Share with your CDO
Eighty-nine percent of APAC data leaders say unreliable data is blocking AI pilots from reaching production, according to a survey cited by Richard Scott, SVP APJ at Informatica by Salesforce. His argument: the bottleneck isn’t the AI model, it’s the data layer underneath it. The fix he proposes is headless data management, which strips the user interface away from data governance, quality, and metadata controls so that AI agents can invoke those capabilities directly inside workflows, without a human in the loop.
What this means for your business
The 89% figure is the tell. If nearly every data leader in the region is naming the same bottleneck, the problem isn’t organizational immaturity, it’s architectural. Enterprises that built their data stacks for human analysts, where governance lives inside a dashboard and quality checks require someone to log in and act, are now running those stacks as inputs to AI agents that can’t click buttons. Whether that describes your organization depends on one diagnostic question: can an AI agent in your current environment access clean, governed, contextualized data without a human preparing it first? If the answer requires any hesitation, this story is about you.
The headless framing borrows from a pattern that already worked in web architecture. Content management systems decoupled their content from their front-end presentation layer so the same content could flow to a website, a mobile app, and a third-party integration simultaneously. Headless data management applies the identical logic: decouple governance and quality controls from their dashboards so the same trusted data layer can serve a BI analyst, an autonomous agent, and an external API call at the same time. The architecture isn’t novel. What’s new is the urgency, because agents don’t wait for a data steward to approve a record before acting on it.
Scott writes for Informatica by Salesforce, a vendor with an obvious interest in organizations treating data infrastructure as a strategic priority rather than a back-office cost, which nudges his timeline toward urgency and his diagnosis toward solutions that require a modern data platform. That tilt doesn’t invalidate the argument. The 88% figure on governance gaps compounding across successive AI projects points to a real cost accumulation dynamic: each stalled pilot doesn’t reset to zero, it leaves behind a larger debt of unresolved data quality issues that the next project inherits. The CDO who treats this renewal cycle as a chance to renegotiate a data platform contract should be asking vendors specifically whether their governance controls are callable via API, not just visible in a UI. That’s the question that separates a platform built for the agentic era from one that’s been relabeled for it.
Concept deep-dive: Headless architecture
Headless architecture separates a system’s capabilities from its visual interface, exposing those capabilities as API calls, meaning any application or automated process can use them directly without opening a screen. Think of it as a kitchen that serves any restaurant in the building, not just the one it was originally built for. In a data context, this means governance rules, quality checks, and metadata lookups become functions that an AI agent can call mid-workflow, rather than controls that only activate when a human opens a data management console.
Based on reporting from How headless data services will power AI in APAC, originally published 2026-06-29 02:25:00.
