Unstructured Data Makes Up 80% to 90% of Enterprise Content

Unstructured data—including emails, documents, PDFs, images, videos, chat logs, forms, reports, clinical notes and customer interactions—makes up 80% to 90% of enterprise content, according to TechRadar. The publication says this material, which sits outside structured databases, can provide context for enterprise AI agents.

TechRadar reports that fewer than half of AI projects in the past two years delivered measurable business outcomes. It argues that poor data foundations, fragmented information, weak governance and unclear objectives can limit the value of enterprise AI and agentic automation. The article cites the UK Public Accounts Committee as finding that poor-quality data trapped in legacy systems is a major barrier to AI.

Unlike assistive AI that responds to prompts, enterprise agents are designed to carry out tasks. The article says deploying them calls for audits of workflows and data, along with clear governance and human oversight. Without those foundations, agents could duplicate work, create bottlenecks or take unwanted actions.

As an example, TechRadar describes accounts payable work where agents could capture invoices, structure and validate their data, match them against purchase orders and goods, then post them to a finance system. The article says each step can be logged and that human judgment remains necessary for decisions where it changes outcomes.

TechRadar argues that organizations should first identify where automation can have an impact, refine existing data and build connected infrastructure. It says enterprise agents will shift from assistive tools to governed operators over the next 12 months.