ARTICLE
Date: 30/07/2026
Unstructured Data: The Missing Piece in Your AI Strategy
Most businesses have invested heavily in structured data. The data that actually captures what happens in their business every day lives somewhere else entirely.
What Unstructured Data Actually Is
Structured data lives in databases and systems: rows, columns, fields, records. It is designed to be queried, reported on, and governed. Most enterprise IT investment over the past two decades has gone into building and maintaining it.
Unstructured data is everything else. The photo a technician takes on site, the PDF that comes back from a contractor, an email that documents an agreed change, or an Excel that captures the latest pricing. In the best cases these files are stored somewhere accessible and loosely connected to the records they relate to. In practice, that connection is rarely reliable or complete.
In most businesses, unstructured data significantly outweighs structured data in volume. A photo, a field report, or a signed document can carry information that is just as critical to a business decision as any record in a database.
Why Structured Data Rarely Tells The Whole Story
The natural response is: if unstructured data is so valuable, why not structure it? The answer is that in practice, structuring information takes time, it can be expensive, and often degrades the very thing that makes it valuable.
Consider what it takes to capture a site observation as structured data. Someone has to define the schema, build the form, map the fields to the business objects, and maintain the model as the business changes. A photo of a corroded fitting takes two seconds. In practice, people capture what they can and move on, which is why common data values in structured systems so often end up as "Other" while the key details of the job are in the photo and the description.
Beyond the practicality, some information simply cannot be reduced to a field. A condition rating of 6 out of 10 tells you a number. A photo tells you what you are actually looking at. The job that was completed but not quite as expected, the delivery that arrived with damage, the asset inspection that found something the system had not anticipated (in every case the structured record captures what was intended, while the documents, photos, and notes created at the moment it happened capture what actually occurred).
Where That Reality Lives
Unstructured data does not live in one place. It accumulates wherever work happens: SharePoint folders organised by project or date, network drives named by team or year, cloud storage buckets, email inboxes, mobile devices, third-party portals used by contractors or suppliers. It is created by many different people, in many different formats, stored in many different places, with little consistency in naming, structure, or organisation.
The challenge is not that this data does not exist. It is that it exists everywhere at once, with no consistent way to find it, query it, or connect it to the business records it belongs to.
Why This Matters For AI
An AI working from structured data alone is working from an incomplete picture. It knows what was planned, approved, and recorded. It does not know what actually happened, what was observed on site, what was agreed informally, or what the current state of something looks like right now.
When the AI encounters a question it cannot answer from structured data alone, it does not say it does not know. It constructs the most plausible answer it can from whatever it can find. The answer looks confident. It is often wrong. And because the unstructured data that would have corrected it was never connected to anything, there is no way for the system to catch the error before it reaches someone who acts on it.
What Changes When It Is Connected
Connecting unstructured data to the business records it belongs to does not require replacing systems or migrating data. The files stay where they are. What changes is the intelligence layer that reads and understands them, maps each one to the business entities it relates to, and makes them findable and usable alongside the structured data they complement.
A photo becomes more than a jpeg. It becomes the current condition record for a specific asset, taken during a specific inspection, linked to the work order and the customer account it relates to. A field report becomes more than a PDF. It becomes the operational record of what actually happened on a job, connected to every system that needs to know about it.