Information Classification & AI: From Chaos to Control

Every day, housing authorities process enormous amounts of documents and data: lease agreements, maintenance reports, invoices, emails from tenants, and internal memos. In addition, there are also many external parties—suppliers, service providers, and municipalities—that provide information.

At the same time, tenants are becoming more demanding: one wants to receive their lease digitally, another wants an overview of maintenance work via the app, while someone else wants to receive information by mail, preferably in their own language. (Social) rental housing often consists of a mix of young people, older adults, and people from the Netherlands or from other countries. Each has their own level of familiarity with communication, but above all, they have different expectations.

The volume of information is increasing, and not just because of tenants’ requests. This increase is also the result of mergers; housing associations are getting bigger and bigger. So how do you manage this growing flow of information? Classification provides an important foundation for this. Information classification is the systematic organization of documents and data, which determines the level of protection required and who is authorized to access the information. It helps ensure that sensitive data is processed securely, maintains clarity, and clearly assigns responsibilities.

But is this classification actually feasible? After all, the combination of an aging workforce, increasing workloads due to digitization and sustainability efforts, and the nationwide labor shortage is making it increasingly challenging for housing agencies to find sufficient and qualified staff.

If information is misclassified, it can have serious consequences. Consider a document containing personal data that is accidentally accessible to everyone, an invoice that is processed incorrectly, or internal memos that aren’t properly protected: these are all examples of risks that can lead to data breaches, fines, or inefficiency. No time for classification? That means housing association employees may not be addressing issues that require extra attention.

AI can make a big difference when it comes to classifying information. In practice, this means, for example, that an AI model can analyze incoming documents and automatically recognize whether they are a lease agreement, an invoice, an internal memo, or a form of identification. The model can identify personal data or store documents in the correct location within the central system, such as an EIM system.

AI performs repetitive tasks faster and more consistently than humans, thereby minimizing errors, revealing connections between documents, and ensuring the quality of information. For example, an AI agent can act as your assistant, automatically categorizing incoming tenant inquiries and linking them to relevant documents. This allows housing authority staff to focus their attention on more complex issues or to free up time for coordinating with maintenance and service providers.

AI-powered classification makes processes run more smoothly and securely. Files are up-to-date and correctly classified, allowing for faster and higher-quality audits. For example, when a tenant calls with a question about their lease or previous correspondence, you can immediately locate all relevant documents, regardless of whether they were received digitally, as scans, or via email. This leads to faster responses, fewer risks, and more reliable service.

At OneXillium, we help housing associations with Enterprise Information Management systems that centralize data and keep it up to date. This prevents different versions of documents from circulating and lays a solid foundation for AI applications. The advantage of our Enterprise Information Management system is that we can anonymize the collected data. This makes it useful, for example, for understanding common questions, analyzing trends, or making improvements to service delivery—all without making a direct reference to any individual (GDPR-compliant).