Now that AI is no longer an experimental endeavour but an operational imperative, organisations find themselves caught between the promise of transformative capabilities and the stark reality of implementation hurdles. The true bottleneck, as a recent report by Collibra in partnership with Ascend2 Research highlights, isn’t a lack of data, but a profound lack of confidence in it. This fundamental deficit, termed “fragmented governance,” is proving to be a growing liability, slowing down decision-making, stalling AI initiatives, and eroding trust across systems and teams.
The research, based on a survey of 458 data governance, risk, and compliance professionals, reveals a concerning landscape: 40% of professionals admit they don’t trust their data. The reasons are alarmingly consistent: fragmented systems, inconsistent definitions, opaque ownership, and manual processes simply cannot keep pace with the demands of modern business. The consequences are tangible – teams become overly cautious, leaders resort to gut instinct, and crucial AI models fail in production, effectively stifling innovation.
The report underscores that AI initiatives are particularly vulnerable to unreliable data. A significant 53% of professionals identify data reliability as a top challenge when building and deploying AI models. Furthermore, a mere 45% feel confident in the data used to mitigate AI hallucinations, pointing to a critical vulnerability at the heart of AI adoption. The issue is not just that bad data slows progress; it actively multiplies existing problems, complicating everything from compliance to risk assessment.
So, what distinguishes the organisations that are successfully navigating this complex terrain from those still grappling with uncertainty? The answer lies in a single, organising idea: Data Confidence. These high-confidence organisations, representing 60% of professionals surveyed who report strong trust in their data use, quality, and governance, exhibit measurably different strategies. They are over twice as likely to have unified data governance across all departments compared to their less confident peers (73% vs 33%), enabling more flexible, cross-functional collaboration and control without forcing data relocation or compromising team autonomy.
Breaking Down the Data Silos
One of the most persistent obstacles to data confidence is the pervasive issue of data silos. Despite years of investment in cloud platforms and data catalogs, 37% of professionals report that data silos actively prevent efficient data sharing across teams, compounded by a lack of visibility or discoverability (32%). A staggering 90% of professionals believe they would benefit from improved data sharing, highlighting this as a critical business imperative. Yet, remarkably, only 21% prioritise reducing information silos in their governance improvement plans.
High-confidence organisations tackle this effectively through five core strategies:
- Establishing shared definitions and metadata standards to ensure consistent understanding across departments.
- Implementing federated stewardship models, empowering teams with data domain ownership while adhering to enterprise standards.
- Automating data discovery and access controls to reduce manual effort and boost responsiveness.
- Governing data in place with system-agnostic frameworks, avoiding data duplication and maintaining flexibility.
- Tracking lineage and usage across systems to build trust, support compliance, and enhance contextual understanding for decision-making.
The Data Quality Rescue Mission
Beyond silos, poor data quality stands as the most urgent challenge to fix, cited by 62% of professionals as the top area needing improvement. It compromises everything from audits to AI models and customer retention. A significant 60% of respondents identified data quality as the biggest barrier to confident data use, outweighing visibility, access, or security combined. Alarmingly, 41% report that data quality issues directly impact the trustworthiness of shared data, meaning even when data is accessible, it’s often not used due to mistrust.
Data-confident organisations are addressing this head-on. They employ customised governance frameworks tailored to each department, which 57% of professionals say improves both adoption and impact. They also demonstrate significantly higher visibility into the health and reliability of their data (64% vs 20% for less confident peers), enabling faster anomaly detection and better tracking of changes. These organisations treat data quality as a measurable, accountable business-critical function, embedding governance into workflows and leveraging automation to drive consistency.
Establishing Data-Centric AI Governance
The report particularly emphasises the criticality of data confidence for AI. While the promise of AI is speed, efficiency, and competitive advantage, fragmented governance often prevents safe scaling of data and AI use cases. Organisations with mostly or fully automated data governance workflows report far greater confidence in the data used to deploy AI models (67% vs 26% compared to those with less automation). This gap underscores that automation is key to enabling governance at the speed of data, reducing human error, and improving consistency.
High-performing organisations rethink AI governance not as an audit activity, but as an active, embedded system managing every step of the AI lifecycle. They achieve this through five foundational governance capabilities:
- Cataloging every AI use case and dataset for traceability and reusability.
- Establishing live links between models, policies, and training data, instantly flagging or blocking risky model behaviours based on governance rules.
- Embedding lineage and business context into models from inception, ensuring transparency on how outcomes are generated.
- Automating visibility and controls across hybrid data environments, ensuring governance follows the data regardless of location or use.
- Creating shared governance workflows that integrate legal, compliance, technical, and business teams early in the process.
This “human-in-the-loop” governance ensures oversight at critical points like data labelling and model deployment, building trust among business users, technical teams, and executives alike.
From Compliance Burden to Business Edge
Compliance, once viewed as a mere checkbox, is rapidly becoming a competitive differentiator. However, 88% of governance professionals report that expanding regulations have made compliance harder, not easier. For AI, 45% of professionals cite ensuring compliance with growing regulations as a top challenge when building and deploying AI models, adding exponential complexity.
Forward-thinking organisations recognise that compliance isn’t just about survival; it’s a strategic framework that builds trust and accelerates decision-making. These organisations don’t just draft policies; they operationalise them by investing in infrastructure that makes compliance actionable, automated, and scalable. This includes unified data governance across platforms, live lineage and usage tracking, automated policy enforcement, embedded privacy-by-design, and cross-functional transparency.
Ultimately, the quest for data confidence represents a profound cultural shift. When organisations adopt a product mindset toward data, treating datasets with defined ownership, documentation, and embedded controls, they significantly improve usability and trust. Among high-confidence organisations, 59% have fully productised their data, nearly three times higher than their less confident peers.
The message is clear: in the age of AI, getting one’s “data house in order” is no longer optional. Unified governance transforms compliance-led teams from perceived bottlenecks into strategic enablers. By cultivating unwavering trust in their data, organisations can accelerate every data use case, ensuring that AI moves both fast and safely, thereby securing a definitive competitive advantage.
