Salesforce’s latest State of Data and Analytics report delivers a clear message from India’s technology leadership: AI ambitions will fail without modern, trusted data foundations. While 89% of India’s data and analytics leaders say their organisations must modernise data strategies for AI to deliver meaningful impact, incomplete, outdated and poor-quality data remains the number one barrier—especially as enterprises move into the agentic AI era.
Data-Driven in Name, Not in Practice
Nearly two-thirds of business leaders (66%) describe their organisations as data-driven, but 52% of data and analytics leaders say their companies still struggle to drive business priorities with data, exposing a gap between perception and reality. Only about half (51%) of business leaders can reliably generate timely insights, and nearly 54% of data leaders say their organisations occasionally or frequently draw incorrect conclusions due to poor business context.
The report stresses that AI cannot compensate for weak data. Despite AI ranking as the top data priority for Indian respondents for a second consecutive year, leaders estimate that around 25% of their organisational data is untrustworthy. This “data paradox” means AI initiatives are being layered on top of fragile foundations, increasing the risk of flawed decisions.
AI as Stress Test for Data Foundations
As enterprises race to adopt AI, existing data architectures are being stress-tested. Fifty-six percent of India’s data and analytics leaders feel pressure to implement AI quickly, yet 39% lack full confidence in the accuracy and relevance of AI outputs because they are driven by disconnected and outdated data.
The consequences are already visible: 94% of data and analytics leaders with AI in production report experiencing inaccurate or misleading AI outputs, and half of those training or fine-tuning their own models say they have wasted significant resources doing so with bad data. The report frames this as a governance and architecture problem, not just a tooling issue.
The Cost of Trapped and Siloed Data
Even when data is high quality, it often remains trapped. Leaders estimate that 26% of their company’s data is siloed, inaccessible or unusable, and 75% believe their most valuable insights sit inside this inaccessible quarter.
This fragmentation has broad impact: more than eight in ten data and analytics leaders cite reduced AI capabilities, obscured customer views, weaker personalisation and missed revenue opportunities due to trapped data. As a result, 89% say unified data is critical for meeting customer expectations, but many are still early in their unification journey.
Zero Copy and Agentic Analytics as Enablers
To break data out of silos, 52% of organisations are adopting zero copy data integration—accessing data across multiple databases without moving or duplicating it. Indian companies using zero copy are 40% more likely to have fully connected customer data sources and 22% more likely to succeed with AI initiatives than those without it, indicating that architectural choices directly influence AI outcomes.
On the consumption side, natural language–driven “agentic analytics” is emerging as a way to overcome data literacy bottlenecks. Sixty-nine percent of data leaders say translating business questions into technical queries is error-prone, while 95% of business leaders believe they would perform better if they could ask data questions in natural language. Agentic analytics, as embodied in new-generation BI platforms, aims to embed AI agents directly into analytics workflows, turning data into contextual insights inside the flow of work.
Governance Gaps in the Agentic Era
Only 52% of data and analytics leaders report having formal data governance frameworks and policies, even as 90% agree that AI demands entirely new approaches to governance and security.
Salesforce India’s leadership stresses that organisations must treat data foundations as a unified, governed and contextual strategic asset if India is to unlock agentic AI at scale. In practice, this means establishing clear ownership, guardrails and quality standards for data feeding AI systems, as well as rethinking security and compliance models for AI agents that operate autonomously across applications and datasets.
