intelligence: All That Businesses Need To Know

Artificial Intelligence (AI) is a high decibel discussion. When the conversations turn to AI, opinions are usually polarized into one of the two extreme camps: those who believe that AI will make our lives better, and those who are convinced it will accelerate human irrelevance. The reality, as always, is somewhat more nuanced. AI technologies are taking disruption to a whole new level. While business leaders are concerned about how AI can be potentially misused, human workforce is getting worried about its role in the future of AI-led business.

In this episode of Infosys Podcast, Infosys anchor Alex speaks with Sudhir Jha, senior vice president and head of product management and strategy, on the latest trends and maturity of AI technologies. Sudhir explains how appropriate the concerns related to AI are and how businesses can leverage AI technologies to stay relevant.


Loading...

Podcast transcript

Alex: Hello everyone! Welcome to Infosys Podcast, this is Alex speaking, your host, and I have Mr. Sudhir Jha with us today. He is the Senior Vice President and the Head of Product Management and Strategy at Infosys. On this podcast we will be talking about AI (that’s short for Artificial Intelligence). So with me I have a list of questions for which our expert will help us find the answers. So let’s first start with Sudhir giving us a brief introduction to AI.

Specifically, what are the current trends in AI at the moment?

Sudhir: So there are a few things. One is AI is still very hard. So it’s hard in terms of skill level required —you still need very highly skilled data scientists to create algorithms, you need a set of highly skilled computing of digital linguistics to basically do natural language processing. So the trend right now is to make tools that actually allow AI to be used by lesser skilled people. And this is similar to how programming sort of started. If you remember, 20 to 30 years ago it was very complex — CE programming, assemblers and compilers and all that stuff — now a seven-year-old person can actually do programming using graphical interfaces. AI sort of initially evolved to the same place where anyone can actually use AI tools. So that’s one trend that actually is happening.

Second is that AI is still somewhat of a black box in the sense that you can create an algorithm and you can basically do predications, but it is hard to explain why the prediction is that way and not the other way. So understanding and explaining AI is another trend that is happening, which basically allows not just algorithm to predict, but also tell you why it is predicting that way or why the results are in the way they are supposed to be. So those are sort of at that level.

In terms of technology level, I think there is still some basic research that needs to happen in terms of how to provide the hardware that performs optimally for the AI kind of technologies — some basic research in terms of how to perform better with much less data. So today most of the accuracy of the algorithms is dependent on how much data you have and how many labeled data you have. That is harder to get in certain circumstances, so how do you actually do that better without having that much data? Those are at that level the technology trends that are also there.

Alex: You see the fear that people have, that lay people have, about AI is quite simply, you see, [that] in the wrong hands AI could be used like an instrument of destruction whereby the AI can take over to such an extent that it could actually be self-sustaining and have such an intelligence of its own form that it can then deem human beings as inefficient and redundant. And human beings can see the other human beings as redundant as well because that workforce is no longer needed when you have an intelligent system to basically run that. What would you add to that?

Sudhir: So there is always a fear. I mean to some extent there is always a fear with every technology. If you go back to going from rocks to hammer and things like that, there’s fear to that. You have cars and accidents and people getting killed, you have Web and you have all these Internet crimes that are there as well. So there is always fear with technology. Any technology in the wrong hands can do much more harm, so that’s always there. I think the responsibility is for the innovators to make sure that they are actually being conscious of that. And this is what I feel, that with every technology that is being developed, there has to be equal sort of effort in trying to create very consciously things that would prevent people from using technology in the bad way. And so already there is that conversation happening, which is good, and there is equal sort of effort being created.

I think that because of the fear you can’t sort of stop using any new technology. So, saying that don’t do AI because there is a fear, that’s not the right thing. The thing is how do you sort of ensure it doesn’t get to that.

I think getting to a place where robots will take over the world is much, much, much distant in the future. We are not even close to where robots can think on their own today, right? Even with all the conversation about self-learning, most of it is what has been taught to the robots or what has been taught to the machines. So, it’s much far distant in the future.

Then you have this trend of how to use AI to provide better human-machine interaction or, in general, better interfaces. People don’t necessarily like to type as much, so how do you do a voice-enabled stuff, how do you do your facial recognition and use emotional intelligence and things like that. So, the whole interaction of humans with the computers and systems is changing and that has to be much better, hopefully in the future.

Alex: That’s how it would benefit a lay person, like myself, but then why do enterprises need AI?

Sudhir: Enterprise[s], in my view, need AI for both staying competitive and staying relevant. To me it seems like companies that didn’t get into Internet or the Web sort of thing, many of them became obsolete. People who didn’t sort of get into mobile bandwagon, many of them have lost market share. So in my mind, you almost ‘have to’ use AI today because otherwise you are not going to be relevant in the future.

It can also be used to differentiate yourself from other people because any tool, any technology gives you that platform to use it in a better way than your competition. And so if you use AI in terms of its full potential, you can better serve your customers, you can improve all your business metrics including sales and margins, and profit, and all kinds of stuff. It allows you to be more efficient, it allows you to scale better so you don’t have to hire 20,000 people if you really want to grow.

So in every dimension you can actually do something very different from your competition and create differentiation. Also, I think it’s almost inevitable that everybody has to use it for that. And it can be used because AI allows you many different technologies and not just one thing — it has machine learning, then it has speech recognition, it has different tools to do different things for the organization that you can leverage across the board for all those benefits.

Alex: So in essence, what you are saying is, if you don’t move with the trends, then you are going to miss the boat…

Sudhir: Exactly.

Alex: …and you would be standing on the harbor wondering where everyone else is going.

Sudhir: Yes, precisely. I mean you most probably are not going to be relevant in 10 years.

Alex: And it’s interesting you say that because I remember the dot com boom. And going back to the industrial revolution, there were these groups who were the laggards who basically went and broke up all the machinery because they didn’t believe in the development of this machinery and the evolution of the machinery, and people thought that of the dot com era as well. That itself is a challenge because there is a lot of people that you have to convince and that, I guess, is a job in itself. So other than that, what are the challenges in adopting AI?

Sudhir: For an enterprise, the challenges are, in my mind, less about technology and more about the human aspects to it. Technologies are there and there are various tools that you can actually leverage that. The human aspect is one. There will always be naysayers and how do you sort of get them to adopt and change their mind about using a new sort of technology? Similar to, again going back to, the dot com [and] Internet revolution, there were many companies that didn’t go there because they thought it was a very small fad, [and that] nobody is going to buy things on Internet, right? I mean that was the thing, 15 years ago, people were saying. Walmart is still catching up to Amazon because of that.

So there would be that sort of stakeholders, some of them will not be convinced that this is the way to go and some people will also not come to the realization and not want to join the party [because] they think it will be impacting them personally. If their job is something that can be eliminated using the AI technologies, then that resistance is also going to be there. So that is one big challenge that needs to be solved, because ultimately organizations are built with people and if the people are not convinced then you are not going to be successful.

The second aspect after that is — do you actually have the right skills in your organization to adopt something new and make it successful? That is again a challenge. As I was saying before, with the new trend there is going to be more tools and more help for people to use AI without a lot of skill. But then the third level comes, which is basically, are you trying to solve the right problem in the first place, because there are certain problems that can be solved today using AI technology and there are certain problems that most probably [are] not the right ones to solve today because the technology is immature and there is still sort of more work to be done.

Selecting the right problem becomes also a challenge and, again, the more experienced you are in AI technologies, the better the chances are that you are selecting the right problem. But if you are just going by the hype, then you might try to solve something that actually is further away than you think.

Alex: You talk about problems and solutions, but then do you look for the gaps, gaps where technologies don’t exist? Does Infosys actually look to fill those particular gaps as well?

Sudhir: The AI and automation platform Infosys NIA, that we launched this year, precisely does that. It takes most of the existing technology, lot of open source technology and it slowly adopts that and then builds things that are sort of missing. So, one is just hardening the open source technology in terms of [being] skill agnostic. But, as I was saying before, there is huge gap in terms of automating things. So, how do you automate the whole data science process where you can have the machine itself learn, basically pick the right algorithms, pick the right features, and do the prediction all by itself and, therefore, you don’t need very high skilled people doing that? Those kind of gaps. Or how do you make it so that it’s very easy to prove why this is working or not working? How do you do automatic calculation of impact that the AI technology is giving you? Those are the things that are missing in most of the technology outside and that’s kind of where Infosys focuses on and built this platform that allows you to do those things in a much more streamlined fashion.

Alex: Okay so you have the next gen platform, the next gen AI platforms. How do next gen AI platforms address these challenges?

Sudhir: So one thing is creating, tooling on these platforms that allow it to be used by folks who are normal business analysts or engineers. They don’t have to be data scientists or don’t have to be computer scientist or something like that. There is a much faster rate of innovation in AI, every six months there are new tools being built and put in open source and things like that. So building a platform that actually can absorb those innovations much more quickly, [and] having a very flexible platform.

AI is not a single tool. As I was saying before, it basically has natural language processing, natural language understanding generation, machine learning, deep learning, all these different kinds of things. How do you have a single platform that actually has all these tools, because different tools are appropriate for different problems in the enterprise? You don’t want to have the enterprise using 20 different tools from 20 different vendors. How do you create a single platform, a single comprehensive platform, that actually can do all those things?

Our focus is on flexibility, comprehensiveness, and then low cost, which is sort of built into our gene – Infosys always wants to be a provider where you are not asking for very large investment from the enterprise. And again, using open source allows us to keep the cost low not only of the platform itself, but also the delivery work of it and also the usage of it. Because a lot of things are automated in the platform, it is much easier to use and it’s much faster to get results – you don’t need a two-year process to implement a particular system or solve a problem. You can do that in three months and that reduces the cost as well.

Alex: You are not looking at a small platform, a small stage. You are looking to address the whole world in that respect and take as many enterprises, businesses like large conglomerate companies, under your umbrella so to speak. Infosys is not necessarily, as you said, a commercial company where you are interested in what the lay person or the consumers intend, but [more about] how you can benefit other enterprises to do that job for you. In the sense, you are just consulting the larger firms, so to speak. Is that something that Infosys is going to consider for the future? Or like Amazon, you know how they have grown or like how they started by selling books and now they have got a TV channel, is there anything that, you can lead on that, Infosys is planning for the future?

Sudhir: So I think that Infosys at least in the near future is not geared to go direct to the consumers. I think the entire DNA of the company is sort of built in understanding the enterprises’ domain, understanding their needs, and serving their needs in the most efficient manner. And I think that is what we do really well, but that doesn’t mean that we will be restricted to just solving the problem they are telling us to solve. We can always find problems for the enterprise.

In fact, one of the things that we have done recently is this Zero Distance initiative where our employees, when they are working with the company, they are trying to understand the different problems that the enterprise might have that they have not actually hired us to do. But we sort of surface that and then we actually solve that. So we definitely are increasing domain that way. We are also moving more into the products area  using products to do innovation and solve even more complex problems. But I think the domain still remains to be the enterprise. I don’t think that direct to consumers is something that, I mean you can never say no, but in the near future we are not geared towards that.

Alex: In a grand scheme of things, how will enterprises be impacted by AI?

Sudhir: The first step is always the efficiency and cost saving. That’s sort of where the enterprise[s] always start. Because that’s the easiest way to start. So if something was costing us US$100 [and] if I use AI technology/any other technology to save US$20 [then] that is one easy metric that enterprises do. Even for AI, if you look at it, a lot of enterprises are starting to use AI to automate and to save cost. That’s one metric.

I think where we are sort of helping the enterprise change, and they are also moving towards, is figuring out how AI can actually address the other business metrics like your sales, and your marketing, and your reach to the consumers. Now, how do you actually launch new products more efficiently? How do you actually expand into different markets more efficiently which then gets not just into the cost savings, but actually in terms of increased revenue, increase profit, increase margin, those kind of things? And that change is happening now where people are sort of starting to measure that as well. So, if you are serving the customer better, are you increasing the customer’s lifetime spent with you? And how do you sort of measure that and use that to fund the AI investment? I think [that] is a positive sign and hopefully will fuel even more growth in the AI field than it is right now.

Alex: How do enterprises get started on their journey with AI?

Sudhir: It always starts with the right problem. And that is the hardest thing. How do you start your journey to solve the right problem for the enterprise? In my view, the problem should invariably be a problem that the company is already solving, but not solving in the most optimal way. Because if you try to solve a completely new problem, it is going to be harder to explain why AI is doing the better job than something else.

If you already are solving a problem for a while, and it is somewhat suboptimal — so what I mean by that is, if you are using rules-based engine to do predictions or doing fraud management using that or you are using humans to enter data from spreadsheets or documents, which you know are inefficient ways of doing things — you can use AI to do it better. Then it is very easy to prove that you actually did something. Like [this] was earlier and this is now, and there is AI between the two, and it is very easy to prove that. So I think, having a problem that is being solved sub-optimally and then using AI to then solve it [in a] slightly more optimal way and then measuring the impact of that is an easy way to get acceptance in the organization, and that’s where you should start.

The other thing important in the problem also is to make sure that you’re using an AI technology to solve the problem that is somewhat mature. So, for example, supervised machine learning is something that has been used for many years, and it’s quite mature. Something like a speech recognition is still sort of getting more mature but maybe [it’s] not as mature. So you don’t want to take a problem where your interface is completely human with lots of accents and you have to get [it] perfect to really make a difference. You use something that actually is somewhat [a] mature technology to start with because it is very important to be successful in the first project that you are doing.

And be extremely sensitive to people’s reaction to it, so that culturally there is no big resistance. You have to involve them in the process, making sure that if this new paradigm is going to have some people lose their job [then] you make it very certain that they will have something else to do. So you reskill them and have already a path for them to have opportunity in the future, so that they don’t resist this change.

Then you want to make sure that there is stakeholder buy-in for [both] the technology option, but also for the people management and how their future is going to be tied to that.

Alex: How do you look at the future? Not only for yourself, but for Infosys as well?

Sudhir: The thing that motivates me is the opportunity that Infosys has and I have, and everybody else has, to shape the future in the way that we think is going to be the best for our next generation. So, I have a kid, 16-year-old, and my thing is, ‘Okay, what are we doing to make sure that the next generation inherits the world in a better way than what we inherited from our parents and grandparents’.

Alex: Thank you so much, Sudhir, for sharing your knowledge today. Thank you all for listening. I hope it was a fun-filled session. And for more information on AI, please visit Infosys.com. And we look forward to you tuning in next time. Thank you! Good bye!