Fix Microsoft 365 Document Processing Errors

Microsoft 365 Document Processing runs entirely on a pay-as-you-go billing model. There's no upfront license to buy and no per-user seat to assign, but that means an active, valid Azure subscription is non-negotiable. Without it, the services simply won't run.

To set this up or repair a broken connection:

1. Sign in to the Microsoft 365 admin center at admin.microsoft.com using a Global Admin or SharePoint Admin account.
2. Navigate to SharePoint admin center → Settings → Document processing.
3. Under Pay-as-you-go billing, click Set up billing (or Edit billing if one was previously configured).
4. In the Azure subscription picker, select the subscription you want to use for billing. If you don't see the right subscription, make sure the account you're signed in with has at least Contributor or Owner access on that Azure subscription.
5. Select a Resource group and confirm the Region matches your Microsoft 365 tenant region. Microsoft documents this requirement, billing meters are region-scoped.
6. Click Save and wait for the confirmation message.

Once saved, head to the Azure portal at portal.azure.com, open Cost Management + Billing, and verify that the SharePoint document processing meters appear under your subscription. You should see service meters like "Syntex" or "SharePoint Document Processing" once the connection is healthy and usage has begun.

What you should see if it worked: Back in the SharePoint admin center, the billing status changes from "Not configured" to a green "Active" indicator. Document processing services will begin appearing and responding within 15–30 minutes of a successful billing connection.

Even with billing connected, individual users can still be blocked from Microsoft 365 Document Processing services if permissions aren't set up correctly at the tenant or site level. This is one of the most common complaints I hear from IT admins after billing is confirmed working, the admin can use the services, but regular users can't.

Here's how the permission model works: any user in your organization can use document processing services, provided they have a valid Microsoft 365 license. That's the pay-as-you-go promise. But "can use" at the platform level doesn't automatically translate to "has access" at the SharePoint site or library level.

Check these specific things:

1. SharePoint site permissions: The user must have at least Contribute permission on the SharePoint document library where they're trying to use autofill columns, taxonomy tagging, or any model-based processing.
2. Content center access: If you're creating or managing models (prebuilt, structured, freeform, or unstructured), the user needs access to your organization's Content Center site. Check that they have Member or Owner access on that site collection.
3. Admin center service toggles: In the SharePoint admin center under Settings → Document processing, individual services like autofill columns, eSignature, and optical character recognition each have their own toggle. Make sure the specific service the user needs is switched On at the tenant level.

To check service-level toggles, go to SharePoint admin center → Settings → Document processing, scroll down past the billing section, and you'll see individual enable/disable toggles for each service. Any service showing as Off is unavailable tenant-wide regardless of billing status.

What you should see if it worked: The affected user can access the document processing features (autofill column setup, OCR, eSignature send button) from within the SharePoint library without receiving permission error dialogs.

Autofill columns use large language models (LLMs) to extract or generate content from files and populate SharePoint metadata columns automatically. When they stop working, you'll typically see one of these symptoms: the column shows a spinning indicator that never resolves, extracted values come back blank, or the column shows an error icon with no explanation.

Work through these checks in order:

1. File type support: Autofill columns work with a specific set of file types. Confirm the files you're targeting are in a supported format, primarily PDF, Word (.docx), PowerPoint (.pptx), and Excel (.xlsx). Files in older binary formats (.doc, .xls) or image-only PDFs (scanned pages with no text layer) may fail to process correctly. For image-heavy PDFs, combine autofill columns with optical character recognition, run OCR on the document first to generate a text layer, then let autofill columns extract from that.
2. Column configuration: Open the column settings for the autofill column in question. Go to the library → Library settings (gear icon) → click the autofill column name. Verify the extraction prompt or source configuration is saved and not blank. If the prompt got corrupted or cleared, the column silently returns nothing.
3. Billing meter check: In the Azure portal, open Cost Management and filter by the resource group tied to document processing. Look for autofill-related meter entries. If there are zero entries despite the column being used, billing isn't metering correctly and you likely need to go back to Step 1.
4. Library item count: Very large document libraries (100,000+ items) can experience autofill processing delays. Microsoft processes these in queues. If you're in a large library, try creating a test library with 5–10 files and the same column configuration to isolate whether the problem is scale-related or configuration-related.

What you should see if it worked: The autofill column populates with extracted or LLM-generated values within a few minutes of uploading a new file to the library. Existing files can be re-processed by selecting them and choosing the refresh/reprocess option from the column menu.

eSignature and document translation are two of the most visible pay-as-you-go services in Microsoft 365 Document Processing, and both have specific failure patterns worth knowing.

For eSignature failures:

1. First, confirm that eSignature is enabled for your tenant. In the SharePoint admin center → Settings → Document processing, find the eSignature toggle and make sure it's On.
2. eSignature only works on files stored in SharePoint document libraries. If a user is trying to send an eSignature request from OneDrive personal storage, it won't work, the file must be in a SharePoint library. Move the file first, then send the request.
3. The sender must have at least Edit permissions on the file. Read-only access is not enough to initiate an eSignature workflow.
4. Check that the recipients have valid email addresses. eSignature requests route through email, and distribution lists or shared mailboxes as recipients can cause silent delivery failures.

For document translation failures:

1. Document translation creates a translated copy in the same SharePoint library as the source file. Verify the user has Contribute permission on that library, they need write access to create the translated copy.
2. Check that the target language you're requesting is in Microsoft's supported language list. Unsupported language codes will cause a silent failure rather than a clear error message.
3. File size limits apply. Documents larger than 40 MB may time out during translation. Split large documents or translate sections separately.

What you should see if it worked: For eSignature, the recipient receives an email with a signing link within a few minutes of the request being sent. For document translation, a new file with the target language code appended to the filename appears in the same library folder, preserving the original file format and structure.

Optical character recognition (OCR) in Microsoft 365 Document Processing extracts printed or handwritten text from images, making scanned documents, photos of whiteboards, and image-only PDFs fully searchable and processable. When OCR isn't working, you lose that text layer entirely, which also breaks downstream services like autofill columns that depend on readable text.

For prebuilt, structured, freeform, and unstructured models, processing errors usually show up as models returning empty results or stuck in a "processing" state.

Here's how to address each:

OCR not processing images:

1. Confirm OCR is enabled at the tenant level: SharePoint admin center → Settings → Document processing → find the Optical character recognition toggle and set it to On.
2. OCR must be enabled on each specific SharePoint document library where you want it to run. Go to the library → Library settings → Document processing settings → enable OCR for that library.
3. OCR runs asynchronously after file upload. Wait 5–10 minutes before concluding it hasn't worked. For large batches, wait longer.

Prebuilt and custom model failures:

1. In your Content Center, open the model and click Run model on demand to trigger a fresh processing pass. If this fails with an error, note the exact error text, it usually contains a code like ModelProcessingError or InsufficientPermissions.
2. For structured and freeform models, verify the model has been trained with enough example documents. Microsoft recommends at minimum five positive example documents for reliable extraction.
3. If you're running models in a custom Power Platform environment, verify the environment is correctly configured, this is a separate setup step from the main pay-as-you-go billing configuration.

# PowerShell: Check if a SharePoint library has document processing configured
Connect-SPOService -Url https://[yourtenant]-admin.sharepoint.com
Get-SPOSite -Identity https://[yourtenant].sharepoint.com/sites/[sitename] | Select-Object Url, SyntexEnabled

What you should see if it worked: After OCR processes a scanned document, searching for text that appears in the image returns the document in SharePoint search results. For models, the metadata columns you configured populate with extracted values after the model runs.