Mr. D.R. Balakrishnan (aka Bali), Senior Vice President, Service Offering Head, Energy, Communications, Services, and AI & Automation, Infosys was in conversation with Ms. Sindhuja Balaji, NASSCOM.

Please read on to find out how rivetting the world is from an AI adoption standpoint, some of the challenges and how they can be addressed. As Mr. Balakrishnan says, “It is the second inflection point of AI.”

Q. We are now in year 2 of the pandemic. Please share your insights on where are we from an AI adoption standpoint. And, what does the future look like?

I think it’s best captured by what the CEO of a Telecom company (our client) told us. Covid has done more to DTx (digital transformation) than what we have seen in the last ten years. To add, even senior citizens who were earlier reluctant to use digital services have started to adapt, to make their daily lives much easier which speaks volumes of where we are headed.

We like to see this as the second inflection point of AI. The first one was when AI entered the consumer space with the hyperscalers and the digital media companies embedding AI in their services and turning the tables on peers who were slower on the uptake. The second inflection point which I have mentioned is about traditional companies using AI in a much bigger way and accelerating their DTx journeys. AI cannot be seen in isolation of course, and it’s the convergence of multiple technologies that have led to maturity, including workplace dynamics and the distributed workforce.  

While digital is the way forward but it has to be done in a sustainable & responsible manner - which brings to the forefront the entire paradigm of Responsible AI. While going beyond PoCs to scale adoptions across organizations, we must see this paradigm as critical. 

Q. As a leading solution provider, how are you future-proofing your AI investments – apportioning resources to companies which are mature in AI adoption and others? 

There are two ways of looking at it. On the one hand, we have to adopt AI in our services, or else we will be cannibalized. On the other, how do we take these services to our clients to ensure their DTx journeys are smooth and they are ahead on the curve. Even traditional IT services such as application maintenance, application development, testing, migration, etc. have AI embedded today. For instance, in application maintenance, the approach is predictive now and AI-led. It is not only about automation but also ways in which we can leverage cognitive automation in the resolution of incidents. 

The traditional approach: An incident occurs and a ticket is raised. The engineer looks at it and fixes the issues.

AI-led approach: Through reverse experience (models based on previous experiences) the incident, now, can be better understood. This learning will eventually go into the model leading to greater accuracy in future predictions. Inasmuch, corrective action can be taken well in advance.

Similarly, in application development, we can use GPT-3 to generate codes and go to the next level. In migration, AI tools can convert from one language to another (C++ to Python or Java to Python or other combinations). In testing as well, we use cognitive automation extensively and leapfrog towards higher efficiency & accuracy. Across services, we use AI to continuously drive improvement.

The other leading thought today is how do we merge front, mid & back-office to see it as One-office? To do this, we cannot be seeing departments in isolation but what the clients want to achieve through their DTx journeys. So, AI is used to rethink processes (discover/democratize & scale AI). Once again, I want to reiterate – whatever we do through AI must be done responsibly and always bearing in mind the elements of security, privacy, ethics, and bias. That is how one can go beyond PoCs to build scale.

Q How do you strike a balance between ready-to-use proprietary solutions and customizing them based on specific needs?    

Yes, we have both horizontal & vertical solutions. In the former, we have a suite of solutions that can be used across industries. For example, we have capabilities that can be used to read PDFs & images (much beyond OCR capabilities) and recognize what action needs to be taken/information that needs to be extracted. This is a horizontal offering and it can be used across multiple industry scenarios. Under Bot Factory, we have created Infosys Cognitive Intelligent Solutions – microbots that can do specific tasks that are industry-agnostic. There are 10K such microbots. These bots can be customized as well – algorithms can be written specifically for let us say a claims process. These bots can be quickly assembled and algorithms can be improvised upon to address a specific need. You can say it’s a microservices approach to leveraging AI & cognitive automation. To add, AI-led Intelligent Document Management solutions will have different use cases for industry verticals – for instance, in Oil & Gas, the requirements will be different from say, the telecom sector. We also have language-based models (NLP) that can interpret speech to text and can be adapted for specific verticals. Then there are industry-specific solutions such as in healthcare, fraud analytics, financial services among many others.

Q. Can you please give us some examples of Infosys-led Edge AI?

Yes, these are use cases where the latency is very low. Autonomous vehicle application is a case in point. Currently, we are working with visionary leaders to build edge AI solutions. Internally, we have developed solutions that can detect covid cases. Our solution which starts with a facial scan can detect elevated temperature levels; if social distancing is maintained in our campuses, etc. Our clients can use these solutions too. Similarly, we have applications that can analyze images captured by drones; detect corrosion in plants, among a host of others.    

Q. Please share your insights on skill-sets that are required to drive this massive opportunity.

Very important question. When we talk about AI we immediately think of data scientists but we also need to understand the lifecycle and the different skill sets that are required in various stages.

To start with, we need to understand what the problem is and how exactly will AI be the game-changer. A significant amount of consulting skills are required to understand what the client needs, their business, the industry dynamics, to build a strong case for AI. In addition, deep architecture capabilities are required to not just build algorithms but also to ensure that the architecture is future-proofed. Design skills are yet another critical area – ultimately it’s about giving a seamless experience to the user. Similarly, there’s the Validation team which will ensure there’s no drift in the model for which a different kind of testing is required. Then there is the ethics part of it – an AI bias will impact the entire organization. While there’s a core team that will drive AI but it has to be an organization-wide approach that is much broader.  

Q. Can you please articulate some of the AI adoption challenges during the pandemic-struck period?

1.     While Boards today understand the implications of not adopting AI, but it’s not always easy to articulate the problem statement and how exactly AI can solve the issue. A lot of effort needs to go into this so that the RoI can be derived. Only then will the PoCs scale up.

2.     Org structure, dynamics, and their complexities.

3.     The tech ecosystem itself – hyperscalers are offering a plethora of products & services, Open Source AI, AI startups, AI product companies. These developments are happening at a rapid pace and often the challenge faced is about keeping up with what’s happening out there.

4.     Skill sets – gathering the right team with the desired combination.

5.     AI ethics – governance, trustworthiness, removing bias, etc.

6.     Regulations.  

Q. Any last piece of advice to leaders?

o  DTx is not a choice any longer – it’s an imperative.

o  Keeping abreast of technology and what it can offer.

o  And how it can combine with business to fetch higher results.

o  Make AI applications ethical & sustainable.

o  You will encounter multiple complexities but they will have to be managed.