AI’s Hidden Bottleneck: Why Unstructured Data Is Derailing Enterprise Innovation

A growing number of initiatives are failing to progress beyond proof of concept despite substantial investment, executive sponsorship, and access to advanced models, even as enterprises accelerate AI adoption. The root cause is rarely algorithmic capability or budgetary commitment. Instead, it lies in a far more foundational challenge: the fragmentation, governance, and operational readiness of unstructured data. With most enterprise information now outside traditional structured databases, organizations face increasing difficulty in making data accessible, secure, compliant, and usable at scale for AI-driven systems.

In this exclusive SecurityInfoWatch Executive Q&A, Nick Burling, Chief Product Officer at Nasuni, examines why unstructured data has become the critical constraint on enterprise AI success and what leaders must do to address it. Burling is responsible for defining and executing Nasuni’s hybrid cloud storage and data services strategy, with a focus on enabling secure, scalable, and intelligent data environments for global enterprises. An accomplished technology executive and entrepreneur, he brings deep experience across the full product lifecycle, having built and led product organizations at Microsoft and IBM, and guided startups from launch through successful exits.

In this conversation, Burling outlines the structural, operational, and security barriers preventing AI from reaching production and explains how modern unstructured data management strategies can transform AI ambition into measurable business outcomes.

SecurityInfoWatch: Why are so many enterprise AI initiatives failing before they ever reach production, despite significant investment and executive sponsorship? 

Nick Burling: Technology leaders are being asked to balance two priorities that often compete with one another. On one hand, boards and executive teams are pressing for sustained reductions in infrastructure costs, without taking on additional operational or security risk. On the other hand, business leaders expect artificial intelligence initiatives to move beyond proof-of-concept phases and deliver tangible, repeatable results at scale.

Without a unified approach to unstructured data, enterprise AI efforts remain constrained, regardless of executive sponsorship or budget allocation.

SIW: Why has unstructured data become the primary bottleneck for enterprise AI, and why is this problem getting worse?

Burling: Unstructured data has become the primary bottleneck for enterprise AI because it is increasingly dispersed, harder to govern, and more difficult to operationalize at scale. Most enterprises now operate across multi-cloud environments that span on-premises infrastructure, public cloud platforms, SaaS applications, and edge locations. While this architecture enables flexibility, it fragments unstructured data across disconnected systems, limiting the visibility and consistency that AI initiatives depend on.

In many organizations, unstructured data lives across file servers and legacy NAS systems, public cloud storage and object stores, SaaS platforms, and regional environments. Each environment relies on its own management tools, security models, and access controls. This lack of standardization makes it difficult to answer fundamental questions about where data resides, who can access it, and whether it is suitable or compliant for AI training or inference. As AI initiatives scale, these gaps become increasingly difficult to manage.

This fragmentation slows AI pipelines at every stage. Data teams spend excessive time locating relevant data, validating it, moving files between environments, and cleaning or reformatting content before models can be trained or deployed. Inconsistent metadata and governance introduce friction that limits scalability and undermines trust in outcomes.

The problem continues to worsen as data volumes grow, and environments become more distributed. Without centralized visibility and control, enterprises are forced to choose between speed, security, and compliance. These trade-offs slow innovation and reinforce unstructured data as a persistent constraint on AI progress.

SIW: How are storage sprawl and backup duplication contributing to rising infrastructure costs that are often invisible to executives?

Burling: Storage sprawl and backup duplication are major contributors to rising infrastructure costs because they quietly increase capacity, complexity, and risk without clear visibility to executives. As unstructured data spreads across on-premises systems, public cloud platforms, and edge environments, organizations routinely overprovision storage to avoid disruption, even as utilization remains uneven and inefficient.

These costs are often hidden within operational budgets rather than surfaced as strategic issues. Common cost drivers include storage sprawl across multiple platforms and regions, backup duplication that creates multiple full copies of the same data, and siloed recovery systems that increase complexity without improving usability. Over time, these layers accumulate, inflating infrastructure spend while obscuring the true cost of data growth.

Ransomware preparedness further accelerates this problem. A 2025 industry survey found that 57% of organizations experienced a ransomware attack in the past 12 months, with nearly a third suffering repeat incidents that affected file availability. In response, many organizations add additional backup and recovery tools, increasing duplication rather than improving resilience. These decisions carry financial consequences that can also erode brand trust.

Improving recovery times requires a different approach. Platforms that capture frequent, immutable file versions provide a stronger foundation for resilience. Maintaining a continuous history of file changes allows organizations to restore clean data states quickly and confidently. When combined with integration into enterprise security monitoring tools, teams gain broader visibility across file environments, enabling earlier detection and more effective automated responses while reducing unnecessary infrastructure duplication.

SIW: What role does cybersecurity, particularly ransomware resilience, play in the AI data readiness conversation?

Burling: Cybersecurity, particularly ransomware resilience, is a critical part of the AI data readiness conversation because AI depends on secure, trusted, and consistently governed data. As unstructured data continues to grow and spread across environments, security gaps increasingly determine whether data can be safely analyzed and used to drive reliable outcomes. Without strong security foundations, organizations are forced to limit AI initiatives to reduce risk, slowing progress before projects ever reach production.

Ransomware has amplified this challenge. Attacks that disrupt access to file data directly impact AI pipelines and undermine confidence in underlying datasets. In response, many organizations add layers of security and backup tools, but without a unified strategy, this often increases complexity and duplication rather than improving resilience. Fragmented environments make it harder to analyze activity, detect anomalies, and respond quickly to incidents.

A unified path forward requires aligning storage, security, governance, and AI enablement into a single data management approach. When unstructured data is clearly organized and governed, security teams gain the visibility they need to analyze behavior, identify threats earlier, and take more efficient action. At the same time, well-structured data makes AI systems more effective by ensuring models are trained and run on clean, reliable inputs.

By modernizing unstructured data management, organizations can reduce infrastructure sprawl, strengthen ransomware resilience, and make data usable for AI at scale. As enterprises move into 2026, AI success will be determined less by model selection and more by the readiness of secure, well-governed data.

SIW: What does it take to make unstructured data “AI-ready” in a large enterprise environment?

Burling: Making unstructured data “AI-ready” in a large enterprise environment starts with consolidating and organizing it so it can be consistently accessed, governed, and analyzed. AI initiatives cannot scale when data remains fragmented across silos with inconsistent metadata, permissions, and formats. Before advanced analytics or machine learning can deliver value, unstructured data must be unified into a coherent foundation.

Once consolidated, unstructured data becomes usable for analytics and machine learning by enabling consistent indexing, governance, and access control. This foundation supports a range of AI-driven use cases that rely on accurate, trusted inputs rather than isolated data sets. Institutional knowledge discovery becomes more effective when teams can search across historical files to accelerate research and decision-making. Automated document analysis helps legal, compliance, and professional services teams extract insights while maintaining control over sensitive content. In manufacturing environments, predictive maintenance and quality analysis depend on the ability to analyze large volumes of files generated across systems and locations.

Secure generative AI data requires that existing access controls and governance policies are enforced consistently, ensuring users only see and use data they are authorized to access. This balance between accessibility and protection is essential in large enterprises.

Ultimately, making unstructured data AI-ready is less about transforming the data itself and more about transforming how it is managed. When data is consolidated, organized, and governed at scale, organizations can move AI initiatives from experimentation to production with greater confidence and efficiency.

SIW: As enterprises plan for 2026, what strategic shift should executives make to avoid becoming part of Gartner’s 60% failure forecast?

Burling: Gartner’s forecast highlights a growing disconnect between AI ambition and data readiness. At the core of this challenge is unstructured data. Gartner estimates that up to 80% of enterprise data is unstructured, and in most organizations, much of it remains fragmented across silos. When this data is poorly prepared for advanced analytics, AI initiatives struggle to move beyond experimentation and into production.

This reality forces executives to confront a critical question: Is our data actually AI-ready? While organizations increasingly recognize the potential value locked inside their unstructured data, they often lack the visibility, governance, and control required to operationalize it at scale. As a result, unstructured data has become a primary reason AI projects stall or are ultimately abandoned.

As enterprises continue to plan for this year, the strategic shift executives must make is moving away from treating AI as a standalone technology initiative and toward treating data readiness as a foundational business priority. Unstructured data is expanding faster than any other data type, placing a growing strain on storage environments. While this data consumes an increasing share of storage budgets and security resources, it also forms the foundation on which modern analytics, machine learning, and generative AI depend.

Executives who take a holistic approach to unstructured data management can reduce risk, control costs, and create a clear path for AI initiatives to progress from pilot to production. Those who do not will continue to face rising infrastructure complexity while AI ambitions remain out of reach.