AI in Microsoft 365: How to Make Your Data AI-Ready and Which Microsoft 365 Settings Are Crucial for Success (Part 2)

AI in Microsoft 365 sounds promising. Copilot can assist with writing texts, summarizing documents, and quickly finding information. But one crucial aspect is often overlooked: Copilot is only as smart as the data it runs on. If your information management isn’t in order, AI won’t just deliver poor results — it can even introduce risks.

Real-World Example

I experienced this firsthand. While searching for information about a Citrix implementation for a client, I came across documentation that didn’t match the current situation at all. Upon closer inspection, it turned out the system had surfaced a report written by an intern over 15 years ago.

This perfectly illustrates what can happen when outdated or irrelevant documents linger in your environment. Copilot will still retrieve them, potentially leading to confusion or even incorrect decisions.

The Key Building Blocks of AI-Ready Data

To ensure Copilot provides reliable and up-to-date answers, focus on these four pillars:

Data Classification

With Microsoft Purview, you can automatically label files—for example, when they contain a citizen service number (BSN) or salary data. This removes the burden from users and keeps sensitive information protected.

Authorizations

Apply the principle of least privilege—only grant access to those who truly need it. This reduces the risk of someone accessing information via Copilot that wasn’t intended for them.

Lifecycle Management

Remove or archive outdated documents so only current information remains accessible. This can be automated: set retention periods so documents are archived after X time and deleted after Y years. This prevents old intern reports or draft documents from influencing AI results.

Structure

Organize data into logical SharePoint sites and Teams channels, and use metadata so Copilot better understands where information belongs. This significantly improves the quality of search results.

Crucial Microsoft 365 Settings

In addition to data hygiene, certain Microsoft 365 settings form the foundation for safe and effective AI use:

• Entra ID (Access Management) – Ensure permissions and groups are tightly configured.
• Microsoft Purview (Classification & Labels) – Automate labeling of sensitive information, such as reports containing BSNs.
• DLP Policies (Data Loss Prevention) – Set rules to prevent data leaks, e.g., blocking accidental emails with BSNs.
• Tenant Settings – Review which AI features are active and align them with your organization’s policies.

What Can Go Wrong Without This Foundation?

Without these basics, you risk:

• Data Leaks: An employee accesses HR documents via Copilot that were never deleted.
• Incorrect Insights: Reports are based on outdated drafts instead of current data.
• Loss of Trust: Employees abandon AI tools after a few bad experiences and revert to old habits.

Microsoft Already Offers Tools

Fortunately, you don’t have to do this alone. Microsoft Purview provides a platform to centrally manage classification, labels, retention periods, and access rights. By using Purview effectively, you ensure sensitive data is automatically protected and Copilot only works with relevant, current information.

What Can You Do Today?

Don’t wait until everything is fully set up. Here are practical steps you can take right now:

Inventory Your Data

• Where are your key documents stored?
• How many outdated files are still in SharePoint or Teams?

Check Access Rights

• Who has access to which folders?
• Are there groups with excessive permissions?

Make Agreements About Storage

• Decide where new documents should be stored (Teams, SharePoint).
• Use clear naming conventions.

Raise Awareness

• Explain to colleagues why data quality matters for AI.
• Start small: pilot with one team or department.