SharePoint Is Not a File Share

The most common thing I find in SharePoint environments older than three years is a file share in disguise. Thousands of documents in flat folder hierarchies. No content types, no metadata. Libraries named “General” and “Archive” and “Old Stuff - DO NOT DELETE.” Guest access granted on request and never audited. Sites for projects that ended in 2019 still sitting in the tenant, full of content nobody owns and nobody has reviewed.

The technology isn’t the problem. SharePoint can absolutely support a disciplined, well-governed information environment. The problem is that SharePoint is easy to use as a file share, and using it that way feels productive right up until it doesn’t. The ceiling hits when search stops returning anything useful, when nobody can find the authoritative version of anything, and when Copilot shows up and makes every governance failure queryable.

Why the file share pattern is so persistent

Blame the folder metaphor. People understand folders; folders are how files have always been organised, and SharePoint makes them trivially easy to create. So most organisations reach for folders first and never develop the discipline for anything better.

Folders are single-axis classification. A document lives in exactly one folder, and finding it means knowing which one. If you don’t know, or the folder tree has grown past the point of navigability, the document is effectively lost unless search rescues it.

Metadata is the alternative. Give a document the right content type, the right managed metadata columns, the right values, and it becomes findable along any of those axes. The folder stops being the primary finding aid. Search works against the metadata. Views and filters work against the metadata. Copilot, when it arrives, uses the metadata as context.

But metadata demands upfront decisions folders don’t. Content types have to exist before the first document does. Values have to be populated, which takes discipline from everyone who uploads a file. The taxonomy has to be maintained. None of it is technically difficult; all of it is organisationally hard. So organisations skip it and use folders.

What it costs

The cost stays mostly invisible until something changes the stakes.

Search quality goes first. Enterprise SharePoint search is powerful, but it needs content to be discoverable through something beyond keywords in the document body. A library full of files named “Final v2 revised APPROVED.docx” with no metadata is a library search can’t usefully rank. Users learn search is unreliable and stop using it. They navigate by hand instead. That works until content volume makes manual navigation impossible too.

Governance goes second. When finding a document means knowing its folder, governance becomes folder governance, and folder governance is almost always informal and personal. The person who created the structure knows the logic. When they leave, the logic leaves with them.

Copilot goes third. Microsoft 365 Copilot uses SharePoint as its primary data source and retrieves content by relevance to the query. In a well-structured environment, it can surface the right information with high confidence. In a folder-based environment with no metadata and chaotic naming, relevance scoring degrades. Copilot still returns results; they just aren’t reliably the right ones. Users lose trust fast, and the bitter part is that the tool is working exactly as designed. The environment wasn’t designed for it.

What good information architecture actually is

In principle, SharePoint information architecture is simple: it’s the set of decisions that determine how content is structured, classified, and found. Content types define what categories of information exist and what metadata each requires. The term store provides a controlled vocabulary so classification stays consistent across the environment. Hub sites and site hierarchies mirror how the organisation’s work is actually structured, not the flat list of projects that happened to request a site.

These decisions belong before content creation, not after. Retrofitting IA onto an environment full of unclassified content is possible but expensive: a metadata migration touching every document, usually via ShareGate or a custom script, plus a content audit to decide what’s still relevant and what gets archived or deleted.

Unglamorous work. But do it, and your SharePoint environment gets more useful as content accumulates. Skip it, and the environment gets less useful every year, until the weight of unclassified content makes starting over cheaper than fixing it.

What this has to do with AI

Short version: Copilot’s quality ceiling is your information architecture quality. A better AI model cannot compensate for a badly structured SharePoint environment. The AI is only as good as the content it can reach and the structure that makes that content retrievable and trustworthy.

Treating SharePoint as a file share was never a technical decision. It was a strategic one, made by default instead of deliberately. The cost was always there. AI just made it visible.