Microsoft Teams and SharePoint 365 captioning for L&D: live event captions, recorded meeting transcripts, SharePoint Stream video libraries, Viva Learning content, and how to get 365 content into a WCAG-compliant caption workflow

Why Microsoft 365 is now the most important captioning blind spot in enterprise L&D

The canonical enterprise L&D caption workflow was designed in the era of the standalone LMS. An instructional designer creates a course in Articulate Storyline, exports SCORM, uploads to Cornerstone or Docebo, and the caption review process happens at the upload step. The workflow is linear, the platform is known, and the caption policy is enforced at a single chokepoint.

That model no longer describes how enterprise training works. The average mid-market organisation now has three or four informal training surfaces running in parallel with the LMS:

• Teams meetings where managers walk through new product features, compliance procedures, or onboarding steps — recorded and shared to OneDrive or SharePoint
• Teams Town Halls where leadership communicates policy changes, strategic direction, or merger integration guidance to hundreds of employees simultaneously
• SharePoint team sites and department intranet pages where SMEs post how-to video walkthroughs, screen recordings, and training updates between formal LMS course releases
• Viva Learning assignments that pull content from both the LMS and SharePoint simultaneously, surfacing all of it in a single employee learning feed

The volume on these informal surfaces often exceeds the LMS. A fifty-person L&D team at a 2,000-employee SaaS company typically produces three or four formal LMS courses per quarter. The same organisation probably generates thirty or forty informal training recordings per month through Teams — product walkthroughs, sales enablement updates, compliance reminders, manager briefings. None of that content goes through the formal caption workflow because it was never considered “L&D content” when the caption policy was written.

The ADA Title II compliance picture for public entities and the ADA Title I picture for private employers do not make a distinction between formal and informal training content. If a recording is made available to employees as training or required viewing, the same WCAG 2.1 AA obligation that applies to an LMS course applies to the Teams recording. The auto-captions that WCAG 2.1 AA does not accept as compliant are the same auto-captions that Microsoft 365 generates by default across every one of these surfaces.

Teams Live Events and Town Halls: the auto-caption accuracy problem

What Teams generates

Teams Live Events (now being superseded by Teams Town Hall in the Microsoft product roadmap) and Town Halls provide two distinct caption experiences. During the event, Azure AI Speech generates real-time captions displayed in the viewer panel. After the event, the same speech recognition engine produces a transcript that is attached to the recording. Both the real-time captions and the post-event transcript are available to attendees, but neither is delivered as a standalone SRT or VTT file without an additional download step.

The real-time caption stream during a live event is the live caption equivalent of auto-captions on a recorded video: it is generated by the same Azure Cognitive Services infrastructure, constrained by the same accuracy limitations, and subject to the same vocabulary and speaker-accent variability. What it is not is a professionally reviewed, synchronised caption file meeting WCAG 2.1 AA. The distinction matters because real-time CART (Communication Access Realtime Translation) captioning — human-generated at speed during a live event — is the WCAG 2.1 AA-compliant path for live events. CART captioning for a Town Hall runs $120–$180 per hour depending on subject matter complexity. Auto-captions are free. The price differential explains why most organisations use auto-captions and the accuracy differential explains why they should not for any Town Hall that constitutes required professional development or policy communication.

Accuracy in practice

Microsoft has published accuracy benchmarks for Azure Cognitive Services speech recognition in controlled conditions: 90%+ on clean audio with standard American or British English. Enterprise training content rarely meets the conditions of that benchmark. The accuracy degradation factors in a typical Town Hall or recorded Teams meeting include:

• Product names and internal terminology. A SaaS company’s product manager describing the Q3 release pipeline will use product feature names, SDK terms, and internal shorthand that are not in the Azure language model. Azure will substitute the phonetically nearest common word. The accuracy degradation on product-vocabulary-dense content is the same problem that affects all auto-caption tools, including Whisper, and is the reason glossary-biased decoding exists as a technique.
• Meeting room audio. A Town Hall broadcast from a conference room with room mic pickup, background HVAC noise, and multiple speakers introduces acoustic variability that degrades recognition accuracy. Studies of automatic speech recognition accuracy in meeting contexts find accuracy rates 10–15 percentage points lower than clean studio audio benchmarks.
• Speaker accents. Azure Cognitive Services performs best on native speaker American and British English. For multinational organisations where Town Hall presenters include speakers from South Asia, East Asia, Eastern Europe, or Latin America, recognition accuracy drops further. This matters particularly for the population most likely to need captions: non-native English speakers in the audience are disproportionately represented among the people who depend on captions to follow content delivered in fast, accent-heavy English.
• Technical and regulatory vocabulary. Legal, compliance, financial, and technical terminology in any regulated sector — healthcare, financial services, pharmaceutical, government contracting — is systematically under-represented in general speech recognition training data. The words that matter most for compliance communication accuracy are exactly the words that auto-captions get wrong most frequently.

The practical outcome is that a Teams Town Hall recording with auto-captions attached will typically have caption accuracy in the 78–85% range for content with any technical or product-specific vocabulary. WCAG 2.1 AA requires accuracy sufficient to convey the information in the audio, and the DCMP standard used by regulators and courts as the operational 99% benchmark is the threshold most compliance programmes need to meet. The auto-caption transcript attached to a Town Hall recording is not a compliant caption file — it is a starting draft.

Post-event transcript access and download

After a Teams Live Event or Town Hall, the auto-generated transcript is accessible to the event organiser and, depending on permissions configuration, to attendees. The organiser can download the transcript in .VTT format from the Teams admin centre recording management interface. This is the starting point for the caption remediation workflow:

1. Download the auto-generated .VTT from the event recording page in Teams or SharePoint
2. Submit the .VTT and the recording (or audio file) to the caption pipeline with the organisation’s company glossary loaded
3. Receive the corrected, WCAG 2.1 AA-compliant VTT
4. Replace the auto-generated VTT on the event recording page with the corrected VTT
5. Verify that the corrected VTT is served to attendees who access the recording after the event

The critical timing constraint is the accommodation window. If a hearing-impaired employee was in the Town Hall live (relying on the real-time auto-captions, which were also inaccurate), they may request access to an accurate post-event transcript. The ADA expectation for providing an accommodation in a timely manner means this corrected VTT needs to be available quickly — not weeks after the event. Building the VTT download-QA-re-upload cycle into the day-after-event routine is the infrastructure solution.

Teams meeting recordings: the OneDrive vs. SharePoint routing decision

Where recordings go determines how they are accessed

Teams meeting recordings route to one of two locations depending on the meeting context and the admin policy set by the tenant administrator. Channel meetings (scheduled from a Teams channel) route to the SharePoint document library of the team that owns the channel. Non-channel meetings (ad hoc calls, calendar-scheduled meetings without a channel context) route to the OneDrive personal drive of the person who initiated the recording.

This routing distinction is the first compliance gap. Content saved to a SharePoint team site is discoverable, accessible to site members, and visible in Viva Learning if the SharePoint site is connected as a learning content source. Content saved to a meeting organiser’s personal OneDrive drive is discoverable only to the organiser and explicitly shared recipients. Neither location provides a caption review workflow. Neither enforces any accuracy standard. But the OneDrive case is significantly harder to inventory and audit for compliance purposes because the content is not in a shared library — it is in an individual’s personal storage.

For remote and hybrid teams where Teams is the primary communication and informal training infrastructure, the volume of training-relevant recordings in personal OneDrive drives can be substantial. A manager who records a weekly product update with the team, saves it to their OneDrive, and shares the link via Teams chat has produced a training recording that exists outside every caption compliance inventory the L&D team has.

The tenant-level solution: route all recordings to SharePoint

The administrative fix for the OneDrive routing problem is to configure the Teams meeting recording policy to save all recordings to SharePoint rather than OneDrive. This is a tenant-level policy set in the Teams admin centre under Meeting Policies → Recording and Transcription → Meeting recording storage. When recordings route to SharePoint, they are accessible to the site members and discoverable to the L&D team’s caption compliance inventory process.

The practical challenge is that this policy applies to all Teams meetings, not only to meetings that constitute training content. Routing all recordings to SharePoint means that confidential 1:1 meetings, performance conversations, and executive planning sessions also route to a shared SharePoint location. Most tenants run a hybrid policy: channel meetings (which are inherently team-accessible) route to SharePoint, non-channel meetings default to OneDrive with explicit guidance to meeting organisers to move training-relevant recordings to the appropriate SharePoint site after the meeting.

The guidance-based approach requires a clear definition of what constitutes a “training-relevant recording” in the caption policy. Without that definition, meeting organisers will make inconsistent judgments about what needs to be captioned. The policy gap is not just about Microsoft 365 — it reflects a broader ambiguity about what the organisation’s caption policy applies to. Extending the caption policy scope explicitly to cover Teams recordings — as discussed in the compliance policy gap section below — is the prerequisite for making the routing guidance enforceable.

Auto-generated VTT format and its limitations

When Teams generates a transcript for a meeting recording, it produces a .VTT file that is stored alongside the recording in SharePoint or OneDrive. The VTT is accessible via the recording player’s “Transcript” panel and downloadable from the recording file’s details pane in SharePoint. The auto-generated VTT has four structural limitations relevant to WCAG compliance:

1. Speaker attribution without correction. Teams VTT includes speaker names when it can identify them from meeting participants. Speaker identification accuracy is separate from transcription accuracy — misidentified speakers produce captions that attribute words to the wrong person, which is a different kind of error from simple word-level inaccuracy but equally problematic for content that requires attribution (compliance training where a speaker’s authority matters, for example).
2. No punctuation inference. Auto-generated VTT from Teams meeting transcripts frequently lacks punctuation, particularly for extended sentences or run-on speech. Missing punctuation makes captions harder to read and degrades the experience for hearing-impaired users reading captions as their primary access channel.
3. No glossary or domain vocabulary. The Azure Cognitive Services transcription model that generates Teams captions uses a general language model. It has no awareness of the organisation’s product names, internal terminology, regulatory acronyms, or domain vocabulary. A GlossCap Team plan or Org plan glossary applied to the same audio would produce substantially better accuracy on the technical terms that matter most.
4. Timestamp drift. For long recordings, VTT timestamps can drift slightly from the actual audio timing. This is rarely perceptible in short meeting segments but becomes noticeable in hour-long Town Halls or extended training sessions where the caption timing drifts a half-second from the audio. Correcting timestamp drift is part of the caption QA workflow, not just word-level accuracy correction.

The download-QA-re-upload workflow applies here in the same way it applies to Town Hall recordings. The distinction is volume: most organisations generate far more Teams meeting recordings than Town Halls. A prioritisation decision — which recordings require caption QA before sharing — is necessary. The caption policy for training content should specify the threshold: any recording that will be assigned as required viewing for one or more employees requires corrected captions before the assignment is made.

SharePoint Stream video libraries: no accuracy standard, no audit workflow

How SharePoint Stream caption generation works

SharePoint Stream (on SharePoint) — the current iteration of Microsoft’s video platform, following the retirement of classic SharePoint Stream — automatically generates captions when a video file is uploaded to a SharePoint document library and played through the Stream video player. The caption generation uses the same Azure Cognitive Services speech recognition infrastructure as Teams, with the same accuracy limitations.

Unlike Teams meeting recordings, which generate transcripts as a byproduct of the recording process, SharePoint Stream caption generation happens on upload. When a video is uploaded to a document library, the SharePoint backend submits it for automatic transcription. The resulting VTT is attached to the video and served when the Stream player is opened. For the typical SharePoint video library — a department SharePoint site where an SME uploads a screen recording of how to use the new expense reporting system — the auto-generated captions are the only captions that will ever exist unless someone explicitly downloads, corrects, and re-uploads them.

The SharePoint Stream caption architecture has three compliance problems that LMS caption workflows do not typically encounter:

• No upload-time review gate. LMS platforms that have been configured with a caption compliance policy enforce review at upload: the content owner must attach a caption file before publishing. SharePoint document libraries have no such gate. Video files can be uploaded, shared, and assigned as training content without any caption review step.
• No audit log for caption accuracy. SharePoint tracks file access and version history, but it does not log whether the captions attached to a video were auto-generated or manually reviewed, or what accuracy review process they went through. An audit of the SharePoint video library for caption compliance evidence would need to examine individual VTT files rather than rely on any platform-level audit trail.
• No glossary support. SharePoint Stream caption generation does not accept a custom vocabulary or glossary. The domain-specific accuracy problem that affects Teams auto-captions applies here equally. A SharePoint video library used by a pharmaceutical L&D team to host GxP compliance training will have systematically inaccurate captions on drug names, process identifiers, and regulatory citations regardless of the upload configuration.

Downloading the auto-VTT from SharePoint Stream

The SharePoint Stream player includes a “Transcript” panel that appears when auto-generated captions are available. From this panel, the caption file can be downloaded as a .VTT. The download path varies slightly between SharePoint Online UI versions, but the general flow is:

1. Open the video in the SharePoint document library using the Stream player (click the video thumbnail)
2. In the Stream player, click the “Transcript” or “CC” icon in the player toolbar
3. In the transcript sidebar, click the download or export option (represented as a download icon or “…” menu)
4. Save the .VTT file to a local working directory
5. Submit the .VTT and the original video to the caption pipeline for accuracy QA

After the corrected VTT is returned from the caption pipeline:

1. Return to the video in SharePoint
2. Open the file details pane (right-click the video file → Details)
3. In the Captions section, upload the corrected VTT file
4. Set the language on the VTT to English (or the appropriate language)
5. Verify that the Stream player now serves the corrected captions

If multiple caption files are attached to a video in SharePoint Stream (for example, the auto-generated file and the corrected file both exist), the player will present a language selection to the viewer. It is best practice to delete the auto-generated file before uploading the corrected version to avoid viewer confusion. SharePoint document library file management (delete the old .VTT from the file details) handles this.

Volume management: which SharePoint videos need caption QA?

A large SharePoint tenant may have thousands of videos distributed across hundreds of document libraries. Prioritising which videos require caption QA before the compliance audit is a risk-based exercise analogous to the caption compliance self-assessment prioritisation framework. The high-priority categories for SharePoint video library caption QA are:

• Videos on SharePoint sites that are designated as training or onboarding resources
• Videos embedded in SharePoint pages that are linked from any LMS course or Viva Learning assignment
• Videos on sites accessible to groups that include employees with known hearing accommodations
• Videos that have been viewed more than a threshold number of times (SharePoint analytics provides view counts), indicating they function as de facto training resources even if not formally designated as such

The remaining videos — casual meeting recordings, project status updates, informal team communication — are lower priority but not zero priority. The organisation’s caption policy should specify a policy horizon for legacy SharePoint video library content: a phased remediation timeline that addresses high-priority content first and works through the library over a defined period. This is the same structure that OCR has accepted for legacy LMS content remediation, and it applies equally to SharePoint video libraries.

Microsoft Viva Learning: caption quality inheritance and the LMS connector

What Viva Learning does and does not do with captions

Microsoft Viva Learning is an employee learning aggregation surface built into Microsoft Teams. It pulls learning content from multiple sources — configured LMS providers (Cornerstone, SAP SuccessFactors, Saba, 360Learning, and others via Microsoft’s Learning Management System connector framework), SharePoint sites designated as learning content repositories, and Microsoft’s own content catalogue (including LinkedIn Learning). The aggregated content appears in a Teams app and in a personal learning tab where employees can browse, bookmark, and complete assignments.

Viva Learning’s caption handling is purely passive: it displays whatever captions came with the content from its source. Viva Learning does not generate captions. It does not apply its own speech recognition to content. It does not verify or validate the accuracy of captions before surfacing content in an employee’s learning feed. The caption quality of content in Viva Learning is entirely determined by the source — the LMS, the SharePoint library, or the LinkedIn Learning platform — before Viva Learning ever touches it.

This has a direct implication for caption compliance strategy. If your organisation uses Viva Learning, the caption compliance audit needs to trace content to its source rather than auditing Viva Learning directly. A Viva Learning assignment that surfaces an LMS course via the Cornerstone connector is captioned at Cornerstone’s caption quality, not at any Viva Learning standard. A Viva Learning assignment that pulls from a SharePoint site is captioned at SharePoint Stream’s auto-generated quality unless the SharePoint VTT has been corrected. A Viva Learning assignment that surfaces LinkedIn Learning content is captioned at LinkedIn Learning’s caption quality, which is variable and not verified against WCAG 2.1 AA by the platform.

The LMS connector: caption pass-through

When Viva Learning connects to an LMS via the Microsoft Learning Management System connector, it imports content metadata (course titles, descriptions, completion status, thumbnails) and streams the actual course content from the LMS rather than copying it. The employee viewing a course through Viva Learning is actually viewing it through the LMS player, embedded in the Viva Learning interface. Captions in this model are delivered by the LMS player, not by Viva Learning itself.

The compliance implication is positive: if your LMS content is captioned to WCAG 2.1 AA standard — for example, via a glossary-biased caption workflow that delivers corrected SRT and VTT to your LMS — that caption quality flows through to Viva Learning automatically. No additional caption work is required at the Viva Learning level for LMS-sourced content. The caption compliance infrastructure you have built for LMS delivery covers Viva Learning delivery as well, provided the LMS connector is the delivery path.

The risk case is SharePoint-sourced Viva Learning content. When Viva Learning is configured to surface content from a SharePoint site (via the SharePoint knowledge base learning content configuration), it surfaces videos with whatever captions are attached at SharePoint. This is the loop that closes: LMS content with corrected captions flows through cleanly; SharePoint content with auto-generated captions flows through with the auto-generated accuracy, not the corrected accuracy. The SharePoint VTT correction workflow described above is the prerequisite for WCAG 2.1 AA compliance on Viva Learning assignments sourced from SharePoint.

LinkedIn Learning in Viva Learning: the third-party caption problem

Many organisations connect LinkedIn Learning to Viva Learning via Microsoft’s native LinkedIn Learning integration, surfacing LinkedIn Learning content directly in the Viva Learning employee feed. LinkedIn Learning operates its own captioning programme: most courses have captions, and LinkedIn Learning has published a commitment to captioning its content library. The quality of those captions is variable. LinkedIn Learning courses with instructor-uploaded captions are typically accurate for the presenter’s content; LinkedIn Learning courses that rely on the platform’s auto-generated captions are subject to the same accuracy limitations as any auto-caption system.

The ADA compliance picture for LinkedIn Learning content assigned as training has been covered in detail elsewhere in this corpus (see the post on third-party content provider caption obligations). The key point for Viva Learning planning is that LinkedIn Learning’s captions cannot be audited or corrected by the organisation using the Viva Learning connector. The organisation receives whatever captions LinkedIn Learning attaches to the course. If those captions are inaccurate for a critical word (a drug name in a pharmaceutical compliance course, an accessibility standard citation in an HR training module), the organisation has no corrective pathway through Viva Learning or LinkedIn Learning — only an inquiry to LinkedIn Learning to correct the course or a decision to source equivalent captioned content elsewhere. This is a systemic limitation of third-party content platforms that is not unique to LinkedIn Learning and is not solvable at the Viva Learning configuration level.

The practical implication: for required compliance training or role-mandatory professional development, relying on third-party platform content (LinkedIn Learning, Coursera, Udemy Business) surfaced through Viva Learning means accepting the platform’s caption quality rather than controlling it. For content where accuracy is material — regulatory compliance, safety training, anything with a legal obligation attached — L&D teams are better served by producing the content internally or by a vendor whose caption output can be verified, rather than by assigning LinkedIn Learning courses through Viva Learning and assuming the captions meet WCAG 2.1 AA.

Three workflows for getting Microsoft 365 content into a formal WCAG caption pipeline

Workflow A: Teams recording → SharePoint VTT download → glossary QA → corrected VTT re-upload

This workflow applies to Teams meeting recordings and Teams Live Event / Town Hall recordings that are stored in SharePoint after the event. It is the highest-volume workflow for most enterprises because it addresses the largest content category: informal training content produced through Teams.

Step 1 — Ensure recording routes to SharePoint. For channel meetings, confirm that the Teams recording policy routes the recording to the channel’s SharePoint document library. For non-channel meetings, configure the meeting recording policy at the tenant level to route to SharePoint where appropriate, or establish guidance for meeting organisers to move training-relevant recordings from OneDrive to the designated training content SharePoint site within 24 hours of the meeting. The L&D team maintains a SharePoint site or document library specifically designated for training recordings — this is the inventory point for caption compliance review.

Step 2 — Trigger caption generation. Teams automatically generates a transcript VTT for recordings. Confirm the transcript is present in the SharePoint recording file’s details pane before proceeding. If the transcript has not been generated (which can happen for recordings below a length threshold or for recordings where transcription was disabled in the meeting settings), open the video in Stream and check whether manual transcription can be initiated.

Step 3 — Download the auto-generated VTT. From the Stream player on the SharePoint recording, open the Transcript panel and download the .VTT. Save it with the recording file name for traceability.

Step 4 — Submit to the caption pipeline with glossary. Submit the VTT and the recording audio to the caption pipeline. If using GlossCap, the company glossary (product names, internal terminology, regulatory citations) is applied during the accuracy QA step. The result is a corrected VTT with glossary-specific terms corrected to the exact spelling in the glossary. For a tech company, this means SDK names, API identifiers, and product feature names that Azure Cognitive Services would have substituted with phonetic approximations are now correct.

Step 5 — Upload corrected VTT to SharePoint. Delete the auto-generated VTT from the SharePoint file’s caption slot. Upload the corrected VTT. Verify in the Stream player that the corrected captions are served. Log the caption QA completion in the team’s caption tracking sheet (date, recording, word error rate before/after, reviewer).

Step 6 — Share the captioned recording. Only after the corrected VTT is in place should the recording link be shared with employees as required viewing or assigned through Viva Learning. The sharing step is the compliance gate: uncaptioned or auto-captioned recordings should not be shared as training assignments before caption QA is complete.

The cycle time for this workflow depends on the caption pipeline turnaround. Same-day turnaround from the recording to corrected VTT delivery is achievable for content up to 60 minutes. For most informal training recordings (15–30 minutes), a 4–8 hour turnaround from recording end to corrected VTT upload is feasible. This means a recording from a Tuesday afternoon meeting can have corrected captions by Wednesday morning — a reasonable timeline for most training assignment contexts.

Workflow B: SharePoint Stream library video → transcript download → caption pipeline → VTT upload to library

This workflow addresses the stock of videos already in SharePoint document libraries that were uploaded without going through the Teams recording process. These are typically screen recordings, narrated slide presentations, SME how-to videos, and other informal content that was recorded outside Teams and uploaded directly to SharePoint. The async video content common in remote and hybrid work environments lands heavily in this category.

Step 1 — Inventory the SharePoint video library. Use SharePoint search with the file type filter for video (MP4, MOV, WMV) combined with the site scope for the training content libraries. SharePoint site analytics provides a view-count report for video files, which is useful for prioritising the inventory: videos with higher view counts are likely functioning as de facto training resources even if they were not formally designated as such.

Step 2 — Check whether auto-generated captions exist. For each video in the priority inventory, open it in the Stream player and check whether the Transcript panel contains a transcript. Videos uploaded before SharePoint Stream’s auto-transcription feature was enabled may have no captions at all — not even auto-generated ones. Videos uploaded recently should have auto-generated transcripts if the language is supported. The caption status (none, auto-generated, manually reviewed) is the starting point for the QA workflow.

Step 3 — Download auto-VTT or generate from audio. If auto-generated captions exist, download the VTT as described in the SharePoint Stream section above. If no captions exist, the caption pipeline will need to generate them from scratch using the video audio. Submit the video file (or extracted audio) to the caption pipeline with the company glossary.

Step 4 — Return corrected VTT and upload. Upload the corrected VTT to the SharePoint file, replacing any existing auto-generated VTT. For videos with no prior captions, this is an add rather than a replace.

Step 5 — Tag the file for audit traceability. SharePoint custom metadata columns can be used to add caption compliance tracking fields to document library video files: a “Caption Status” column (Auto-generated / QA Complete / Uncaptioned), a “Caption QA Date” column, and a “Caption Reviewer” column. These columns create the WCAG 2.1 AA compliance documentation trail that an OCR complaint review or ADA litigation discovery process would look for. SharePoint list views can then filter on Caption Status to identify all QA-complete vs. pending videos at a glance.

The custom metadata approach is the SharePoint-native equivalent of the caption tracking infrastructure that LMS platforms typically provide as part of their course management workflow. It is the closest approximation of a compliance audit trail for SharePoint video library content without deploying a third-party SharePoint video governance tool.

Workflow C: Viva Learning LMS connector → ensure LMS captions pass WCAG before Viva surfaces them

This workflow applies when Viva Learning is configured to surface content from an LMS via the Microsoft Learning Management System connector. The caption compliance work happens entirely at the LMS level; Viva Learning configuration is just the verification step.

Step 1 — Identify all LMS content sources connected to Viva Learning. Viva Learning connectors can pull from multiple LMS providers simultaneously. Map each connector: which LMS, which content catalogues within that LMS, and whether those catalogues include video content that would require captions.

Step 2 — Audit caption quality on LMS video content before Viva surfaces it. Run the LMS caption compliance review on the content that will be surfaced through Viva Learning before activating the connector for that content catalogue. The LMS caption ingestion workflow that covers Cornerstone, Docebo, Kaltura, TalentLMS, and other platforms applies here: ensure that every video course in the connected catalogue has a compliant VTT or SRT file before the Viva Learning connector surfaces it in employee learning feeds. The caption discoverability benefits of well-structured caption files also apply to Viva Learning content surfacing — accurate captions improve the metadata quality that Viva Learning uses to surface content in learning recommendations.

Step 3 — Configure the Viva Learning connector for the audited content catalogues. Connect only the content catalogues where LMS caption compliance has been verified. If a catalogue has mixed caption status (some courses captioned, some not), either complete caption QA on the full catalogue before connecting or configure the LMS to restrict the Viva Learning connector scope to the captioned subset.

Step 4 — Monitor for new content added to LMS catalogues surfaced through Viva Learning. New LMS courses added to a connected catalogue will automatically appear in Viva Learning unless the connector is configured with content filters. Establish a pre-publication caption verification step in the LMS course publication workflow: no course can be published to a Viva Learning-connected catalogue until caption QA is complete. The SCORM and xAPI caption delivery tracking framework provides the publication gate infrastructure at the LMS level.

Step 5 — Handle SharePoint-sourced Viva Learning content separately. If Viva Learning is also configured to surface content from SharePoint sites (via SharePoint knowledge base learning content), run Workflow B for those SharePoint libraries before enabling the Viva Learning integration for that SharePoint source. SharePoint-sourced and LMS-sourced content in Viva Learning require separate caption compliance workflows because the caption architecture of the two sources is different.

The Microsoft 365 caption compliance policy gap

Why the LMS caption policy does not cover Microsoft 365

The typical enterprise caption policy was drafted when the LMS was the only significant video training surface. It specifies that all video content published to the LMS must have WCAG 2.1 AA captions before publication, defines the review process, names the approving authority, and sets the turnaround time for caption production. This is a sound policy for LMS content.

The policy contains a structural gap: it defines scope as “video content published to the LMS.” Teams recordings, SharePoint videos, and Viva Learning assignments are not “published to the LMS.” They fall outside the policy scope by definition, not by oversight. When the policy was written, the implicit assumption was that the LMS was the complete universe of formal training content delivery. That assumption no longer holds.

The result is a dual-standard reality that many enterprise L&D teams now operate under without recognising it: LMS courses have compliant captions because the caption policy requires it; Teams recordings and SharePoint videos do not have compliant captions because the caption policy does not cover them. Both categories of content are used for the same purpose — training employees — and both are subject to the same ADA obligations. The compliance exposure from the informal video surface is often larger than the remaining LMS compliance exposure precisely because it has been unmanaged.

Extending the caption policy to cover Microsoft 365

The policy extension requires two definitional changes and one operational addition. The definitional changes:

1. Expand the definition of “training content” to include “any video recording, live session, or on-demand video made available to employees in connection with their employment, professional development, or compliance obligations, regardless of the platform on which it is delivered or stored.” This definition explicitly brings Teams recordings, SharePoint videos, and Viva Learning assignments into scope.
2. Expand the definition of “training content platforms” to list, by name, Microsoft Teams (including Live Events and Town Halls), SharePoint (including Stream video libraries), and Microsoft Viva Learning in addition to the named LMS platforms.

The operational addition is a pre-distribution caption verification step for Microsoft 365 content: “Before distributing a Teams recording, SharePoint video, or Viva Learning assignment to employees for training purposes, the content owner must confirm that WCAG 2.1 AA-compliant captions are in place on the content. Distribution of training content without compliant captions is not permitted regardless of the urgency of the distribution.”

The “regardless of urgency” clause is important. The most common reason Microsoft 365 training content ships without captions is urgency: a manager records a training session, wants to share it immediately, and the caption QA step is a delay they choose to skip. The caption policy must be explicit that urgency is not a basis for distributing non-compliant content. The operational solution to urgency is short caption pipeline turnaround time (achievable with the right tooling), not policy exceptions.

The governance structure for Microsoft 365 caption compliance

LMS caption compliance is typically governed by L&D operations: the instructional designer or content publisher who uploads the course is the checkpoint. Microsoft 365 caption compliance requires a different governance structure because the content owners are distributed — they are managers, SMEs, and department leads who record Teams meetings without going through L&D operations at all.

The governance structure that works for Microsoft 365 training content is a content owner accountability model supported by L&D infrastructure:

• Content owner accountability. The person who records the Teams meeting or uploads the SharePoint video is responsible for ensuring it is captioned before sharing it as training. This is the same model that applies to any other training content — the creator owns the compliance obligation. The caption policy makes this explicit.
• L&D infrastructure support. L&D operations provides the caption pipeline, the SharePoint upload workflow documentation, and the glossary (so content owners do not need to manage terminology themselves). The barrier to caption compliance is reduced to: submit the file, receive the corrected VTT, upload it. The L&D team does not need to produce every caption file centrally; it needs to provide the infrastructure that lets content owners do it efficiently.
• Teams channel and SharePoint site configuration. L&D operations works with the IT administrator to configure Teams channel meeting recording policies and SharePoint library caption metadata columns as described in the workflows above. These are one-time configuration steps that create the infrastructure for the accountability model.
• Periodic audit. L&D operations or the accessibility coordinator runs a quarterly review of SharePoint video libraries in training-designated sites and Teams recording libraries to confirm caption status metadata is current. The review does not need to sample every video — it can use the custom metadata columns to identify uncaptioned or auto-captioned-only videos in the training content libraries and trigger a remediation queue.

This governance structure mirrors the building-caption-compliance-programme framework for LMS content but adapted to the distributed content-owner reality of Microsoft 365. It is documented in the caption compliance programme build guide as the informal channel extension to the programme scope.

The audit trail gap: what regulators will ask for

An OCR complaint or ADA litigation discovery process will ask for documentation of the organisation’s caption compliance programme, evidence of caption compliance on the specific content at issue, and evidence of the process by which compliance is maintained. For LMS content, this documentation typically exists: the LMS tracks caption file attachment, publication dates, and course completion records. For Microsoft 365 content, this documentation does not exist by default unless it has been explicitly created.

The SharePoint custom metadata column approach described in Workflow B is the nearest equivalent of an audit trail for SharePoint video library content. For Teams recordings, a Teams recording inventory maintained in a SharePoint list (with columns for Recording Title, Recording Date, Meeting Organiser, Caption Status, Caption QA Date, VTT File Path) provides the documentation infrastructure for an audit response.

Neither of these approaches is as automated as LMS audit reporting. They require active maintenance by L&D operations or the accessibility coordinator. But they exist, they are discoverable, and they provide the evidence that “the organisation has a documented process for caption compliance on Microsoft 365 training content” — which is the standard OCR and courts apply in assessing whether a compliance programme is adequate. The absence of any documentation on Microsoft 365 caption compliance is harder to defend than an imperfect but documented programme in active use.

Putting it together: the Microsoft 365 caption compliance checklist

Translating the above into an actionable checklist for L&D operations and accessibility coordinators:

Policy

• Caption policy explicitly names Microsoft Teams, SharePoint Stream, and Viva Learning as covered platforms
• Caption policy defines “training content” to include recordings and informal video regardless of platform
• Caption policy designates a content owner accountability model for Microsoft 365 content
• Caption policy prohibits distribution of training content without compliant captions regardless of urgency

Infrastructure

• Teams channel meeting recording policy routes recordings to SharePoint team site (not OneDrive for training-designated channels)
• SharePoint training content libraries have custom metadata columns: Caption Status, Caption QA Date, Caption Reviewer, VTT File Path
• Organisation glossary is loaded in the caption pipeline and available for Teams recording and SharePoint video caption QA
• Caption pipeline has documented turnaround time commitment (target: same-day for recordings up to 60 minutes)
• Viva Learning connectors are scoped to LMS catalogues where caption compliance has been verified
• SharePoint-sourced Viva Learning content has been reviewed for caption compliance before Viva Learning surfacing is activated

Process

• Teams Town Hall and Live Event recordings: VTT downloaded, submitted for QA, corrected VTT re-uploaded within 24 hours of event end
• Teams meeting training recordings: VTT downloaded, submitted for QA, corrected VTT re-uploaded before recording is shared as required viewing
• SharePoint video library: new uploads flagged for caption QA, Caption Status set to Auto-generated until QA is complete, QA Complete after corrected VTT is uploaded
• New LMS courses connected to Viva Learning: caption QA verified in LMS before Viva Learning surfacing is enabled
• Quarterly audit: SharePoint libraries filtered by Caption Status = Uncaptioned or Auto-generated-only to identify backlog

Documentation

• Teams recording inventory list with Caption Status tracking
• SharePoint video library metadata columns populated and current
• Caption QA log (date, recording, WER before/after, reviewer, VTT file path) retained for at least three years
• Caption policy document version-controlled with effective date and approving authority