When to Utilize Syntex?
Syntex is a solution that should be implemented when a client requires the following:
Auto-classification of documents in their SharePoint Online Environment
Auto- Metadata extraction from documents
Content assembly
Classification and extraction across multiple SharePoint sites in the tenant. With Syntex document understanding, one model can be deployed to various sites.
Approval processes where Teams notifications can be sent after a successful file upload into SharePoint
Clients who require content processing locally within their own tenant for security and regulation compliance
Benefits
Clients with Syntex will experience the following benefits:
Reduced manual labor and avoid human error by automatically tagging, labeling and organizing content
Improve the searchability of files
No code models that can be implemented OOB
Syntex integrates with Purview to enforce security and compliance with automatically applied sensitivity and retention labels to classified files
Enhanced document collaboration with annotation, redaction, and PDF management
Ai classification can help drive workflows and business process automation
Improved auditing and understanding of document types stored in SharePoint Online
Premium taxonomy services (SKOS term sets, pushing content types to hub sites)
Overview
Syntex can be accessed from the Microsoft Content Center. When creating new models, you can map to an existing content type or create a new content type for the given model. You must have Global admin or SharePoint admin permissions to be able to access the Microsoft 365 admin center and set up Syntex. To create a document understanding model you need to:
Add example files (5 of the same type and 1 negative example)
Train the classifier
Label the documents as positive or negative examples
Train and provide explanations for what distinguishes your classification
test your model against new documents
Syntex will provide a model accuracy score you can use for adjustment
After classifying the document, you can add extractors which are pieces of metadata Syntex can extract and store in columns
Label the examples
Provide explanations of how Syntex can find the metadata
Test against new files
Syntex will provide a model accuracy score you can use for adjustment
Finally, the model can be applied to any SharePoint document library.
Components
Content Center: A content center is a site template for creating, training, and managing Microsoft Syntex models. A Default site is created as part of Syntex setup and additional content centers can be created afterward. Models can be organized by groups (such as departments) and can be created locally in teams or communication sites.
Models
Models: Enable users to classify a file of a particular business type and extract specific entity information from it. There are several different model types in Microsoft Syntex including form processing, document understanding, and prebuilt models which include Invoice and receipt processing as of December 2022.
Enterprise Models: An enterprise model is created and trained in the content center and can be deployed to multiple sites and libraries across your tenant.
Local models: Microsoft Syntex provides an option to create and train models locally on your own SharePoint site. These models can be used ONLY on the site where they're created. To create a local model, go to a SharePoint document library, select the files you want to analyze, and then select the Classify and extract button.
Model Types: (for more details see here) Syntex has 3 model types:
Unstructured document processing
Freeform document processing (semi-structured)
Structured document processing (Layout)
Structured Document Processing = Layout and Freeform methods (Formerly form processing): For structured documents (forms, invoices, anything with a template/table). Applied locally to a library and leverage AI builder. Unlike models built in AI builder, these models will NOT automatically create power automate flows you can modify. These models rely on clicking and locations within the document to extract information. Layout - can extract multiple lines from a table Freeform - cannot extract multiple line items
Unstructured document processing = Training method (Formerly Document Understanding: Unstructured documents (statements of work, emails, anything without unique spacing/tables). These models need training and explanations to help Syntex understand them.
Licensing
To use Microsoft Syntex, you must have a license for each Syntex user. Microsoft Syntex is available only to customers currently licensed for Microsoft 365 F1, F3, E3, A3, E5, A5, Office 365 F3, E1, A1, E3, A3, E5, A5, Microsoft 365 Business Basic, Business Standard, Business Premium, or SharePoint Online K, Plan 1, or Plan 2. (Source) These apps must be assigned to users:
Syntex
Syntex - SPO type
Common Data Service for Syntex
Unstructured Document Processing: Require Ai builder credits. You need to allocate AI Builder credits to the custom environment before you can create document processing models. (Source) Structured Document Processing: For each Syntex license, you are allocated 3,500 AI Builder credits per license, per month pooled at the tenant level, with a maximum allocation of 1 million credits per month. The allocation renews each month and unused credits DO NOT carry over.(Source) Supported File Types Form Processing structured models support: JPG, PNG, or PDF format (text or scanned) are supported. Text-embedded PDFs are better because there will not be any errors in character extraction and location. (Source) Document Understanding unstructured models support: .csv, .doc, .docx, .eml, .heic, .heif, .htm, .html, .jpeg, .jpg, .md, .msg, .pdf, .png, .ppt, .pptx, .rtf, .tif, .tiff, .txt, .xls, and .xlsx. (Source)
Syntex Integrations with Power Automate
Syntex has multiple use cases, but most processes involve 1 of 2 core processes which are content AI or content assembly. Based on these universal processes Microsoft has deployed 1 Power Automate trigger and 1 action based around Syntex.
The trigger = When a file is classified by a Microsoft Syntex model
The action = Generate document using SharePoint Syntex (for content assembly)
After adding the trigger you just need to specify the site and library.
For further reading please see here: SharePoint Syntex - new support for full document automation scenarios with Power Automate
Deleting a Syntex Model
Remove the model from any applied libraries.
Select "model settings" in the top left corner.
Edit "sites where this model is applied" to " No sites" and then select the "save" button.
From the Content Center Models page select the model and then select the delete button.
If the model was power by AI builder ( freeform or layout) you'll also need to delete the model from AI builder.
For further reading please see here.
Resources and further reading
Syntex integration with Power Automate: https://www.sharepointnutsandbolts.com/2022/09/Syntex-Power-Automate-trigger-action.html
About Cadence Solutions
Jordan Uytterhagen founded Cadence Solutions starting on the client side of the table. His mandate has been to help organizations struggling with digital transformation implement projects without losing their trust and confidence. Our solutions include automation of human resources, finance, accounts payable, contract management, document capture, drawing and records management, as well as managed services. Cadence Solutions has proven, time and again, that our client's projects will be successful because we are authentic with unmatched experience.
Further information: Jordan Uytterhagen, Managing Director Cadence Solutions 1.833.454.0577 | jordan@cadencesolutions.ca