Copilot Cowork landed. Your SharePoint still needs to be ready first.

Microsoft updated Copilot Cowork on 5 May 2026. The new chapter brings Cowork to iOS and Android, ships a wave of plugins (Power BI via Fabric IQ, Dynamics 365 sales, customer service, and ERP, plus third-party connectors for LSEG, Miro, monday.com, and S&P Global Energy), and sharpens the emphasis on Skills as the primary unit of reusable workflow.

The thesis I want to land in this article is the one most other Cowork pieces are missing: Cowork makes your SharePoint foundation more important, not less. The same hygiene that determines whether regular Copilot answers correctly determines whether Cowork picks the right document, the right version, and the right context for the user's task. Cowork is more confident than regular Copilot. That means it surfaces wrong answers more confidently too.

This article covers what is actually new in the May 2026 update, why Cowork amplifies the SharePoint readiness story rather than replacing it, the five things in your tenant that determine Cowork accuracy, the mobile and plugin angles, and the thirty-minute pre-Cowork audit you can run today.

What's actually new in Cowork (May 2026)

A short recap because most readers are arriving with the new headline in their head.

Cowork itself is not new. It launched 9 March 2026 and became available through the Microsoft 365 Frontier program on 30 March. The May 5 update is the next chapter and it adds three things.

Mobile. Cowork is now on iOS and Android. Before May 2026 it lived only on the desktop and the web. Users can now trigger Cowork tasks on the train, between meetings, or at a customer site.

Plugins. Cowork now ships native integrations with Power BI (via Fabric IQ), Dynamics 365 (sales, customer service, ERP), and third-party connectors for LSEG, Miro, monday.com, and S&P Global Energy. Custom plugins are also possible. Microsoft is signalling that Cowork wants to reach further than the M365 surface alone.

Skills emphasis. The May 5 announcement positions Skills as one of the three pillars (Skills, integrations, devices). The 13 built-in Cowork Skills and the 20 custom Skills per user pattern is now the central story for reusable Cowork workflows. The deep dive on the Skills landscape across Cowork, AI in SharePoint, and Copilot Studio lives on the sharepointaiskills.com Skills comparison piece.

What did not change in the May 2026 update: the underlying dependency on Work IQ, the SharePoint grounding model, the permission inheritance, the licensing, or the readiness work. All of that is the same.

Why Cowork makes SharePoint readiness more important, not less

Most Cowork articles position it as a feature. This one positions it as a force multiplier on whatever your tenant looks like today.

Cowork is more accurate than regular Copilot for two reasons. First, it has multi-step planning, so it can chain multiple SharePoint queries rather than relying on one shot. Second, it operates over Work IQ, which gives it richer context (calendar, email, Teams, files, relationships) than a single Copilot prompt would have.

That accuracy is wonderful when your foundation is solid. When the foundation is messy, the same accuracy works against you. Cowork picks the wrong document with confidence. It cites a draft policy from 2022 because that draft was the best text match, then writes an executive briefing on top of that draft, then sends a Teams message announcing the briefing is ready. Three steps of compounding error from one bad piece of SharePoint hygiene.

Regular Copilot would have given the same wrong answer, but the user would have been one step closer to noticing. Cowork makes the steps invisible. The user sees the finished output, not the bad source.

The fix is the same fix it has always been. Better metadata. Tighter permissions. One source of truth per topic. The work has not changed. The consequences of skipping it have.

Work IQ reads SharePoint. Crap SharePoint, crap Cowork.

Microsoft's official guidance on the secure foundation for Copilot is consistent on this point: Work IQ is the intelligence layer; it draws on Microsoft Graph signals, including SharePoint content; it scopes to the asking user's permissions.

In practice this means Cowork's quality is bounded by the quality of what it can read. The text content of your documents, the metadata around them, the freshness, the ownership signals, the permissions structure. All of it feeds Cowork's planning and Cowork's outputs.

The implication is simple. Every piece of advice in the SharePoint metadata for Copilot article applies to Cowork. Every permissions cleanup item in the SharePoint permissions and Copilot oversharing article applies to Cowork. The diagnostic framework in why Copilot can't find your SharePoint files applies to Cowork.

Nothing in the May 2026 update changes that. Cowork inherits the foundation it sits on.

The five things in your tenant that determine Cowork accuracy

These are the items I check first in any tenant before turning Cowork on for broader use. None of them are Cowork-specific. All of them have outsized impact on whether Cowork answers correctly.

Document Status column populated. Without it, Cowork reaches for any version it can find. With it, Cowork prefers Approved over Draft and surfaces Archived only when the user explicitly asks for historical content. The five-column metadata pattern is in SharePoint metadata for Copilot.

Owner column populated. Cowork uses Owner as a confidence signal when it cites a document and as the contact reference when answering follow-up questions. A document with no Owner is treated as orphaned content and deprioritised in Cowork's output.

Permissions audited and tightened. Cowork inherits the user's permissions and surfaces what they technically have access to. Broad permissions become Cowork-discoverable. The permissions cleanup framework covers the four highest-risk patterns to fix first.

Folder depth flattened where possible. Cowork reads metadata better than folder paths. A document three folders deep with no metadata is a weaker retrieval candidate than a flat library with proper columns. The library setup pattern is in how to set up a SharePoint document library.

Stale document cleanup. The "two truths" problem hits Cowork harder than regular Copilot because Cowork ships answers, not just citations. Two versions of the same policy in different libraries means Cowork picks one and writes a briefing on top of it. The diagnostic for this pattern is in why Copilot can't find your SharePoint files.

Cowork on mobile changes the question

This is the section most other Cowork articles skip and it is the most practically important part of the May 2026 update.

Cowork on iOS and Android means users will trigger document creation, email sends, and Teams posts from contexts where they are not double-checking the source. The user is on a train. They are between meetings. They are at a customer site. They will accept Cowork's draft because they do not have time to verify it.

If your SharePoint has stale documents or wrong metadata, the consequences of bad citations multiply on mobile. A wrong answer reviewed on a phone screen is harder to spot than a wrong answer reviewed on a desktop with the source document open in another tab. The user trusts the small confident answer more, not less.

The practical implication: do the readiness work BEFORE you let mobile Cowork loose on your tenant. The mobile experience is good enough that it will get used. Make sure what it surfaces is worth surfacing.

Cowork plugins (Power BI, Dynamics, third-party) and where SharePoint still fits

The plugin announcement is the part that sounds most disruptive and is actually the least so for SharePoint readiness.

Plugins extend Cowork's reach into Power BI, Dynamics, and third-party systems. They give Cowork structured data sources that are independent of SharePoint. So at first read, plugins look like they reduce the importance of SharePoint as the grounding layer.

The reality: plugins handle structured data. SharePoint handles unstructured content. The plugins do not replace SharePoint as the source of truth for documents, pages, lists, knowledge, policies, contracts, and the long tail of organisational content. They add new sources alongside SharePoint, not instead of it.

For most teams, the work Cowork does will continue to lean heavily on SharePoint content because that is where the unstructured knowledge lives. Plugins are a useful expansion. They do not change the readiness conversation.

The deeper compare-and-contrast on Cowork plugins versus Copilot Studio agents is a different article (Copilot Studio still has the broader connector library and the audit features). For SharePoint readiness, plugins are interesting but not load-bearing.

The thirty-minute pre-Cowork audit

If you have one window to act before broader Cowork rollout, this is what to do.

1. Open the top three libraries Cowork users will hit first. For most teams that is HR policies, project documents, and customer records.
2. Check Document Status, Owner, and Document Type columns are populated for the top 50 most-viewed documents in each library.
3. Run three test prompts in Cowork against each library. Document the answers and the cited sources.
4. Note any "I didn't expect Cowork to find that" moments. These are oversharing flags. Investigate the permissions on each surfaced item.
5. Decide whether to expand the Cowork rollout or fix the surfaced issues first.

The output is a short list of specific fixes for each library, prioritised by frequency of access. Fix the high-impact items before the broader rollout. The long tail can be picked up by AI in SharePoint autofill in the background.

The full six-week version of this work is in How to prepare SharePoint for Microsoft Copilot. The thirty-minute version above is the abbreviated pre-Cowork sprint.