Fix Microsoft 365 Document Processing Issues Fast
Why Microsoft 365 Document Processing Keeps Breaking
You sat down, opened SharePoint, clicked where you expected a document processing option to be , and nothing. Or worse, the option is there but the model just won't run. Maybe you get a vague "service not available" banner, or autofill columns simply refuses to extract anything from your uploaded contracts. I know this is frustrating, especially when your whole team is waiting on that invoice automation to work before end of quarter.
Microsoft 365 document processing , what used to be called Microsoft Syntex, went through a significant branding and licensing overhaul. The core functionality is identical, but the way you pay for it, activate it, and administer it changed substantially. Most of the problems I see come directly from that transition. Teams that were using Syntex under a per-user license are suddenly finding features locked or behaving unexpectedly as those licenses age out. Administrators who set up billing months ago are discovering their Azure subscription wasn't linked correctly. Developers building automated content workflows hit walls because service meters weren't enabled at the tenant level.
Here are the specific things that go wrong most often:
- No Azure subscription linked: Document processing is now a pure pay-as-you-go service billed through Azure. If no active Azure subscription is connected to your Microsoft 365 tenant, none of the document processing services will function, not OCR, not document translation, not autofill columns, nothing.
- Expired per-user licenses with no replacement billing: Per-user licenses for these services are no longer available for new purchase. Organizations with existing licenses that have lapsed but haven't switched to pay-as-you-go hit a hard wall where models stop running and library settings disappear.
- Service not enabled at the SharePoint admin level: Even with billing in place, individual document processing services need to be explicitly turned on inside the Microsoft 365 admin center. Autofill columns, unstructured document processing, prebuilt document processing, each one has its own toggle.
- Government Community Cloud (GCC) limitations: If your organization is on GCC, pay-as-you-go licensing is not yet available at all. You must maintain active per-user licenses to keep using services. This catches a lot of government contractors off guard.
- AI Builder credit confusion: Microsoft announced a progressive end to AI Builder credits in October 2025. Organizations that were relying on AI Builder credits for document processing capacity are finding those credits no longer apply, and the billing model underneath has shifted entirely.
The frustrating thing is that Microsoft's error messages almost never tell you which of these is actually the problem. You get generic "unable to process" messages or features that silently stop working without any notification. That's why I wrote this guide, to walk you through diagnosing and fixing each one of these systematically.
The Quick Fix, Try This First
Before you go deep into troubleshooting, try this single check. It resolves the majority of Microsoft 365 document processing failures I encounter, especially after the Syntex-to-document-processing migration.
Open the Microsoft 365 admin center at admin.microsoft.com. In the left navigation, go to Settings > Org settings, then click the Services tab. Scroll down until you find Document processing. Click it. You'll see a panel on the right that shows whether pay-as-you-go billing is configured and which individual services are enabled.
If the billing status shows "Not configured" or "No Azure subscription linked", that's your answer. Everything flows from that. Without an active Azure subscription connected here, none of the downstream services will run, no matter how many licenses you've assigned or how many models you've built.
If billing looks fine in the admin center, the next thing to check is whether the specific service you're trying to use is toggled on. Microsoft 365 document processing is not one monolithic service, it's about ten distinct services (autofill columns, document translation, eSignature, OCR, content assembly, image tagging, taxonomy tagging, prebuilt models, structured and freeform models, and unstructured models) and each one can be independently enabled or disabled. It's entirely possible to have billing active but a specific service like optical character recognition switched off, which would explain why OCR isn't extracting text from your scanned PDFs even though invoice processing seems fine.
Once you identify and fix either the billing link or the service toggle, go back to your SharePoint document library, refresh the page, and try re-applying the model. In most cases the feature reappears immediately, no sign-out required.
This is the foundation. Microsoft 365 document processing is billed entirely through Azure service meters, there's no separate Microsoft 365 subscription cost for consumption. Your organization needs an active Azure subscription, and that subscription needs to be explicitly linked to your tenant for document processing billing to work.
Here's how to set it up or verify it's set up correctly:
- Go to admin.microsoft.com and sign in as a Global Admin or SharePoint Admin.
- Navigate to Settings > Org settings > Services.
- Find and click Document processing in the list.
- In the panel that opens, look for the Azure subscription section. If it says "No subscription linked," click Set up pay-as-you-go billing.
- You'll be taken through a wizard where you select your Azure subscription, resource group, and region. Pick the region closest to your users for lowest latency on processing jobs.
- Confirm the Azure subscription is active (not suspended or over-budget) by opening portal.azure.com, navigating to Subscriptions, and checking the status column shows "Active."
Once the subscription is linked, service meters for document processing appear in your Azure Cost Management dashboard under the namespace Microsoft.SharePoint. You can use the SharePoint cost calculator (accessible from the document processing settings panel) to model your expected monthly spend before you start processing large document volumes.
If the billing setup wizard fails partway through, check that the account you're using has Contributor or Owner role on the Azure subscription, Reader access is not enough to authorize the billing link.
Success indicator: The Document processing settings panel shows your subscription name, status as "Active," and a green checkmark next to billing configuration.
Having billing connected doesn't automatically turn on every service. Microsoft 365 document processing consists of ten distinct pay-as-you-go services, and you need to enable each one your organization plans to use. This is easy to miss because the UI doesn't shout at you, disabled services just quietly don't appear in SharePoint.
Back in the Document processing panel at Settings > Org settings > Services > Document processing, scroll through the list of available services. You'll see toggles for:
- Autofill columns
- Document translation
- eSignature
- Optical character recognition (OCR)
- Content assembly
- Image tagging
- Taxonomy tagging
- Prebuilt document processing
- Structured and freeform document processing
- Unstructured document processing
Enable only what you actually need right now. Each enabled service starts incurring costs the moment it processes a file, even in test scenarios. Turning on image tagging on every library when you only need invoice extraction is the kind of thing that shows up as a surprise on your Azure bill at month end.
For most enterprise document processing workflows involving contracts and invoices, you'll want to enable Prebuilt document processing (works out of the box for invoices, receipts, and contracts without training), Unstructured document processing (for classifying varied document types), and Structured and freeform document processing (for extracting specific fields from forms).
If you're running Microsoft 365 document translation to localize content across SharePoint libraries, that's its own separate toggle and its own per-page pricing meter in Azure.
Success indicator: After enabling a service, go to a SharePoint document library, click the gear icon, choose Library settings, and look under Automate, the newly enabled service should now appear as an option.
Once services are enabled at the tenant level, you configure document processing at the individual SharePoint library level. This is where most users get confused, document processing models are applied per-library, not per-site or per-tenant. A model you build for one library doesn't automatically apply anywhere else.
Navigate to the SharePoint document library where you want document processing to work:
- Open the document library.
- Click Automate in the top command bar (if you don't see this, the service may still be disabled, go back to Step 2).
- Select Apply a document processing model.
- You'll see a panel asking whether you want to create a new model or apply an existing one. If this is your first time, choose Create a new model.
- Pick the model type that fits your content, Prebuilt, Structured/Freeform, or Unstructured. For invoices and receipts, choose Prebuilt and save yourself hours of training.
For SharePoint document processing autofill columns specifically, the setup path is slightly different. Go to the column you want to autofill, click the column header dropdown, select Column settings > Edit, and look for the Autofill section at the bottom of the editing panel. This uses large language models to either extract content from the document or generate it based on document context, you configure a prompt that tells the LLM what to pull out.
One thing worth knowing: models applied to a library run on every new file uploaded going forward. They do not retroactively process existing files unless you explicitly trigger a Run model command on those files manually.
Success indicator: After configuring, upload a test document to the library. Within a few seconds to a couple of minutes (depending on document complexity and service load), the columns you configured should populate automatically.
Sometimes the model is set up, billing is connected, the service is enabled, and files just sit there with columns blank. No error, no processing indicator, just silence. Here's how to unstick it.
First, check whether the model is actually applied to the library and in an active state. Go to Automate > View applied models on the library. If you see your model listed but it shows a status of Paused or has a yellow warning icon, click it and look for a reactivation option.
If the model appears active but still isn't processing, try triggering it manually on a specific file:
- Select the file in the library by clicking the checkbox next to it.
- Click Automate in the command bar.
- Choose Run model on selected files.
Watch whether it processes immediately or returns an error. If you get an error code here, note it, errors starting with 400 typically indicate a model configuration problem (bad prompt, missing required field), while 503 errors indicate service availability issues on Microsoft's end.
For unstructured document processing models that won't classify correctly, the culprit is usually insufficient training examples. Microsoft recommends at least five example documents per content type, but I've found that eight to ten gives dramatically better accuracy. If you trained on five and you're seeing misclassifications, go back to Content Center > Models > [Your Model] > Train and add more labeled examples, then retrain and republish the model.
Also check that the SharePoint site hosting your Content Center hasn't had its document processing model publishing permissions changed. Go to the Site permissions of your Content Center site and confirm that site members have at least Contribute access, read-only Content Center access prevents model updates from propagating to libraries.
Success indicator: Manual run completes and the column values populate. Going forward, new uploads should process automatically within one to three minutes.
If your issue is with a specific service rather than the whole document processing stack, here are targeted fixes for the three most commonly broken individual services.
Optical Character Recognition (OCR) not extracting text: OCR in Microsoft 365 document processing scans image-based PDFs and photographs to make them text-searchable in SharePoint. If it's not working, first confirm you've enabled OCR specifically (it has its own toggle separate from other document processing services). Then check the file type, OCR supports JPG, PNG, TIFF, and PDF formats. Files uploaded as HEIC (common from iPhone photos) won't trigger OCR without conversion. Finally, verify the file is under 20MB; oversized files fail silently.
Document translation producing no output file: When document translation runs, it creates a translated copy in the same library with the target language appended to the filename (e.g., Contract_ES.docx for Spanish). If no output file appears, check that the destination library has storage space and that the account running the translation has Write permissions to that library, not just the source library. Translation supports all Microsoft 365-supported languages, but the translation itself takes anywhere from 30 seconds to a few minutes depending on document length.
eSignature requests not sending: The Microsoft 365 eSignature service keeps documents inside your Microsoft 365 environment during the signing process, a key compliance advantage. If signature requests fail to send, verify the recipient email address is valid and reachable. Also confirm that the document is stored in OneDrive or a SharePoint library, not a local file, eSignature does not support local file paths. Check that the sender has a Microsoft 365 license (any tier) assigned.
Success indicator: For OCR, search SharePoint for a word that appears only in the image content of your test file, it should return results. For translation, a new file with the language suffix appears in the library. For eSignature, the recipient receives an email within two to three minutes.
Advanced Troubleshooting for Microsoft 365 Document Processing
If the steps above didn't resolve your issue, you're dealing with something at the tenant configuration, Group Policy, or enterprise identity layer. These are less common but they do happen, especially in large organizations with complex Azure architectures.
AI Builder credits have ended, update your billing model now: In October 2025, Microsoft announced the progressive end of AI Builder credits. If your document processing workflows relied on AI Builder credits for processing capacity, those credits are gone. You need to have pure pay-as-you-go Azure billing configured. Go to the Microsoft 365 admin center, find Document processing settings, and verify the billing source is an Azure subscription, not AI Builder credits. If you're still seeing AI Builder credit references in your tenant, it means the migration hasn't completed and you should proactively switch to Azure billing before your credits zero out entirely.
GCC tenant limitations: Government Community Cloud organizations face a specific wall here. Pay-as-you-go licensing and the services that require it are not yet available in GCC. If you're on GCC, you must maintain active per-user licenses. Per-user licenses are no longer sold new, but existing licenses can still be assigned to new users. Once those existing licenses expire, your options are limited, contact your Microsoft account team about GCC roadmap timelines for pay-as-you-go availability.
Custom Power Platform environment for structured models: Structured and freeform document processing models that run inside a custom Power Platform environment have additional configuration requirements. The environment must have the correct data policies applied and must be linked to your SharePoint tenant correctly. If your structured models work fine in the default environment but fail in a custom environment, check the Power Platform admin center at admin.powerplatform.microsoft.com, go to Environments > [Your Environment] > Settings > Features and confirm document processing services are enabled at the environment level.
Event log and diagnostic analysis: SharePoint document processing doesn't write to the Windows Event Viewer, but it does write to the Unified Audit Log in Microsoft Purview compliance portal. Go to compliance.microsoft.com, navigate to Audit, and filter activities for SharePoint file operations and Content Understanding. Look for failed events, they'll show the error type and affected file, giving you a much clearer picture than any SharePoint UI error message.
PowerShell diagnostic commands, if you have the SharePoint Online Management Shell installed, you can check document processing model status across your tenant:
# Connect to SharePoint Online
Connect-SPOService -Url https://yourtenant-admin.sharepoint.com
# Check syntex (document processing) settings at tenant level
Get-SPOTenant | Select-Object ContentUnderstandingEnabled, SyntexEnabled
# List all content centers in the tenant
Get-SPOSite -Template "CONTENTCTR#0" | Select-Object Title, Url, StorageUsageCurrentIf ContentUnderstandingEnabled returns False, that's your issue, it needs to be enabled either via the admin center UI or via Set-SPOTenant -ContentUnderstandingEnabled $true.
Prevention & Best Practices for Document Processing
Once you have Microsoft 365 document processing running properly, keeping it running comes down to a few consistent habits. I've seen well-configured document processing setups break simply because no one was watching the billing or the model health, small things compound over time.
Set up Azure cost alerts for your document processing meters. In the Azure portal, go to Cost Management + Billing > Budgets and create a budget specifically for the Microsoft.SharePoint meter namespace. Set an alert at 80% of your expected monthly spend. Document processing costs can spike unexpectedly if someone applies a model to a library containing tens of thousands of historical files and then triggers a bulk reprocessing run.
Keep your document processing models maintained and retrained. Unstructured models especially tend to drift in accuracy as the composition of your document library changes over time. Build a quarterly review into your IT calendar, check model accuracy reports in the Content Center (go to the model, click Model activity), look for declining confidence scores, and add new training examples if you see accuracy below 85%.
For teams using autofill columns, document your LLM prompts in a central location. Prompts that work today may behave differently as Microsoft updates the underlying LLM, keeping a change log with the prompt text, the date it was written, and the expected output makes troubleshooting regressions dramatically faster.
If you're running document translation at scale, monitor the SharePoint storage in your tenant. Translation creates a new copy of every translated document in the same library, on a library with thousands of files being translated to multiple languages, storage grows fast. The SharePoint admin center shows storage per-site; build a monthly check into your process.
- Create an Azure cost budget alert at 80% of expected monthly document processing spend, surprises on the bill are avoidable.
- Only enable the specific document processing services your organization actually uses, idle enabled services add no cost, but keeping the list tight makes auditing and troubleshooting much cleaner.
- Run new document processing models on a small test library with 20–30 representative files before applying to production libraries with thousands of documents.
- Schedule a quarterly model accuracy review using the Model activity dashboard in your Content Center, catching accuracy drift early is far cheaper than fixing misclassified content after the fact.
Frequently Asked Questions
What happened to Microsoft Syntex, is it gone?
Microsoft Syntex isn't gone, it was rebranded. The pay-as-you-go services that were part of Microsoft Syntex are now officially called "document processing for Microsoft 365." The features and functionality are identical; only the name and some UI labels changed. If you had Syntex set up and it suddenly seemed to disappear, check that you're looking in the right place in the admin center under the "Document processing" label rather than searching for "Syntex."
Can I still use my old Syntex per-user license, or do I have to switch to pay-as-you-go?
You can still use existing per-user licenses, they haven't been deactivated. Per-user licenses just can't be newly purchased anymore, so once yours expire, pay-as-you-go is the only path forward. While your per-user license is active, you can apply unstructured models to libraries, create prebuilt and structured models, upload content, run models on demand, and use content assembly and taxonomy services. The main limitation versus pay-as-you-go is that some newer services require Azure billing, if you hit a wall on a specific service, that's likely why.
My organization is on GCC, why can't I set up pay-as-you-go billing for document processing?
Government Community Cloud (GCC) tenants don't yet have access to pay-as-you-go billing for document processing. This is a Microsoft platform limitation, pay-as-you-go and the Azure billing integration it requires are not currently available for GCC. Your only option right now is to maintain active per-user licenses. Contact your Microsoft account representative for the most current GCC roadmap information, as availability timelines change and your rep will have the latest information.
Why did my autofill columns stop populating after a few weeks of working fine?
The most common cause is an Azure billing issue, either the subscription went over budget, the subscription was suspended, or billing authorization expired. Check the Azure portal for subscription status first. The second common cause is that the LLM prompt powering your autofill column was written against a specific document structure that changed. If your document template was updated, the prompt may no longer extract the right field. Go to the column settings and edit the autofill prompt to match the new document layout. Finally, check whether someone accidentally disabled the document processing service at the tenant level, it takes about 30 seconds to verify in the admin center.
How much does Microsoft 365 document processing actually cost, I can't find a clear number?
Pricing is per-transaction and varies by service, there's no single flat number. Each service has its own Azure meter rate: OCR is priced per page extracted, document translation is priced per character translated, eSignature is priced per signature request, and so on. The best way to estimate your costs before committing is to use the SharePoint cost calculator, which is linked directly from the Document processing settings panel in the Microsoft 365 admin center. Enter your expected monthly volumes and it gives you a projected monthly spend. Microsoft also offered included capacity at no cost through December 2025 for organizations with pay-as-you-go set up, check whether that still applies to your tenant.
Can any user in my organization use document processing, or do I need to assign licenses?
Under the pay-as-you-go model, any user in your tenant with a Microsoft 365 license can use document processing services, you don't need to assign separate document processing licenses to individual users. The cost is tracked at the tenant level and billed to your Azure subscription. This is one of the real advantages of the pay-as-you-go approach over the old per-user licensing model: you're not gatekeeping access based on whether someone has a specific add-on license. Just make sure the specific services they need are enabled at the admin level and that the SharePoint libraries they're working in have models applied.