Fix Microsoft 365 Document Processing Issues

This is where almost everyone starts when Microsoft 365 document processing isn't working. Without an active Azure billing link, every single document processing service, autofill columns, OCR, document translation, eSignature, content assembly, you name it, will be completely unavailable to your users.

Sign in to admin.microsoft.com with a Global Administrator or SharePoint Administrator account. In the left navigation panel, go to Setup, then scroll down to the Files and content section. Click Use content AI with Microsoft Syntex (the admin center may still show the Syntex name in certain builds, this is the same service). On the next screen, click Set up pay-as-you-go.

You'll be prompted to select or create an Azure subscription. If your organization already has Azure, pick the appropriate subscription from the dropdown. If not, you'll need to create one, you'll need an Azure account with at least Contributor-level access on the subscription. Select the Azure region that matches your Microsoft 365 data residency (for most worldwide tenants this is fine to leave at the suggested default). Accept the terms of service and click Save.

After saving, wait about 10–15 minutes. Document processing service availability propagates across SharePoint asynchronously, it's not instant. Then navigate to a SharePoint document library, go to Library settings > Document processing (under the gear icon) and confirm the option is now present and not grayed out.

# You can verify billing setup is active via SharePoint Admin Center PowerShell
# Connect first:
Connect-SPOService -Url https://yourtenant-admin.sharepoint.com

# Check Syntex/Document Processing settings:
Get-SPOTenant | Select-Object SyntexEnabled, AIBuilderEnabled

If the property returns True for SyntexEnabled, your billing link is live and services should be accessible across the tenant.

Here's something a lot of people miss: setting up pay-as-you-go billing doesn't automatically turn on every document processing service. You have to deliberately enable the ones your organization wants to use. This is by design, Microsoft doesn't want you accidentally incurring charges for services you didn't ask for.

Go back to the Microsoft 365 admin center. Navigate to Settings > Org settings > Services and select Document processing. You'll see a list of all available services. The ones you can enable include: Autofill columns, Document translation, eSignature, Optical character recognition, Content assembly, Image tagging, Taxonomy tagging, Prebuilt document processing, Structured and freeform document processing, and Unstructured document processing.

Toggle on only the services your organization actually plans to use. If your team wants to automatically extract data from invoices and receipts, enable Prebuilt document processing. If you need to automatically translate documents in SharePoint libraries for multilingual teams, enable Document translation. If your legal team wants digital signatures without leaving Microsoft 365, enable eSignature.

After enabling services, SharePoint site owners and library owners will gain access to configure those services on their specific libraries. The controls appear in the library settings under Automate or under the gear menu, depending on the service type.

One common mistake: enabling Image tagging and then being surprised when it runs on every new file upload automatically. Review each service's behavior before enabling it broadly, some are passive and need to be applied manually; others can run continuously on a library once configured.

# To check which services are enabled on a specific SharePoint site:
Get-SPOSite -Identity https://yourtenant.sharepoint.com/sites/yoursite | 
  Select-Object Url, SyntexEnabled

Autofill columns and content assembly are two of the most-used Microsoft 365 document processing features, and they have their own quirks when things go wrong. I want to walk through both because the failure modes are different.

Autofill columns not working: This feature uses large language models (LLMs) to extract or generate metadata automatically when files land in your library. The most common failure is a mismatch between the column type you've configured and the data the LLM is trying to extract. If you've set up an autofill column that should pull "Invoice Total" from PDFs but it keeps returning blank or wrong values, check two things: first, verify the column is of the right data type (currency, not single line of text, for monetary values). Second, review your prompt configuration, go to the library, click the column name, select Column settings > Edit, and look at the Autofill section to review or refine the extraction prompt.

Content assembly not generating documents: Content assembly is built for automating standard, repetitive business documents, contracts, statements of work, letters of consent. If document generation fails, the most common culprits are template misconfiguration and missing placeholder fields. Navigate to the content center site where your template lives. Go to Models > Content assembly, find your template, and click Edit template. Verify all placeholder fields are correctly mapped to SharePoint columns or form fields. Any unmapped placeholder will cause the generation to fail silently.

Also check that the site collection where your content center is hosted has document processing enabled, it's not inherited automatically from the tenant-level setting.

# PowerShell to verify a site collection has the content center feature active:
Get-SPOSite -Identity https://yourtenant.sharepoint.com/sites/contentcenter |
  Select-Object Url, Template

A content center site should show CONTENTCTR#0 as its template. If it shows a standard team site template, you haven't actually created a content center, you've just created a regular site and need to either convert it or create a new content center from the SharePoint admin center under Active sites > Create.

These two services trip people up during initial setup more than during day-to-day use. Let me cover the specific gotchas I see repeatedly.

Document translation not appearing in library menus: Document translation creates a translated copy of a file in the same SharePoint document library, preserving the original format and structure. If the option doesn't appear when you right-click a file or check the command bar, first confirm the service is enabled at the tenant level (Step 2 above). Then check file type, document translation works on Word, PowerPoint, PDF, and a range of other formats, but it won't work on files it can't parse (corrupted files, zero-byte files, or unsupported formats). The service supports all standard Microsoft 365-supported languages and dialects, so language isn't typically the issue.

If translation is enabled but the translated copy comes back with formatting errors, that's usually a complex layout issue. Tables with merged cells, embedded objects, and heavily formatted PDFs can lose fidelity in translation. The original file is always preserved, so nothing is lost, you just may need to do minor formatting cleanup on the translated output.

eSignature requests failing to send: eSignature keeps your content inside Microsoft 365 while it's being reviewed and signed, which is the whole point for compliance-sensitive teams. The most frequent setup error is permissions. The person initiating the eSignature request must have at minimum Contribute permissions on the library where the document lives. Read permissions are not sufficient to initiate a request.

Also verify that external sharing is permitted on the SharePoint site if signers are outside your organization. Go to SharePoint admin center > Active sites, select the site, click Policies > External sharing, and set it to at least New and existing guests. Without this, external signers will receive the email but hit an access error when they try to open the document.

# Check external sharing setting on a site via PowerShell:
Get-SPOSite -Identity https://yourtenant.sharepoint.com/sites/yoursite |
  Select-Object Url, SharingCapability

This is where the complexity really picks up, and where I see enterprise IT teams spend the most time troubleshooting. These three model types are the core of Microsoft 365's intelligent document processing, they're what actually reads and classifies your documents.

Prebuilt models are ready-made for common document types like contracts, invoices, and receipts. If a prebuilt model isn't extracting fields correctly, the most likely cause is that your documents don't match the expected schema for that prebuilt type. For example, the invoice prebuilt model expects certain standard invoice fields. If your vendor invoices use highly non-standard layouts, switch to a structured or freeform model instead where you can train it on your specific documents.

Structured and freeform models are for documents where information can appear anywhere. If you created a model but it's not being applied to new uploads in your library, check the model's application settings: go to the content center, select your model, click Where the model is applied, and confirm your target library is listed. Models don't apply retroactively by default, you need to manually run them on existing files or enable continuous processing.

Unstructured models classify documents that vary in composition. These are the most powerful but also the most sensitive to training data quality. If your unstructured model is misclassifying documents, you likely need more training examples, Microsoft recommends at least 5 examples per class, but 20+ gives significantly better accuracy. Go to Models > Train model in your content center and add more labeled examples.

If you have active per-user licenses, you can still create prebuilt, structured, and freeform models and apply them to libraries, this capability isn't locked to pay-as-you-go. But running those models on demand or continuously does consume pay-as-you-go meters once your per-user license processing allowance is exhausted.

# Manually trigger a model run on an existing library via PnP PowerShell:
Invoke-PnPSiteDesign -Identity "your-model-site-design-id"

# Or via the SharePoint REST API:
POST https://yourtenant.sharepoint.com/sites/contentcenter/_api/machinelearning/publications
Content-Type: application/json
{
  "ModelUniqueId": "your-model-guid",
  "TargetSiteUrl": "https://yourtenant.sharepoint.com/sites/targetsite",
  "TargetWebServerRelativeUrl": "/sites/targetsite",
  "TargetLibraryServerRelativeUrl": "/sites/targetsite/Shared Documents",
  "ViewOption": "NewViewAsDefault"
}

After triggering, check the library's Document processing activity log (gear icon > Library settings > Document processing) for processing status and any error codes returned.