Target is to double organisations' AI adoption rate: TCS CTO Harrick Vin

It has been more than two years since ChatGPT, an artificial intelligence (AI) chatbot built by OpenAI, took the technology world by storm. As the euphoria ebbs, enterprises are adopting AI slowly. HARRICK VIN, chief technology officer (CTO) of Tata Consultancy Services (TCS), told Avik Das and Shivani Shinde on the sidelines of the Nasscom Technology and Leadership Forum in Mumbai what organisations should do to scale up adoption. Edited excerpts from the interview: 

Adoption of generative artificial intelligence (GenAI) is slow across enterprises, according to analysts. What is the reason?  

Adoption is about value realisation and what we have seen is low double-digit percentage: Less than 15-25 per cent of all use cases going into production. The uptick is really slow, almost a trickle, for two reasons: Have people picked the right problems and have we built the right solutions? When GenAI started, everybody picked up random use cases which had no business value. 

The other thing is the way enterprises have engineered the solution. We keep talking about the four levels of those models: World-wise, industry-, enterprise-, and activity-wise. When you build agents for the solutions, one has to understand the combination, as choosing the right model and orchestrating them is an art form. 

Also, we underestimated the effort involved in adopting and scaling these models. When we adopt AI in an organisation, it changes the role of the employees to trainers and interrogators of machines from just doers. You do not just reskill people on AI but they also have to master soft skills. From being a software engineer used to writing code, you have to ask why am I building a particular feature, how should I architect a solution. It is more of what and why rather than the how, that takes time.  Organisations will have to be designed for change rather for functionality. So we talk of a perpetually adaptive enterprise.  

TCS last year announced the creation of WisdomNext. Can you share updates? 

The reason for setting up the AI Cloud unit was to build the foundation that is often on the Cloud. Layer zero was essentially compute infrastructure, which was well integrated, on top of which we have the various models. Then all the hyperscalers have their models and it allowed us to play with all of them. WisdomNext cuts across the hyperscalers. What we see is that while proofs of concept have accelerated, it does not mean that adoption is increasing. The target for this year is to push more for these AI-led redefinitions and doubling the adoption rate. 

What makes an organisation AI-enabled? 

That is where assist, augment, and transform come in. When an enterprise is looking to start, it will typically be in the assist stage. When we say transform, it is about value creation such as what is the future of the business, using technology such as GenAI. It is not just using one technology but a combination of technology that will help you define the value. GenAI is not the magic app and that is where the mismatch between hype and reality happens. In any technology, we overestimate the short term and underestimate the longer. Right now, I would say, about 85-90 per cent is assisting, and about 10 per cent augmenting.  

Will the political climate in the US put the brakes on GenAI adoption? 

I do not think there will be a slowdown. There will be small hiccups along the way but that is fine. The bigger challenge is for organisations to have a clearly defined AI strategy rather than ad hoc representation. If you have the right strategy, then you can implement the solutions better. 

There are three phases in this implementation: Assist, augment, and transform. Assist is mainly in improving productivity and the initial deployments are self-funded, which will continue. Augment involves making people better, helping them take better decisions. 

One issue is that despite GenAI’s potential, actual adoption among employees is low. Is it true for TCS? 

I feel it is more about AI for skilling rather than reskilling or skilling for AI. It is also about how you use technology to nudge people to become better. Using AI for talent engagement is critical for organisations at this stage.  

We are trying to get people to adapt to these technologies and use it in their day to day work. We have to make people feel that those tools are good for them, something that helps them elevate. That is where change management and talent transformation goes hand in hand. It is a good time to focus on genius partners, and not just the task-level copilots. That augments my ability to make better decisions.  

How are you helping organisations create AI strategies?  

Several organisations have felt the need to create such a strategy because of the low adoption, and a mismatch between expectations and reality. They need to realise that strategy is a team sport and not a technology problem. It requires teams from business, security, compliance, technology and operations. That realisation is setting in slowly.