Fix Microsoft 365 Document Processing Issues
Why Microsoft 365 Document Processing Breaks
I've seen this exact situation play out in enterprise environments more times than I can count. Your SharePoint admin confidently clicks Set up document processing, follows what looks like a perfectly straightforward wizard , and then nothing works. Models refuse to run, autofill columns stay stubbornly blank, eSignature requests never send, and your OCR jobs silently fail. The error messages, when they appear at all, are vague. "Something went wrong" is not a diagnosis.
Here's the core reason this happens so often: Microsoft 365 document processing is not a traditional click-and-install feature. It's a collection of pay-as-you-go AI services that sit on top of an Azure billing subscription. If the Azure plumbing isn't connected correctly, every single downstream service fails, and Microsoft's UI gives you almost no feedback about why.
Before April 2025, these services were bundled under the "Microsoft Syntex" brand. If you've ever searched for Syntex documentation and landed on a page that now says "document processing for Microsoft 365," that's why. The rebrand was cosmetic, the features and functionality didn't change, but it created a wave of confusion because older setup guides, YouTube walkthroughs, and even some internal IT documentation still reference Syntex toggles that have been renamed or moved.
There's also an important licensing shift that trips up a lot of teams. Per-user licenses for these services are no longer available for purchase. If your organization was coasting on old Syntex per-user licenses, you may have noticed things starting to break as those licenses expired. The path forward is pay-as-you-go via Azure, but that requires an active Azure subscription linked to your Microsoft 365 tenant, which is a non-trivial configuration step that many tenants simply haven't done.
On top of that, Microsoft announced in October 2025 the progressive end of AI Builder credits. If your document processing workflows were relying on AI Builder credit consumption rather than the native pay-as-you-go meters, you're going to hit walls. This affects Power Platform integrations in particular.
For Government Community Cloud (GCC) organizations: pay-as-you-go licensing isn't available yet in your environment. You can continue on per-user licenses for now, but you can't access the pay-as-you-go service tier. This isn't a bug, it's a documented limitation.
The good news? Almost every Microsoft 365 document processing setup problem traces back to a handful of root causes: a missing or misconfigured Azure subscription link, incorrect admin permissions, a service not activated at the tenant level, or a SharePoint library that hasn't had a model applied correctly. Work through the steps below and you'll fix 95% of these problems. Browse all Microsoft fix guides →
The Quick Fix, Try This First
Before you spend an hour digging through Azure portal settings, do this one check. It resolves the majority of "document processing not working" complaints I see from admins who've set things up but can't get services to fire.
Go to the Microsoft 365 admin center at admin.microsoft.com. In the left navigation, expand Settings and click Org settings. Scroll down to find SharePoint in the list (not the SharePoint admin center, this is the org-level settings panel). From there, look for the document processing billing configuration panel.
What you're checking is whether pay-as-you-go billing has been connected. If you see a banner that says "Connect an Azure subscription to enable document processing services", that's your problem right there. Nothing will work until that link exists.
To connect it: click Set up pay-as-you-go billing, sign in with a Global Administrator or SharePoint Administrator account, and follow the wizard. You'll need to select an existing Azure subscription (or create one). The subscription must be active, a subscription in a disabled or expired state won't work even if it appears in the dropdown.
Once connected, wait about 5–10 minutes for the service meters to activate in your tenant. Then go back to your SharePoint document library, navigate to Library settings → Document processing, and check whether the service options now appear. If they do, you're in business.
If the billing is already connected and things still aren't working, keep reading, the step-by-step section covers every scenario I've encountered.
This is where every document processing setup either succeeds or dies. Without a valid Azure subscription link, the document processing services have no billing meter to charge against, so they simply refuse to run.
Sign into the SharePoint admin center at yourtenant-admin.sharepoint.com using a Global Admin or SharePoint Admin account. In the left navigation, click Settings. Look for the Document processing section. If you see a "Pay-as-you-go not configured" message, that confirms the root cause.
Click Go to billing setup. This redirects you to a panel where you select:
- Your Azure subscription (it must be in an Active state, not Disabled, Expired, or Past Due)
- A Resource Group within that subscription
- The Azure region for your billing meters
If you don't have an Azure subscription at all, you'll need to create one first at portal.azure.com. A Pay-As-You-Go subscription works fine here, there's no need for an Enterprise Agreement or reserved capacity.
After completing the wizard, you should see a confirmation screen stating "Pay-as-you-go billing is active." Give the system 10–15 minutes to propagate. Then return to a SharePoint document library and check Library settings → Document processing, the model creation options should now be accessible.
If the billing confirmation appears but services still don't show up in your library settings after 20 minutes, clear your browser cache and try in an InPrivate/Incognito window. Cached authentication tokens occasionally block the UI refresh.
Connecting Azure billing activates the billing infrastructure, but individual services still need to be turned on at the tenant level. I've seen plenty of setups where billing was connected correctly but OCR, autofill columns, or document translation still weren't available because the service switch was never flipped.
In the Microsoft 365 admin center (admin.microsoft.com), go to Settings → Org settings → Services. Scroll to find SharePoint. Within the SharePoint settings panel, look for the Document processing services section.
Here you'll find individual 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
Make sure the specific service you need is toggled On. Save your changes. Changes here can take up to 15 minutes to reflect in SharePoint libraries.
One thing I want to flag: if you're in a GCC (Government Community Cloud) tenant, these pay-as-you-go toggles won't appear. GCC organizations are on per-user licensing until Microsoft extends pay-as-you-go availability to that cloud environment. This is a known, documented limitation, not a bug with your setup.
When the service is properly enabled, you'll see the relevant options appear inside SharePoint document libraries under the Automate menu in the library toolbar.
Autofill columns are one of the most used, and most broken, features in Microsoft 365 document processing. The concept is simple: you define a column in a SharePoint library, configure it with a prompt or extraction rule, and the large language model fills it in automatically when a file is uploaded. In practice, I see three failure modes constantly.
Failure Mode 1, Column configured but never runs: This almost always means the autofill column wasn't properly saved or the library isn't syncing triggers. Go to the library, click the column header, select Column settings → Edit, and confirm the autofill configuration is saved. Then manually trigger it: select one or more files, click Automate → Autofill columns in the toolbar.
Failure Mode 2, Column runs but returns blank values: The LLM couldn't extract or generate the requested content from the document. This usually means the prompt is too vague, or the document format isn't being read correctly. Check the file type, autofill columns work best with DOCX, XLSX, PPTX, and PDF files. If you're uploading scanned PDFs without text layers, you need OCR enabled first (Step 4). Refine your extraction prompt to be more specific about where in the document the information appears.
Failure Mode 3, "Service unavailable" error on autofill: This means the pay-as-you-go meters aren't active yet, or the service was turned off at the tenant level. Go back through Steps 1 and 2 to confirm billing and service activation.
When autofill columns work correctly, you'll see a small processing indicator on the file row and then watch the column value populate within a few seconds to a couple of minutes depending on document size.
OCR is the engine that makes scanned documents readable by every other document processing service. If your prebuilt models aren't extracting data from invoices, if your unstructured models aren't classifying contracts correctly, and if autofill columns are returning blank on image-heavy PDFs, OCR is probably either not enabled or not running on those files.
First, confirm OCR is turned on. In the SharePoint admin center, navigate to Settings → Document processing and verify Optical character recognition shows as Active.
Then, enable OCR on the specific library where it's needed. Open the SharePoint document library, go to Library settings, and find Optical character recognition under the Document processing section. Toggle it on for that library.
Important: OCR processes new uploads automatically once enabled, but it does not retroactively process files already in the library. To run OCR on existing files, select them, then click Automate → Optical character recognition from the toolbar.
You can verify OCR ran successfully by checking the file's properties. After OCR processing, SharePoint updates the file's extracted text metadata. If you search for a phrase you know is in a scanned PDF and it now appears in search results, OCR is working.
One common gotcha: OCR consumes pay-as-you-go meters per page. If your Azure subscription has spending limits set at a very low threshold, OCR jobs on large PDF batches can hit the cap and silently stop. Check your Azure subscription's Cost Management + Billing → Budgets to make sure you haven't hit a limit.
eSignature in Microsoft 365 document processing keeps your documents inside the Microsoft 365 ecosystem while they're being reviewed and signed, no third-party DocuSign connection required. When it breaks, the failure usually falls into one of two buckets: setup problems or delivery problems.
Setup Problems: eSignature requires pay-as-you-go billing to be active (Step 1) and the eSignature service to be enabled at the tenant level (Step 2). Beyond that, the person creating the signature request must have at least Contribute permissions to the SharePoint library where the document lives. Read-only access isn't enough, the service needs to write back the signed document and update status metadata.
Delivery Problems: Recipients aren't getting their signature request emails. First check whether the recipient's email address is correct, this sounds obvious, but I've seen typos kill more eSignature workflows than any technical issue. Next, check your tenant's outbound email configuration. If your Microsoft 365 tenant has strict outbound mail flow rules or anti-spam policies, eSignature notification emails can get blocked before they leave the tenant.
To send an eSignature request from a document library: select the document, click Automate in the library toolbar, and choose Request signatures. Fill in the recipient details and signature field placements, then send. The document status column in the library will update to "Pending signatures" if the request was sent correctly.
If you see an error during the request creation flow, specifically anything mentioning "service not configured", go back and verify the eSignature service is toggled on in tenant settings and that billing is connected. Refreshing the page after making those changes (rather than navigating away and back) sometimes helps the UI pick up the updated service state faster.
Advanced Troubleshooting for Microsoft 365 Document Processing
If the steps above haven't fully resolved your Microsoft 365 document processing issues, you're likely dealing with one of a few more complex scenarios: Power Platform integration problems stemming from the AI Builder credit end-of-life, per-user license expiration, custom Power Platform environment misconfigurations, or enterprise network-level blocks.
AI Builder Credit End-of-Life Impact
In October 2025, Microsoft announced the progressive end of AI Builder credits. If your document processing pipelines were built on Power Automate flows that consumed AI Builder credits for document extraction, those flows may already be failing or will start failing as the credit pool winds down. The fix is to migrate those flows to use the native document processing services via pay-as-you-go meters rather than AI Builder credit consumption. This typically means rebuilding the Power Automate actions to use the updated SharePoint-native document processing connectors rather than the AI Builder Document Processing connector.
Per-User License Expiration
Per-user licenses for document processing services are no longer available to purchase, but if your organization has existing licenses, they can still be assigned to new users. Once they expire, though, there's no renewal path. You must switch to pay-as-you-go. If you're seeing users suddenly lose access to model creation or content assembly, check their license assignments in the Microsoft 365 admin center → Users → Active users → Licenses and apps. If their Syntex/document processing license shows as expired, that's the cause.
Custom Power Platform Environment Issues
If you're running structured or freeform document processing models in a custom Power Platform environment, there's an additional configuration step that's easy to miss. You need to set up that custom environment explicitly for document processing, the default Power Platform environment is configured automatically, but custom environments are not. Microsoft's documentation for this configuration is at the Power Platform admin center under Environments → [your environment] → Settings → Features.
Network-Level and Firewall Blocks
Enterprise environments with strict egress filtering sometimes block the Azure service endpoints that document processing uses for its AI inference calls. If everything looks configured correctly in admin portals but models still fail to run, ask your network team to check whether traffic to *.ai.azure.com and *.cognitiveservices.azure.com endpoints is permitted from your SharePoint Online IP ranges. These are the endpoints backing the OCR and LLM services.
Using the SharePoint Cost Calculator
If your concern isn't a hard failure but unexpectedly high Azure charges from document processing meters, Microsoft provides a SharePoint cost calculator tool specifically to help estimate pay-as-you-go spend. You can find it linked from the document processing billing documentation in the Microsoft 365 admin center. Input your expected monthly document volumes by service type to model costs before they hit your Azure bill.
Prevention & Best Practices for Document Processing
I've seen enough broken SharePoint document processing setups to know that the organizations that have the smoothest experience are the ones that did a bit of upfront planning. Here's what makes the difference.
Create a dedicated Azure subscription for Microsoft 365 pay-as-you-go services. Mixing document processing billing meters into a general-purpose Azure subscription makes cost tracking a nightmare. A dedicated subscription lets you set budget alerts specifically for document processing spend, gives your Microsoft 365 admins cost visibility without needing broad Azure access, and makes it easier to spot runaway usage (like a misconfigured OCR rule processing thousands of pages per day).
Set up Azure cost alerts before you go live. In the Azure portal, navigate to your subscription, go to Cost Management + Billing → Budgets, and create a monthly budget with alert thresholds at 80% and 100% of your expected spend. Without these alerts, it's entirely possible to accumulate unexpected charges from high-volume document processing before anyone notices.
Test new models in a non-production library first. Before applying an unstructured classification model or a prebuilt extraction model to a document library that contains thousands of existing files, test it on a small pilot library with 20–50 representative documents. This lets you validate extraction accuracy without burning through meters on documents that will reveal model quality issues you could have caught early.
Document your service configuration. Keep a record of which Azure subscription is linked, which services are enabled, which libraries have models applied, and which Power Automate flows depend on document processing. This sounds tedious but it pays off enormously when you're troubleshooting at 11pm because a procurement workflow stopped working.
Monitor the Microsoft 365 service health dashboard regularly. Document processing services occasionally have partial outages that affect specific regions or specific service tiers. Before spending time debugging a problem that appeared suddenly and without any configuration changes, check the Microsoft 365 admin center → Health → Service health dashboard to see if there's an active incident.
- Set Azure budget alerts at 80% of expected monthly spend before enabling high-volume OCR or autofill columns
- Enable OCR on a library before applying any AI models, models perform significantly better on text-layer PDFs than on raw scanned images
- Use the SharePoint cost calculator linked in the admin center to model costs for each service before activating across your whole tenant
- Review license assignments quarterly, expired per-user licenses silently break model creation access for affected users without any obvious error message
Frequently Asked Questions
What is the difference between Microsoft Syntex and Microsoft 365 document processing?
They're the same thing, Microsoft rebranded Syntex as "document processing for Microsoft 365" to better reflect where the services sit within the product stack. The rebrand happened in 2024/2025 and was cosmetic only; all the features, models, and configurations carried over without changes. If you see older guides referencing "Syntex admin settings" or "Syntex pay-as-you-go," those instructions still apply, just look for equivalently named settings in the updated admin UI. The core services (autofill columns, document translation, eSignature, OCR, prebuilt models, etc.) are all identical to what was launched under the Syntex name.
Why is document processing greyed out in my SharePoint library settings?
The most common cause is that pay-as-you-go billing hasn't been connected to your tenant, or it was connected but the specific service you need hasn't been enabled in Microsoft 365 admin center org settings. It can also happen if you're accessing the library with an account that doesn't have at least Site Owner or SharePoint Administrator permissions, the document processing configuration UI is hidden from users with lower permission levels. Work through Steps 1 and 2 in this guide to confirm billing and service activation, then verify your account permissions.
How much does Microsoft 365 document processing cost per month?
Document processing uses pay-as-you-go meters billed through your Azure subscription, there's no flat monthly fee. Each service has its own per-transaction or per-page rate. For example, OCR is billed per page processed, autofill columns are billed per column fill operation, and document translation is billed per character translated. Microsoft provides a SharePoint cost calculator (linked from the admin center billing documentation) to estimate your specific monthly costs based on expected volume. Through December 2025, organizations with pay-as-you-go billing set up received a limited amount of free monthly capacity on selected services to try them out at no cost.
Can I still use per-user Syntex licenses I purchased before the change?
Yes, existing per-user licenses remain valid and can still be assigned to new users in your tenant until they expire. With an active per-user license, you can apply unstructured models to libraries, create prebuilt/structured/freeform models, upload content to libraries with applied models, run models on demand, and use content assembly, taxonomy services, content query, annotations, and document library rules. However, per-user licenses are no longer available to purchase or renew. Once your existing licenses expire, you'll need to switch to pay-as-you-go billing to continue accessing those services. Plan for that transition before your license expiration date.
Why isn't document translation available in my tenant even though pay-as-you-go is set up?
Pay-as-you-go billing being active is necessary but not sufficient, each individual service also needs to be switched on in the tenant-level settings. Go to Microsoft 365 admin center, then Settings → Org settings → Services → SharePoint, and find the Document processing services section. Confirm that Document translation is specifically toggled on. If you're in a GCC tenant, note that pay-as-you-go services including document translation aren't yet available in that cloud environment, per-user licensing is the only current option for GCC. Allow 10–15 minutes after toggling a service on for the change to propagate before testing again.
My unstructured document processing model was working fine and suddenly stopped classifying documents, what happened?
A few things can cause a sudden stop without any configuration changes on your end. First, check the Microsoft 365 service health dashboard (admin.microsoft.com → Health → Service health) for any active incidents affecting SharePoint or document processing services. Second, check your Azure subscription status, if a billing payment failed or a spending limit was hit, the subscription can enter a degraded state that stops all pay-as-you-go meters. Third, if your model was trained using AI Builder credits, the progressive end of AI Builder credits announced in October 2025 may be the cause, those credit-based models need to be migrated to the native pay-as-you-go approach. If the model was working and the Azure subscription is healthy with no service incidents, open a Microsoft support ticket with your tenant ID and the library URL where the model is applied.