Large contracts, typically $500 million and above, are a key revenue driver for technology outsourcing companies.
“In large deals, we are using GenAI to quickly understand the current operations of our customers, business processes and the tech to become productive in transitioning and then becoming operational. We are also doing a lot of model innovation, not just limiting to use cases but also creating a lot of specialised, fine-tuned models to help us deliver our services better and these are IPs (internet protocols) specific to us, helping us differentiate and deliver better,” Tarafdar said.
According to him, the impact of the ongoing pilots will be visible only after the next 6-12 months. He said, “We will have to reach a number … For them to go full out in production and scale, I think will take around 6 to 12 months, when some of these outcomes will be more available.”
Sectors and outcomes
Discover the stories of your interest
The Bengaluru-based IT major is seeing a lot of AI and GenAI adoption in financial services and communication, media and telecommunications (CMT), especially with telcos that are investing a lot in AI, Tarafdar said.
“Within manufacturing, in the industrial and automotive clients, logistics and consumer goods companies are picking it up. Every other segment is also experimenting,” he said, adding that the company is making a lot of innovations in AI models, building on top of its cloud offering and AI platform Cobalt and Topaz, making it a full-stack solutions provider.
“In the last one year, a lot of GenAI solutions have gone into production and are being used at scale by our customers,” Tarafdar said. “For example, a bank rolled out its own version of a GPT, currently being used by 20,000-plus users within the bank, helping them with about a 10-15% productivity boost. Another is customer service, an area where a lot of the AI assistants have been rolled out.”
Among other use cases by clients, he said a company launched 3-4 apps within weeks in the organisation using Infosys platforms to roll out GenAI offerings. In Topaz, the software services exporter is looking to get clients with data readiness and responsible AI.
Infosys is also using AI in software engineering, employing code assistants to write and test codes faster. Additionally, AI is being used for migration and modernisation of existing solutions. Internally, Infosys is seeing productivity benefits for employees in software engineering improve 10-30%, he said.
“The upcoming areas where now the clients are starting to look beyond productivity and efficiency is to see how they reimagine the business process using a combination of digital, cloud and AI and GenAI because the value will be much higher,” Tarafdar said.
Talent advantage
“In terms of talent, India is in a very good spot,” Tarafdar said.
“With the kind of engineering and AI talent we have, applications (apps) obviously are a big area for us. With the developer base in India, we will have a very strong play in building apps … I think the apps is where the volume will be at play and our talent will play a significant role. We are also getting deeper into the stack with platforms and models and this deep expertise is required to create truly innovative solutions in future,” he said.
He thinks India needs to keep scaling higher and quickly to create more use cases and solve more problems by leveraging the power of cloud, digital and AI. “The faster we do, the faster we can create an impact,” he said.
Tarafdar also mentioned three challenges in implementing AI. The first being change in management efforts and mindset to roll out adoption. “The accuracy and outcome of AI solutions improve over a period of time, which means the more we use, the better it becomes. That requires a significant change in management effort, mindset to be rolled out within the organisation… unless the adoption happens, you won’t get the benefits of productivity, efficiency, cost, etc.”
The second element is data readiness for GenAI solutions as most of the projects consume about 60% of the time in data preparation today.
“A lot of data gaps must be fixed, and a lot of test data will have to be created for validating that the AI solution is performing right. So, there is a significant amount of work involved in data readiness through pre-training, post-training rollout, which sometimes if not estimated properly up front, creates issues,” Tarafdar said.
The third, he said, “is making sure that these are built in a responsible by-design manner, which means today in most of the countries, AI regulations are coming into play”.
