For large companies, especially those in regulated fields like finance and insurance, managing data is a major focus. Many have spent years building data governance programs. These efforts often start with setting policies, categorizing data sources, creating data catalogs, and communicating what’s non-negotiable. But most of the attention has been on structured data—think data warehouses and relational databases. Now, with AI playing a bigger role, there’s a growing need to govern unstructured data used to train language models and power AI systems.

The Growing Role of Unstructured Data in AI

Unstructured data makes up the majority of information in many enterprises. Unlike structured data, which is organized in rows and columns, unstructured data includes documents, images, videos, emails, and more. AI has introduced new tools, like vector databases and large language models, that help extract meaning from these messy data sources. This shift pushes organizations to rethink their data governance approaches.

Experts highlight that controlling unstructured data is more complex. Ashish Mohindroo, from Nutanix, emphasizes asking key questions like who needs daily access and how to keep data safe from unauthorized use or accidental loss. These questions are straightforward for structured data but much harder for unstructured sources, which often contain sensitive information hidden deep within documents or files.

Challenges and Strategies for Unstructured Data Governance

Joanne Friedman, CEO of ReilAI, points out that organizations need to move beyond static access controls. Instead, they should adopt a more flexible approach based on governed autonomy, which involves contract-based safety measures. This means understanding the context of data, not just its content, and connecting assets intelligently rather than reacting to issues after they happen.

Managing unstructured data requires understanding the risk inside each piece of content. Amanda Levay from Redactable explains that sensitive details often hide in places people don’t review regularly. She advocates for controls that prevent risky documents from entering workflows and systems that flag files containing confidential information. Catching issues at the right moment is critical to avoiding data leaks or compliance breaches.

Defining rules for unstructured data is more difficult than for structured data. For example, rules for accessing financial transactions are clear-cut, but policies for contracts, health records, or other documents are far more complex. Friedman notes that establishing effective controls requires understanding the nuances and potential risks embedded in unstructured content.

Leveraging AI to Improve Data Governance

AI itself can be part of the solution. Advanced tools like large language models help organizations interpret and classify unstructured data at scale. These models can identify sensitive information, automate tagging, and enforce policies more efficiently. This reduces the risk of human error and speeds up compliance processes.

However, relying on AI also introduces new challenges. Ensuring that AI systems are transparent, secure, and aligned with governance policies is vital. Organizations must develop frameworks to monitor AI’s decisions and maintain control over how unstructured data is accessed and used.

Ultimately, managing unstructured data in AI environments requires a mix of technological tools, thoughtful policies, and continuous oversight. As data sources grow in volume and complexity, companies need adaptable strategies that prioritize safety, compliance, and accessibility. AI can be a powerful ally in this effort—if used wisely.

Inspired by

• https://www.infoworld.com/article/4160979/addressing-the-challenges-of-unstructured-data-governance-for-ai.html

Sources

• starcio.com
• starcio.com
• nutanix.com
• reilai.com
• collibra.com