Indian IT firms are often first in line to implement enterprise AI, but a closer look at delivery reveals constraints rooted more in data governance than in algorithms. Practitioners working hands-on with AI systems say data readiness across the sector is best described as “partial.”
Data readiness is important for organisations to make informed decisions. With high-quality and clean data, they can innovate faster, offer better customer experiences, and ensure regulatory compliance.
Sourajyoti Datta, a data scientist at Vaillant Group with about 12 years of experience in the field, told AIM that Indian IT firms today have “strong data engineering talent and growing governance frameworks,” but outcomes still depend heavily on the client’s underlying data maturity.
“The biggest constraint is not models but data,” he said, pointing to fragmented architectures, inconsistent quality, and immature lineage as reasons why only a minority of AI initiatives make it into stable, value-generating production.
This unevenness cuts across tiers. Datta notes that large Indian IT firms now operate large cloud-native data practices and proprietary GenAI platforms, while mid-sized players often move faster on narrower, domain-specific problems.
The trade-off, however, is scale. Mid-tier firms can execute quickly within defined scopes, but struggle to replicate that success consistently across complex enterprise environments.
Then there’s the question of whether Indian IT companies are building their own data platforms or relying on external ecosystems.
Datta argued that most firms are building core data lakes, lakehouses and machine learning operations (MLOps) blueprints in-house on top of hyperscaler infrastructure, using partnerships and selective acquisitions to accelerate rather than replace those foundations. In his view, data platforms are increasingly treated as internal intellectual property, even as vendors continue to depend on cloud providers and specialist partners to fill gaps.
Pravat Jena, a senior data scientist at Dell with 14 years of experience in data strategy, took a more sceptical view. He said most Indian IT firms remain “strong at pilots and PoCs (proofs-of-concept),” but only a subset are consistently ready for production AI at scale.
The limiting factors, he argued, are operating models and governance maturity rather than technical capability. “Most firms are not building core platforms fully in-house,” Jena said. “They rely heavily on hyperscale-native stacks, vendor tools and partnerships, with limited proprietary differentiation.”
Both Datta and Jena agreed that data quality and governance remain the weakest layers in AI programmes. Jena described governance frameworks as often existing “on paper,” with uneven implementation across legacy systems and multi-cloud environments.
The result is that AI systems scale slowly beyond controlled use cases.
Datta noted that despite regulatory pressure from India’s Digital Personal Data Protection Act and emerging AI governance guidelines, only a minority of enterprises feel adequately prepared to support scalable AI workloads. According to a 2025 McKinsey report, while almost all companies worldwide invest in AI, only 1% believe they are at maturity.
This gap between intent and execution explains why many AI projects stall before reaching enterprise-wide deployment. MIT places the failure rate of generative AI pilots and PoCs at 95%.
“Failures typically originate upstream,” Jena said, citing poor data quality, inconsistent definitions and unclear ownership as the primary causes. Model development, by contrast, is rarely the first point of failure.
From an industry-wide vantage point, the picture looks slightly more optimistic, particularly for the largest firms.
Biswajeet Mahapatra, principal analyst at research firm Forrester, said tier-1 Indian IT companies are largely ready to deliver production-grade AI systems, especially in regulated sectors such as banking, financial services, and telecom. Their advantage is flanked by established cloud-native architectures, mature governance frameworks, and experience in large-scale data engineering.
Meanwhile, tier-2 firms, he added, show uneven readiness, often excelling in niche engagements but lacking the enterprise-grade consistency needed to scale beyond pilots.
Mahapatra described the prevailing approach to data platforms as “hybrid”. Firms typically build internal accelerators and reference architectures while partnering with vendors such as Databricks, Snowflake, and Confluent for specialised capabilities.
Indian IT firms rely on selective acquisitionsto strengthen data engineering or AI operations, particularly among tier-1 players, while smaller firms rely more heavily on pre-integrated partner solutions, he said. These acquisitions plug data gaps, strengthening cloud data platforms, embedded analytics, governance layers and AI-native engineering depth needed to move enterprise AI beyond pilots into production.
This is exactly what TCS tried to accomplish with the $700-million acquisition of Coastal Cloud, a US tech consulting firm specialised in Salesforce solutions. This is a faster way for TCS to gain deep workflow expertise and talent required to tap into AI opportunities.
Experts flagged that ownership of client data platforms remains limited. Indian IT companies still primarily act as system integrators and managed service providers, executing architectures defined by hyperscalers and clients.
Full end-to-end custodianship is rare, although Mahapatra noted that long-term managed services and outcome-based contracts are gradually expanding the scope of operational responsibility for large vendors.
Talent Bottleneck
Talent depth is another concern. Datta and Jena both point to a shortage of elite senior data platform architects capable of designing and running enterprise-scale systems over long periods, even as tool-level and model-centric skills are widely available.
Indian IT services companies are training at an unprecedented scale to prepare their workforces for an AI-led future. Yet even as hundreds of thousands of employees are reskilled in data and generative AI, the sector continues to chase a much smaller pool of specialised talent—data architects, machine learning engineers, and AI platform builders—whose skills cannot be created quickly or cheaply.
Mahapatra agreed that tier-1 firms have strong senior talent pools, but demand far exceeds supply, while tier-2 firms remain more focused on implementation-level skills.
Datta expected hybrid platform orchestrators to emerge, with large firms combining proprietary IP-led platforms and long-term operations while remaining tied to hyperscalers and client sovereignty.
Jena anticipated a bifurcation, where a small group evolves into long-term custodians of enterprise data and AI platforms, while many others continue as project-based implementers.
Meanwhile, Mahapatra said that tier-1 firms will deepen custodial roles through managed platforms and outcome-linked service models, leaving tier-2 firms at a crossroads unless they invest significantly in governance authority and platform operations.
The post Indian IT’s AI Moment Runs Into a Data Readiness Check appeared first on Analytics India Magazine.
