Share with your CDO
A quarter of Indian enterprises are already running agentic AI, the kind that takes autonomous action on loan approvals, vendor reconciliations, and customer triage, not just generating text for a human to review. Gaurav Pathak of Informatica makes the case that the constraint isn’t model quality or deployment tooling: it’s whether the data those agents consume carries verified governance, consent status, and lineage in real time. With India’s DPDP Rules setting a May 2027 deadline and the RBI’s FREE-AI framework already requiring explainable, auditable AI decisions in financial services, the window to retrofit data infrastructure is closing.
What this means for your business
The organizations most exposed here aren’t the ones that haven’t deployed agents yet. They’re the ones that have. Every enterprise that moved fast to put an agent in production inherited whatever data quality and governance posture existed before, which EY’s survey suggests is patchy for 78% of Indian enterprises still untangling basic system integration. If your data layer was built around a human logging into a dashboard and catching errors before they propagated, it was never designed to serve an autonomous system making hundreds of decisions per hour without pausing for review.
Pathak’s concept of “headless data management,” making governance rules, quality checks, and consent verification callable services rather than features locked inside a platform someone has to open, is the right framing, even if Informatica, whose metadata and data catalog products sit squarely in this architecture, has an obvious commercial interest in accelerating that shift. The argument holds regardless of who’s selling it. An agent acting on personal data without real-time consent verification isn’t a latent compliance risk under the DPDP framework; it’s a live violation the moment the processing happens. That’s a binary condition, not a risk to be weighed against speed-to-market.
The leading indicator to watch is how your organization’s data governance tooling is licensed and deployed. If governance capabilities live entirely inside platforms that require human access to activate, they simply don’t exist from the agent’s perspective. The architectural decision this reframes isn’t whether to invest in better data quality; most CDOs already know they need it. It’s whether the current vendor stack can expose that quality as a programmatic service an agent can call, or whether a renegotiation is already overdue before May 2027 makes it urgent.
Concept deep-dive: Model Context Protocol
Model Context Protocol (MCP) is an emerging open standard that gives AI agents a consistent way to discover and call external data services, roughly analogous to how REST APIs let any web application pull product data from a catalog without rebuilding the catalog for each new app. Without something like MCP, every new agent deployment requires custom integration code to reach governed data sources. With it, governance and quality controls travel as accessible services rather than being rebuilt from scratch each time an agent is added.
Based on reporting from The real bottleneck in India’s agentic AI race isn’t the AI, originally published 2026-07-30 08:14:00.
