Episode Summary: In this episode of the AI in Industry podcast, we interview Nikhil Malhotra, Creator and Head of Maker’s Lab at Tech Mahindra, about how artificial intelligence changed the nature of IT services and business services in general. Malhotra talks about what businesses should consider to make themselves relevant for the future. In addition, he discusses the philosophy shift that has to happen for people to be appreciative of the process of problem-solving, and to see profit and growth from AI. We hope business leaders in the IT services industry will take from this interview the low-hanging fruit applications in the IT services industry.
Subscribe to our AI in Industry Podcast with your favorite podcast service:
Expertise: Artificial intelligence, natural language processing
Brief Recognition: Malhotra holds a Master’s of Technology in Distributed Computing Systems from RMIT University. Previously, Malhotra served as Senior Architect at AT&T and as Technical Lead at IBM.
(03:10) Let’s start by talking about the potential benefits of data science in AI to the technology services industry in India. It’s a booming part of the economy and global companies have offices here. How can IT services companies make the best of the coming trends in AI?
Nikhil Malhotra: The IT services industry has seen a lot of boom in the last few years. In a lot of ways, from the employment perspective to the IT services we can provide to the outside world, the world started changing in 2014 or 2015 when AI started hitting us. The IT services industry was quick to realize this. This upside to that AI was, even though there is lesser research in the IT services industry, the bandwagon of AI has picked up. We started off with automation At Tech Mahindra, I run a small R&D center called Maker’s Lab. We really look at the R&D part of AI, deep neural networks and even techniques beyond that.
If you look at the practical side of what we are doing with customers, it starts from business process efficiencies. Can we make the process efficient? The first thing that happens from a business process efficiency standpoint is to go into automation. Can you write scripts to automate this stuff so you don’t have to do redundant stuff? I think it is through automation that you get insight and say “because I automated and made sure that some part of it was removed from a human perspective, I’d get into the insights and say whether I can do a predictive mechanism or not.
More often than not, we are looking at more automation from an Indian service standpoint. There are few cases trickling in that talk about the AI, whether it is a prediction about customer churn in a telecoms scenario or about chatbots which have become so omnipresent these days that every company has a chatbot.
These are the upsides we are seeing within the services industry.
(05:45) There are firms that “automate” their technology work or other work with human brains. What you’re saying is can we replace that with scripts? Is that for internal operations with the Indian IT services or is this for clients? How do you imagine that shift happening?
NM: The shift is both ways. I think it happens internally as well because you’ve got to do it yourself before you apply something to the customers. For example, from a Tech Mahindra perspective, we have about 16 to 20 chatbots that are functional, which is all for different needs, whether for HR, booking tickets, or finding whatever is going on. To me, chatbots are less conversational and more automated.
They figure out what’s going on in a simple use case. They’re not the Google duplex that you see when Google launched them or something that people are imagining saying they are near-human conversation pieces. This is something we are applying internally.
But some of these cases which are automating testing, and a …. interface to a chatbot in certain cases, are also being applied to our customers. There are different needs. Some are internal to customers. They want the entire piece to happen within their own enterprise and not open it to their own customers. But in some scenarios, we see customers are willing to open it to their end consumers with caveats and with proper disclaimers that say “please don’t treat it as a human. It is essentially a bot that can converse or that can do some automation for you. There are different places where we see this work.
(07:50) What is it like to level up with the skill set of the Indian tech services sector to build these conversational interfaces and automated applications to help clients and help themselves?
NM: What we are doing internally, when I came back from the US in 2014, I started this R&D center called Maker’s Lab, because we make a lot of stuff. That was the whole premise of it. We have a co-R&D where we work on AI and machine learning, We are foraying into content computing at this point and some of those things we do with AI can work on content computing.
But you are right. There is a talent gap. One of the ways we are figuring that out is that we have opened that Maker’s Lab within the sector and we have seen across the world and four in India – Pune, Bangalore, Hyderabad, and Chennai.
We allow R&D associates to come into the lab, see what’s happening. We also want them to participate in some of these disruptive technologies. So that from an Indian perspective, we’ve gone through gradual work on the programming part where we know C Sharps from C++ to Java. But now is the time when design also gets included into some of our conversations. And that design is not really a design based on a language or an algorithm or a piece of code. It is based on a specific algorithm or a specific problem. That’s one part.
The second thing that we are seeing is…people come out of colleges and universities have opened up these channels for disruptive opportunities. They do teach AI and machine learning. I think the practical piece and the nitty gritty of learning AI is still missing. One of the reasons for that is we have to push a lot of effort toward the R&D sector. Why? Because if you look at the machine learning or AI world, typically you have a hundred toolkits that you can work with…You name it and there is a toolkit.
From an Indian IT services standpoint, these toolkits are nothing more than an API (application programming interface) or an SDK (software development kit) which somebody used earlier to create a java program, used Android API to make something from an Android perspective. Now they are using another toolkit to create an algorithm, algorithmic base, or an application. How does an algorithm really work? It’s something that needs to be taught and that’s why there is a gap. You can upskill a person in the service industry to say “here is Tensorflow, read it and learn about it. Then start coding on Tensorflow.” But there is a difference in what a Tensorflow neural network does to what an LST or a long or short term memory does. That’s the gap we have to fulfill. I hope in the next two to three years we would fulfill that gap.
(11:45) Wouldn’t it be great if the services sector could be the place that puts the hands-on work that binds those academics that make them powerful to swell the skill set of the country if that could be done well?
[NM]: Exactly. I think what we’re doing at Maker’s Lab is that we are a conduit between the academe and the business. I have kids coming in from 8 years to about 14 or 23 when they reach college. We do that because we get so many non-jaded ideas from an 8- or a 9-year-old. To take that idea and bring it to fruition, we would have these engineers from college who are being trained on these new technologies. Theoretically, they would also get a corporate feel of how you work across these technologies. So that’s how you become a conduit between the academy and business.
What we find is that when our people who have not been traditionally trained in R&D, they get involved in designing, researching and developing a particular product or a process with these young students.
It’s a lot of fun at the end of the day. You break heads, you break windows but it’s quite fun because you are researching.
(13:23) Is there a way to turn the fact that a lot of those IT customer services technology backend things are happening in India? Can enough of that be congealed to the point where those could become scalable products for the Indian economy, not just a place for a guy to do something competitive?
[NM]: You are absolutely right, and this is something I keep telling in my lab. We as an economy, as a services industry, have to start loving the problem. Traditionally, we’ve always loved the solutions. The reason we are the solution bowl of the world…you outsource the problem, you get a solution. You always love the solution part. You got a problem in a piece of code, on a website, a mobile phone, and now a data science part. We can give you a solution. We’ve lost the touch of loving the problem.
I think that’s where it starts. If you start loving problems, you also start loving the whole process of design to solve that. And if you love the design aspect of solving the problem, you also start researching the around it. What’s the world doing? What is the amount of data do I have today? How would that data affect the end consumer, whether in the US or in India?. Can I take a piece of that data and apply that to the end consumer and make a saleable product out of it? The day we start loving problems, we will get into the research and development wing.
My wishlist is that every services company starts getting hands dirty into the research and development wing. We have a wishlist at Tech Mahindra. But then, this research and development if it gets started in the services industry, it also puts the pressure on universities, academia as well as the other scientific community, to say that now, these other companies also need a research and development person rather than just a heads-down person who actually does…And that’s where it starts changing the process.
(17:44) What are the things that people in this industry should be aware of and steer clear of?
NM: One of the challenges of the AI and machine learning world is getting too attached to it. At this point, we are not cognitive and we have not breached the scale of AI.
The other thing I see is that people have a simple problem but they are trying to make it complex by putting up a machine learning or AI wrap around things. Machine learning and AI are there to stay, but they are there to serve a specific cause and purpose. And that cause and purpose have first to be hypothesized and understood from a human constraint perspective, and then applied. Typically, what we are seeing is the growth in automation that we are building in the services sector in India is on a leap forward into the machine learning world. But the machine learning world requires a) a lot of data, b) a lot of hypothesis from a data science perspective and c) a specific use case where you can apply that AI and machine learning. Not everything can be solved through AI and machine learning. That’s the pitfall that we have if you are in the services world. It looks like a very snazzy term to be utilized when you are talking to your customers and that is something we have to avoid.
(19:10) There is a bit of an interesting tension that you see this a ton in the United States, but I imagine it is the same out there where you have to drop these keywords to seem that you are hip and you are with it. You have to be able to offer it because somebody is going to ask if you can. But there is this insidious force where we’re bragging about capabilities that don’t exist or leveraging and talking about AI, winning a contract by saying AI or trying to, when in fact this shouldn’t be an AI problem. It’s not something that’s going to require AI to get a result.
[NM]: I keep saying to my people, “You gotta love the problem. Don’t love the solution. The why is important. The what and how would come out. The what and the how could be an AI and machine learning piece. It could not be an AI and machine learning piece, but why has to be understood correctly. We have so many examples from history. The steam engine was built somewhere in the 1800s, but unless a motorcar was made in the format that Henry Ford did, it never scaled up.
There was a technology and a design behind it. How did that technology solve something? That is immaterial as long as I could give that seamless experience to my customers. I think that’s the important thing we have to look at. It doesn’t matter if I have to apply AI but there are cases where you have to.
(20:37) Anything else that you would say in terms of risk to avoid in transitioning to AI in the Indian tech sector? Anything in closing that you people should be mindful of?
NM: The sector is here to stay and here to work across. We have to look at a broader perspective of AI and machine learning. ETIO and the government have already placed a few bets on AI and they are working on some of the things that could change the way India looks at AI as well as solving some of our complex problems like agriculture, education, languages a stuff like that. But what people need to be mindful of is the task of getting into it but also the process of getting in.
The process of getting into it is not necessarily generic, which we Indians typically do. We Indians typically have a mindset that if I follow up then I would l lead them to this specific sector, which is not there in the US. We have to be mindful of the fact that you could reach an AI place even if you were an economics student. You’ve gotta love data to reach the AI part of it. If AI is going to be here to stay, AI is going to solve big problems but what we’ve got to look at is a) love the problem but b) also be mindful of the process of finding that problem out.
Subscribe to our AI in Industry Podcast with your favorite podcast service:
Header Image Credit: vbriindia