As enterprises shift from traditional, rule-based systems to AI-driven operations, data is becoming a critical business asset. A typical enterprise AI project now allocates approximately 50% of its budget to data-related work.
Vijay Gopalakrishnan, partner at Deloitte India, said spending varied. His estimate allocates 15% to core AI development, 15% to design, functional requirements and governance, 50% to data-related activities and 20% to infrastructure.
“All these components are critical to supporting AI initiatives end to end, and consequently, each is witnessing growing enterprise demand. However, in recent times, the availability of technology infrastructure and technical skills has become less of a challenge. The greater priority for enterprises today is identifying the right use cases, deploying them effectively, and scaling them in a cost-efficient and secure manner while ensuring a clear return on investment,” he said.
Sridhar Mantha, chief executive of Happiest Minds’ Generative AI Business Services unit, said at Happiest Minds Technologies, data engineering, modernisation, quality, cleansing and governance account for nearly 40% of its data and AI work.
The remaining 60% of the company’s data and AI work involved generative AI, agentic AI, machine learning and statistical modelling.
“At least 70 to 80% of our proposals cover both data and AI, even when the client’s end goal is to build an AI models or agents,” Mantha said. “So we spend a lot of our time fixing the data first, cleaning it, bringing it together and putting governance in place, before the AI work starts.”
Mantha said demand for data preparation, quality and governance had increased as clients moved beyond pilots. Agents draw information from several enterprise systems and act on it, with limited human involvement. Metadata management and data lineage were becoming necessary, he said, so companies could establish where data came from and how it had changed.
Anushree Verma, senior director analyst at Gartner, said the strongest enterprise demand was for agent deployment, as investment shifted from static automation and AI copilots towards systems that can plan, decide and execute. This requires a control architecture, she said.
“The key barriers in moving AI projects from pilot to production are Data readiness, Integration and operations, Risk and compliance, Skills and operating model and Business value and change management. Overall, data is one of the barriers but not the biggest one,” Verma said.
The shift is also reflected in AI revenue reported by large IT services companies. Tata Consultancy Services said annualised AI revenue reached $3.1 billion in Q2 FY27, crossing 10% of its overall revenue. HCLTech reported advanced AI revenue of about $876 million annualised, up 72% year-on-year.
