With the recent announcement of 9 new Generally available built-in ready-to-use Business trainable Classifiers inside Microsoft 365, I wanted to see how they might make my E5 customers’ lives easier retaining content across their tenant.
Up to this point in time, customers could build their own custom trainable classifiers to detect business-like content across their tenant (contracts, tax forms, loan agreements, etc.); however, these 9 new business trainable classifiers really hit the mark for the “typical” kinds of content they may have across their tenant without having to build one of their own!
Here’s the 9 new classifiers:
These trainable classifiers can be used to:
- classify/recommend/automatically apply sensitivity labels to items opened in Office client apps
- automatically classify content and apply retention labels (what this post is about)
- detect matching content in Communication Compliance policies
Advantages of this trainable classifier model
- matching content can be automatically detected across your entire tenant and retention/sensitivity label applied if found
- you don’t have to use other conditions to detect content (E.g., a Contract content type on a library or a specific Site URL) to use as an auto-apply condition on the retention label
- if an end-user stores something in a site they aren’t supposed to, you can still apply a retention label if your auto-apply policy is scoped to all Sites, Microsoft 365 Groups, and OneDrives
- you can detect/monitor when matching content is found in a communication channel – e.g., someone sharing Intellectual property with external parties
Let’s check it out…
I’ll use a demo tenant to see if I can automatically detect agreements as defined by the ‘Agreements’ trainable classifier and apply a retention label if found!
Here’s the types of content the Agreements trainable classifier will look for:
Detects content related to legal agreements such as Non-Disclosure Agreements, statements of work, loan and lease agreements, employment and non-compete agreements
Seeding some data
Since statements of work (SOWs) are included in the Agreements trainable classifier, I’ll randomly seed some actual SOWs from my own company across several SharePoint sites to see if it will automatically apply a retention label when found.
- Sample SharePoint Site A
- Created new ‘SOW Library’ and placed both Quote documents and SOWs in it
- Created new ‘Regular Library’ and placed regular documents (non-agreements) and 2 SOWs in it
- Added some SOWs into the default Documents library
- Sample SharePoint Site B
- Added some SOWs into the default Documents library
Retention Label and Policy Setup
For this test, I created a new Retention Label, Agreement, and an auto-apply label policy called Agreement Label Policy statically scoped to ALL SharePoint sites and Microsoft 365 Groups . The label policy was configured to apply the Agreement retention label to all content matching the Agreements trainable classifier:
The Result
An auto-apply retention label policy can take up to 7 days to apply the retention label; however, in this demo, the next morning the labels were applied to all matching content!
4 separate libraries across 2 SharePoint sites with SOWs stored in them had the Agreement retention label auto-applied by the retention label policy. Documents that weren’t detected as matches by the Agreements trainable classifier, did not.
Monitoring
Now that the trainable classifier has been used as a condition in an auto-apply retention label policy, I can see it in Content Explorer within the Data Classification feature. You can see that the 19 documents labeled above are now showing beside the Agreements trainable classifier:
You can see the individual Label applied activity within Activity explorer as well:
Closing thoughts
I think these nine built-in business trainable classifiers are a fantastic addition to allow you to automatically protect, retain, and monitor content across your tenant in an intelligent manner. This is another example of the value an E5 Compliance license provides in managing compliance at scale with no end-user intervention.
Thanks for reading.
-JCK