Unstructured Data

Unstructured data is free-form information — documents, emails, chat logs, call transcripts, meeting notes, scanned forms — that doesn't fit into the predictable rows and columns of a database. Unlike structured data, where a field is explicitly labeled "SSN" or "date of birth," unstructured data contains the same categories of sensitive information without any label at all: a customer's Social Security number might appear mid-sentence in an email, a patient's diagnosis might be buried in a paragraph of clinical notes, an account number might be mentioned conversationally in a call transcript.

This matters enormously for AI specifically, because unstructured data is estimated to make up the large majority of enterprise data — commonly cited figures put it around 80 to 90 percent — and it's also exactly the kind of content AI tools are best at working with. Summarizing documents, searching across emails, transcribing and analyzing calls: these are core AI use cases precisely because AI excels at reasoning over free-form language in a way traditional software never could. That overlap means AI adoption concentrates risk directly onto the category of data organizations have historically protected the least.

A company might have genuinely strong security around its structured customer database: encrypted at rest, tightly access-controlled, with clearly labeled fields that make audits straightforward — a mature, well-protected system. But that same customer's sensitive information also exists, in parallel, across support email threads, sales call transcripts, Slack messages, meeting notes, and scanned PDF forms — all containing the same categories of personal and financial data, just without any of the structure that made the database easy to protect and audit.

When that company deploys an AI tool to summarize support tickets or search across historical emails, it's pointing AI directly at this unstructured pool — the exact data that was hardest to classify, locate, and protect in the first place. The database being secure doesn't help here at all; the sensitive information living in unstructured form was never covered by that protection to begin with.

Strong structured-data security can create a false sense of overall data protection maturity, because the unstructured half of an organization's data — often the larger half — carries the same underlying sensitive information without any of the same safeguards. This gap tends to go unnoticed precisely because it's harder to see: a database's contents are visible and auditable in a way that thousands of scattered documents, emails, and transcripts simply aren't.

Protecting a well-labeled database is the easier half of the problem. The harder, larger half is the free-form content sitting everywhere else — which is exactly where AI tools are being pointed.

Questa AI's entity-detection engine is specifically built to handle unstructured, natural-language content — documents, chat logs, call transcripts, emails — rather than only recognizing sensitive data in clean, structured database fields. This directly addresses the gap that regex-based and structured-focused tools leave open: sensitive information phrased conversationally, without a label, in the free-form content that makes up most of an organization's data.

Because this is precisely the content most AI copilots, chatbots, and summarization tools are built to process, closing this gap is central to making AI adoption safe rather than just convenient. Questa AI anonymizes sensitive information within unstructured content before it reaches an AI model, extending real protection to the majority of enterprise data that structured-data-focused security tools were never designed to cover.