Indian IT firms cautiously advance GenAI projects, facing scaling hurdles

Despite the hype, top IT firms and analysts caution over slow adoption and high abandonment risks in generative AI projects

A recent study by Gartner predicts that 30 per cent of generative AI (GenAI) projects will be abandoned after the proof-of-concept stage by the end of 2025. The statement is full of concern as GenAI is being seen as the new growth area for the over $200 billion IT (information technology) services industry.

One then wonders about the significance of the recent management commentary made by the top IT services firms and how generative AI is being re-evaluated. For once, the CEOs of India's top four IT services firms were on the same page as they said that generative AI deals continue to be small and hence not yet contributing meaningfully to the revenue.

“Currently, only a few generative AI deals have been fully executed, with the majority still in the proof-of-concept (POC) phase as organisations strive to demonstrate and realise value. A few large IT services firms such as TCS, Wipro, and Infosys have quickly addressed market needs and embraced generative AI early in the process, while other firms have adopted a more conservative approach,” said Biswajit Maity, senior principal analyst, Gartner.

“Although 73 per cent of providers have plans for generative AI, these plans are fairly conservative,” he added.

This is also evident in the fact that only Accenture and Tata Consultancy Services (TCS) gave out separate numbers in terms of deal pipelines from their generative AI projects during their quarterly results.

"In general, there's a lot of talk and discussion with generative AI, but the programmes, even though they're not POCs, the actual projects are not large revenue projects and transformation is not so much what we are seeing," said Salil Parekh, CEO, Infosys during the press conference post the firm's Q1 FY25 results.

Leaders also agree that customers are realising that implementing generative AI is a complex process. It is not as easy as it is being seen in consumer usage. Aiman Ezzat, global CEO, Capgemini, said in an interview with Business Standard: "Clients are now understanding that generative AI is still very complex, and like it happened while using AI, they also need to move bit by bit in generative AI adoption. Generative AI also got a little hyped, especially around productivity and cost cutting. But in reality, it’s not happening. Like AI, generative AI also needs change in processes, its use case by use case, then evaluation, change management etc…it is not a big bang approach. It is good that the hype around generative AI has come down and now we can really get to work."

Gartner also points out that these POCs will be abandoned due to reasons such as poor data quality, inadequate risk controls, escalating costs or unclear business value of generative AI projects, said the report.

This statement resonated with C Vijayakumar, CEO, HCLTech, who during the Q1FY25 results said that there's a need for improvements in data and cloud infrastructure of clients to better implement generative AI.

“To leverage generative AI in an effective manner, a lot of customers need to continue their modernisation journey and the streamlining of data. So, I think there are a lot of prerequisites that are becoming more and more important as we complete these POCs for a number of customers,” he said during the Q1 earnings call.

HCLTech in its first-quarter results this year called out 11 generative AI-specific deals but did not give out any separate pipeline for the same.

A major challenge for organisations is arising in justifying the substantial investment in generative AI for productivity enhancement, which can be difficult to directly translate into financial benefit, said analysts.

TCS, though, remains an outlier among all the players. The largest IT firm in the country doubled its AI and generative AI pipeline in the quarter to $1.5 billion.

K Krithivasan, CEO, TCS, highlighted the shift from the POC stage to enterprise-scale application, during the Q1 FY25 earnings call. "Customers are looking at scaling out the POCs and pilots by implementing necessary guardrails. Mature customers with a solid cloud and data foundation that were able to experiment with multiple generative AI use cases are now looking to reimagine parts of their value chain to make them AI-native."

He also pointed out a sense of caution that clients have while investing in generative AI.

"While the interest is strong, organisations are taking a calibrated approach to measure the risk potential and organisational impact while chalking out their roadmap for generative AI adoption," he added.

He said that while generative AI offers productivity gains to customers, widespread high-level productivity improvements are not yet common.

"When the discussions happen, we do offer customers to explore generative AI for software engineering and building productivity, but it is not becoming a huge demand yet," he said.