Fix Microsoft 365 Document Processing Issues
Why This Is Happening
You set up Microsoft 365 document processing , maybe you were trying to get autofill columns working in SharePoint, or you wanted OCR to extract text from scanned invoices, or your team needed document translation running across a library. And then something broke. Maybe the service just doesn't appear. Maybe your pay-as-you-go billing keeps failing to connect. Maybe models you trained months ago stopped processing files overnight.
I've seen this exact scenario on dozens of tenant configurations. Here's the honest truth: Microsoft 365 document processing is one of the most powerful features in the M365 ecosystem, and one of the most frustrating to get right when something goes sideways. The error messages are often vague. The admin center UI changes without warning. And the transition away from Microsoft Syntex branding to "document processing for Microsoft 365" left a lot of IT admins staring at documentation that doesn't match what they see on screen.
The core architecture hasn't changed though. Microsoft 365 document processing is a suite of AI-powered, pay-as-you-go services that sits on top of SharePoint and uses Azure as its billing backbone. That Azure dependency is the root cause of the majority of issues I see. If your Azure subscription isn't connected correctly, or if it lapses, hits a spending limit, or gets assigned to the wrong billing scope, everything downstream breaks silently.
Beyond billing, there are three other common failure categories. First, tenant-level configuration gaps: the services need to be enabled in the Microsoft 365 admin center before any SharePoint admin can use them. If a Global Admin skipped that step, nothing works no matter what you do in SharePoint settings. Second, license and permission mismatches: per-user licenses for these services are no longer sold as new purchases. Organizations still holding active per-user licenses can continue using them, but once those expire the switch to pay-as-you-go is mandatory, and that transition has its own gotchas. Third, model-specific failures: unstructured, structured, freeform, and prebuilt document processing models each have their own processing pipeline. A model that trained correctly can still fail at inference time if the library isn't configured properly or if the content type mapping breaks.
If you're on a Government Community Cloud (GCC) tenant, I want to save you some troubleshooting time right now: pay-as-you-go licensing is not yet available for GCC organizations. You're limited to per-user licenses until Microsoft rolls out that capability. Document processing services that require pay-as-you-go billing simply won't activate on a GCC tenant, that's by design, not a bug you can fix.
For everyone else, here's how to work through this systematically. Browse all Microsoft fix guides →
The Quick Fix, Try This First
Before you spend an hour digging through Azure portal settings, try this first. The single most common reason Microsoft 365 document processing services don't work is that pay-as-you-go billing isn't properly linked to an active Azure subscription. This one issue accounts for probably 60% of the support tickets I've seen on this topic.
Here's how to check it in under two minutes. Open the Microsoft 365 admin center at admin.microsoft.com. In the left navigation, go to Setup > Files and content. Look for the Document processing section. If you see a banner saying something like "Set up pay-as-you-go billing to use these services," that confirms this is your problem.
To fix it, click Go to SharePoint admin center. In the SharePoint admin center left nav, select Settings. Scroll down and click Document processing (it may still display as "Syntex" in older tenant configurations, they refer to the same thing). Click Go to Azure portal to set up billing. You'll need to:
- Select an active Azure subscription from the dropdown, it must be in an active, non-suspended state
- Select or create a Resource group within that subscription
- Select the closest Azure region to your SharePoint tenant's home region
- Click Save
After saving, go back to the SharePoint admin center and wait about 5 minutes for the configuration to propagate. Reload the Document processing settings page. If billing is now shown as "Active," try running one of your services again. In most cases, this resolves it immediately.
If you don't have access to create or link an Azure subscription, you'll need to loop in whoever manages your Azure billing, typically a Billing Administrator or Owner on the Azure subscription. This is not something a SharePoint admin can bypass.
Every document processing service, OCR, autofill columns, document translation, eSignature, content assembly, image tagging, taxonomy tagging, prebuilt models, structured models, freeform models, unstructured models, all of them require tenant-level activation before they do anything. This is a step that gets skipped constantly, especially when SharePoint admins assume a Microsoft 365 license is all that's needed.
Sign into admin.microsoft.com with a Global Administrator account. Navigate to Settings > Org settings > Services tab. Scroll through the list and look for SharePoint or Document processing. Depending on your tenant's admin center version, the exact label varies, but you're looking for a toggle or checkbox that enables AI-powered document services.
Alternatively, go directly to the SharePoint admin center at yourtenant-admin.sharepoint.com. Navigate to Settings in the left sidebar. You should see a Document processing entry. Click it. If the status shows "Not configured," document processing is disabled at the tenant level and no amount of site-level or library-level configuration will make it work.
Click Activate document processing (the exact button label may read "Enable" or "Turn on" depending on your tenant version). Confirm the activation prompt. Tenant-level changes like this typically take 15–30 minutes to propagate across SharePoint, don't test immediately. Come back after half an hour and try again.
If you don't see the Document processing entry in SharePoint admin settings at all, your tenant may be missing the required Microsoft 365 license at the tenant level. Every user who accesses document processing needs at minimum a Microsoft 365 license (E1, E3, E5, Business Basic, Business Standard, or Business Premium all qualify). The service itself bills via Azure, but base M365 access is still required.
When this step is done correctly, you'll see "Document processing: Active" in your SharePoint admin center settings, and the document processing options will become visible in SharePoint document library settings menus.
This is the single most technically involved step and the one most likely to block you if your Azure environment is complex. Document processing for Microsoft 365 uses Azure meters to bill for service consumption, autofill columns, OCR, document translation, eSignature requests, model processing, every action has a corresponding Azure meter that charges your subscription. No valid Azure subscription link, no processing.
In the SharePoint admin center, go to Settings > Document processing. Click the Azure portal link in the billing section. This takes you directly to the document processing billing setup blade in Azure. You'll see three fields:
Subscription: [Select active Azure subscription]
Resource group: [Select existing or create new]
Region: [Select region closest to your M365 tenant]
A few things to watch carefully here. The Azure subscription you select must be in an Enabled state, not Warned, Disabled, or Expired. If it shows any other status in the Azure portal under Subscriptions, fix the subscription issue first. The resource group doesn't need to be dedicated to document processing, you can reuse an existing one, but using a dedicated resource group makes cost tracking cleaner. For region, pick the Azure region that geographically matches where your SharePoint tenant's data residency is configured. Mismatching regions doesn't always break things, but it can introduce latency and occasionally causes service validation errors.
After clicking Save, you'll get a confirmation banner. Back in the SharePoint admin center, the billing status should update within 5–10 minutes. If it still shows "Not configured" after 15 minutes, go back to the Azure portal and check whether the Microsoft.SharePoint resource provider is registered on your subscription. Navigate to Azure portal > Subscriptions > your subscription > Resource providers. Search for Microsoft.SharePoint. If status shows "NotRegistered," click it and hit Register. This is a common miss on newer Azure subscriptions.
Once billing is active, you can monitor actual charges under Azure portal > Cost Management + Billing > Cost analysis, filtering by the resource group you selected.
With billing live and services enabled, the next common failure point is model configuration at the SharePoint library level. Whether you're working with an unstructured model, a structured or freeform model, a prebuilt model (for invoices, contracts, receipts), or one of the automated services like OCR or autofill columns, each one needs to be explicitly applied to a specific document library. They don't activate globally.
Navigate to the SharePoint document library where you want document processing to run. Click the gear icon (Settings) in the upper right. Select Library settings. Under the Document processing section (this section only appears if tenant-level activation from Step 1 is complete), you'll see options to apply or create models.
For prebuilt models like invoice processing or receipt extraction, click Apply a model > Prebuilt models. Select the model type (Invoice, Receipt, Contract, etc.) and click Add. The model is applied immediately, no training required because Microsoft pre-trained these on millions of documents.
For unstructured models (good for classifying documents that vary in structure), you'll need to go to a Content center site in SharePoint first. If your organization doesn't have a content center, a SharePoint admin needs to create one via SharePoint admin center > Active sites > Create > Content center. From the content center, create your model, provide at least 5 training examples per content type, train it, and then publish it to the target library.
When a model is properly applied to a library, you'll see a new Classified by column appear in the library view (for classification models) or new metadata columns populated with extracted values (for extraction models). If those columns aren't appearing after uploading test documents and waiting a few minutes, the model application didn't complete, go back to Library settings and check the Document processing section for any error messages.
Two of the most commonly used document processing features, autofill columns and optical character recognition, have their own specific failure modes that don't always surface clearly in the UI.
For autofill columns not generating content: Autofill uses large language models to extract or generate content automatically for SharePoint column values. If your autofill columns are sitting empty after files upload, check these in order. First, confirm the column type is supported, autofill works with single-line text, multi-line text, choice, and number columns, but not all column types. Second, check that the autofill definition is saved correctly. Go to Library settings > the column in question > Autofill settings. You should see a prompt or extraction rule defined. If the field is blank, the autofill never got configured properly. Third, autofill only runs on newly uploaded or modified files by default, existing library files won't be retroactively processed unless you manually trigger processing. To process existing files, select them in the library view, click Automate in the command bar, and choose Run document processing.
For OCR not extracting text from images: OCR in Microsoft 365 document processing extracts printed and handwritten text from image files (JPG, PNG, TIFF, BMP) and image-based PDFs. Common failure causes include file format issues (OCR doesn't process native Word or Excel files, it targets image content), file size limits (files over 20MB may not process), and library settings. Verify OCR is enabled on the specific library: Library settings > Document processing > confirm "Optical character recognition" shows as active.
After OCR runs successfully, extracted text appears in the file's metadata and becomes searchable via SharePoint search. If you're uploading an image file and the search index still shows no text content after 30 minutes, check the SharePoint Unified Audit Log for processing errors. Go to Microsoft 365 compliance center > Audit, filter by SharePoint activities and look for any document processing failure events against the file in question.
If OCR processing is consistently failing on specific file types, ensure those types aren't blocked by your tenant's SharePoint file type exclusion list under SharePoint admin center > Settings > Sync.
eSignature and document translation are two of the most visible pay-as-you-go document processing services, and they both have distinct setup requirements that trip people up.
For eSignature, the most common issue is that the "Request signatures" option doesn't appear in the SharePoint document library command bar. This almost always means one of two things: eSignature hasn't been enabled in the admin center, or the user doesn't have the right permissions. To enable it, go to Microsoft 365 admin center > Setup > Files and content > eSignature. Click Set up and follow the activation flow. You'll need to confirm your Azure billing connection here as well, eSignature is a paid service billed per signature request. Once activated, the "Request signatures" button appears in SharePoint document libraries for PDF files. Note: eSignature currently supports PDF format only. Trying to initiate a signature request on a .docx file directly won't work, convert to PDF first.
For document translation, the service creates a translated copy of a document in the same SharePoint library while preserving the original formatting. If translated documents aren't being created, or if you get an error when trying to translate, first verify the service is enabled: SharePoint admin center > Settings > Document processing > check that Document translation is listed as active. Second, confirm the source language is one of the supported languages, Microsoft supports translation for a wide range of languages and dialects, but not every file language is auto-detected reliably. If auto-detection is failing, try explicitly specifying the source language in the translation request dialog. Third, document translation only supports certain file types natively: DOCX, PPTX, XLSX, PDF, HTML, and a few others. Plain TXT files and older .doc format files may fail silently.
After a successful translation request, the translated file appears in the same library with a language code appended to the filename (e.g., Report_FR.docx for a French translation). If the file doesn't appear within 10 minutes, check the document library's information panel on the source file, it often shows a translation status indicator that reveals whether processing failed.
Advanced Troubleshooting
When the standard steps don't get you unstuck, it's time to look deeper. Here are the scenarios I see most often in enterprise environments where document processing breaks in non-obvious ways.
Per-user license transition issues. Per-user licenses for document processing services are no longer available for new purchase. If your organization has existing per-user licenses, they still work, but only for a specific subset of features: applying unstructured models to libraries, creating prebuilt/structured/freeform models, uploading content to libraries with applied models, running models on demand, using content assembly and taxonomy services, content query and annotations, and document library automation rules. If users are trying to access features outside this list with a per-user license and getting access denied errors, they're hitting the license boundary. Once those licenses expire, pay-as-you-go is the only path forward.
AI Builder credits ending. Microsoft announced in October 2025 that AI Builder credits are being progressively wound down. If your document processing models were previously consuming AI Builder credits (common in environments that used Power Automate flows with document processing), those flows will start failing as credits are depleted. The fix is migrating those workflows to use the native pay-as-you-go document processing services directly in SharePoint rather than routing through Power Automate's AI Builder actions. Check the Microsoft documentation on the End of AI Builder credits for the specific timeline affecting your region.
Custom Power Platform environments. If you're running structured or freeform document processing models that connect to a custom Power Platform environment (rather than the default environment), there's additional configuration required. The Power Platform environment must be explicitly registered with your SharePoint tenant's document processing setup. Go to SharePoint admin center > Settings > Document processing > Power Platform and ensure the correct environment is linked. Models trained in one Power Platform environment won't be visible when browsing from a different environment, this is a common source of "I can't find my model" confusion.
Event log and audit analysis. For persistent failures that don't generate visible error messages, the Microsoft 365 Unified Audit Log is your best diagnostic tool. Go to compliance.microsoft.com > Audit. Set the date range to cover when failures occurred. In the Activities filter, search for "document" to surface document processing-specific audit events. Look for events with "Failed" or "Error" in the operation details, these entries will include correlation IDs that Microsoft Support can use to trace the failure through their backend systems.
SharePoint content center model publishing failures. If you've trained a model in a content center site and publishing to a target library fails, check that the account doing the publishing has at minimum Site Owner permissions on the target library's site. Contribute-level permissions aren't sufficient for model publishing operations. Also verify the target library's content types are not locked, if the library's content type management is set to "Read only" through a content type hub, model publishing will silently fail without a clear error.
Prevention & Best Practices
I've watched a lot of organizations get document processing working, then break it six months later because of something preventable. Here's what to build into your operational habits from day one.
Monitor Azure billing alerts proactively. Because document processing bills through Azure meters, a misconfigured library or a runaway Power Automate flow can rack up unexpected charges. Set an Azure budget alert on the resource group you're using for document processing. In Azure portal > Cost Management + Billing > Budgets, create a monthly budget and set alert thresholds at 50%, 80%, and 100%. This gives you early warning before costs spiral and before Azure auto-disables your subscription due to a spending limit breach, which would silently kill all document processing services for your entire tenant.
Document your model configurations. Model training in content centers is not automatically backed up or version-controlled. If a content center site gets accidentally deleted or a model gets overwritten, you lose your training data. Keep an external record of: which models are trained on which content types, the training examples used for each model, which libraries each model is published to, and which columns are mapped to which extracted fields. SharePoint's own site export feature doesn't reliably capture content center model configurations.
Test pay-as-you-go services in a non-production library first. Before deploying document processing models to high-traffic production libraries, stand up a test library, publish the model there, and run a batch of representative test files through it. This lets you validate extraction accuracy and catch configuration errors without affecting real business data. Unstructured models in particular benefit heavily from iterative training refinement, the first version rarely gives you the accuracy you need for production workflows.
Track the AI Builder credits deprecation timeline. If any of your existing automation flows use AI Builder document processing actions, audit them now. The progressive end of AI Builder credits (announced October 2025) means any flow depending on those credits will eventually stop working. Identify affected flows in Power Automate > My flows and plan migration to native SharePoint document processing services before the credits run out for your region.
- Set Azure budget alerts on your document processing resource group so billing surprises don't suspend your services mid-month
- Assign a dedicated SharePoint admin as the document processing owner, a single person who monitors service health, billing, and model accuracy
- Use the SharePoint cost calculator Microsoft provides to estimate monthly spend before you roll out document processing across large libraries
- Review per-user license expiration dates now and plan the pay-as-you-go transition in advance, a last-minute scramble when licenses expire will disrupt your team's workflows
Frequently Asked Questions
Why can't I see the document processing option in my SharePoint library settings?
This almost always means document processing hasn't been enabled at the tenant level. A Global Administrator needs to activate it in the SharePoint admin center under Settings > Document processing, and pay-as-you-go billing needs to be linked to an active Azure subscription. Until both of those steps are complete, the option simply doesn't appear in library settings, SharePoint doesn't show it as disabled or grayed out, it just omits it entirely, which is genuinely confusing. If both those things are confirmed and it's still not appearing, try clearing your browser cache and checking from a private window to rule out a stale UI state.
How much does Microsoft 365 document processing actually cost per month?
The costs depend entirely on which services you use and how much volume runs through them, each service has its own Azure meter with its own per-unit rate. Microsoft provides a SharePoint cost calculator (linked from the official document processing pricing page) that lets you input estimated usage volumes and see projected monthly costs. There's no base fee or minimum commitment, you pay only when services are actually used. Through December 2025, Microsoft included a limited amount of free capacity each month for selected services to let organizations trial them, but that free tier period has now ended.
Can I still use Microsoft Syntex features or has everything been renamed?
Microsoft Syntex has been rebranded to "document processing for Microsoft 365", but the underlying features and functionality are completely unchanged. If you had Syntex models trained and published, they continue working exactly as before. The renaming is purely cosmetic from a functional standpoint. You'll see "Syntex" still referenced in some older admin UI screens and in Power Platform integrations. Over time Microsoft is updating those labels, but don't let the terminology mismatch confuse you, it's the same product.
My organization is on GCC, why won't document processing services activate?
Pay-as-you-go licensing for document processing is not yet available for Government Community Cloud (GCC) organizations, this is a deliberate compliance and sovereignty decision by Microsoft, not a bug. GCC tenants can only use document processing services through per-user licenses. Since per-user licenses are no longer sold as new purchases, GCC organizations with no existing per-user licenses cannot currently access these services at all. Microsoft has indicated pay-as-you-go will eventually come to GCC, but no specific date has been announced. Check the official Microsoft 365 Government roadmap for updates.
I trained an unstructured model but it's classifying documents incorrectly, what should I do?
Unstructured models need a minimum of five example documents per content type to train, but in practice you'll want 10–20 examples with good variety for reliable classification. If accuracy is poor, go back to the content center and add more training examples, especially examples that are being misclassified, so the model learns to distinguish edge cases. Also check your explanation phrases: explanations (the keyword lists and phrase patterns you define to help the model identify content types) have the biggest single impact on classification accuracy. Make sure explanations for each content type are genuinely distinctive and don't overlap heavily with other content types in your model.
Document translation is creating translated files but the formatting is completely broken, is there a fix?
Document translation in Microsoft 365 is designed to preserve the original format and structure of the source file, but complex formatting, especially tables with merged cells, embedded objects, custom styles, or heavy use of text boxes, doesn't always survive translation perfectly. This is a known limitation of machine translation applied to richly formatted documents. The best mitigation is to simplify the source document's formatting before translation where possible. For PDFs, the translation accuracy for layout preservation is heavily dependent on whether the PDF is a native PDF (text-based) or a scanned image-based PDF, scanned PDFs produce much poorer translation layout results than native PDFs.