A Skill on a project's SharePoint site can read three weekly reports, identify recurring themes, summarise subcontractor performance, list outstanding actions, and produce an overall assessment. The user types one sentence. The Skill does the rest. The same site demo also shows updating list metadata through chat and applying conditional formatting with no JSON.
Three capabilities in nine minutes of video, all built on the same Skill pattern. This is the worked example for teams who want to see Skills run end to end on real content.
What changes when AI works your way, not the other way
Most teams hit the same wall with Copilot in SharePoint in the first few weeks. The AI is capable but generic. Ask for a status summary and you get a status summary, but it has the wrong shape, missing sections, or includes things the team does not need.
The fix is Skills. A Skill captures the shape of work the team wants every time. The cross-report analysis demo shows this exactly: same input, same shape of output, every run, regardless of who runs it.
If you want the full explainer on what Skills are and how to build them, the SharePoint AI Skills deep-dive covers anatomy, authoring, and the comparison to Copilot Studio agents. This article is the worked walkthrough.
The cross-report analysis Skill
The Skill lives at /Agent Assets/Skills/cross-report-analysis/SKILL.md on a project's SharePoint site. It defines the structure for combining multiple weekly site reports into one cross-week summary.
The five output sections defined in the SKILL.md:
- Work completed by the weeks. What got done in each week of the analysis.
- Issue log recurring themes. Issues that appeared in more than one week, with the pattern across weeks.
- Subcontractor performance notes. Observations about each subcontractor across the time range.
- Outstanding actions. Items still open at the end of the time range, with how long they have been open.
- Summary assessment. Overall narrative on the project's health across the weeks.
In the demo, the user opens chat and says: "I want cross-report analysis over three weeks, weeks five, six, and seven."
AI in SharePoint loads the cross-report-analysis Skill because the request matches the trigger conditions defined in the SKILL.md's When to use section. The user does not have to invoke the Skill by name; AI in SharePoint figures it out.
The Skill then:
- Looks at the site reports library metadata
- Finds weeks 5, 6, and 7
- Reads all three reports
- Applies the structure defined in the SKILL.md
- Produces the cross-week summary in the exact five sections above
The output reaches the user in under a minute. Same shape every time.
Updating list metadata through chat
The same demo includes a second pattern that uses AI in SharePoint directly, not via a Skill: updating metadata on list items through natural language.
The user types: "I know that week four has been submitted. Can you please update the week four site report status to be submitted?"
AI in SharePoint:
- Locates the week four item in the site reports list
- Checks the schema for the Status column
- Confirms the value "Submitted" is valid for the Status choice column
- Applies the update
No list settings UI, no manual edit. The user describes the change; AI in SharePoint makes the change. Existing SharePoint permissions on the list and item apply, so the user can only make changes they would have been authorised to make manually.
This is not technically a Skill (no SKILL.md file required) — it is in-built AI in SharePoint behaviour for working with list items conversationally.
Enhancing library readability without JSON
The third pattern in the demo shows AI in SharePoint applying view formatting without anyone writing JSON.
The user asks: "Can you make this list look a little bit prettier? For our choice columns, let's look at adding some pills with different colours."
AI in SharePoint:
- Identifies which columns are choice columns
- Applies pill formatting with distinct colours per choice value
- Surfaces the formatted view to the user
In SharePoint terms, this is the equivalent of writing JSON in the column or view formatting screens. Manually it takes 15-30 minutes per column and most users never get past the documentation. Through chat, it takes one sentence.
The same pattern works for:
- Conditional formatting. "Any client with a contract renewal date this month should be highlighted in red." Applied through chat, no JSON written by the user.
- Email highlighting. "Show email addresses in blue so they stand out."
- Status badges. "Add a green badge for completed items and a red badge for blocked items."
- Row colouring. "Colour-code rows by priority."
This is not the same as a Skill. It is a built-in capability of AI in SharePoint when working with lists. But it pairs naturally with Skills, because once you have a Skill that produces a list, you can make the list readable in one further chat.
How the three patterns combine
The demo shows three separate AI in SharePoint behaviours: invoking a Skill, updating metadata, and applying formatting. Each one stands alone. The combination is what makes the team's day-to-day work change.
A project manager opens chat on the site Friday afternoon. Three actions take three sentences.
Sentence one: "Run the cross-report analysis Skill across weeks 5, 6, and 7." The Skill runs, the cross-week summary appears.
Sentence two: "Update the week four report status to submitted." The list item updates.
Sentence three: "Apply pill formatting to the status column with green for submitted and amber for in-progress." The view updates.
Three sentences, three minutes of work, three jobs done. Previously: an hour of summary writing, a manual list edit, and a half-day's worth of JSON formatting that nobody on the team had time to do.
What makes the Skill work
Three things from the SKILL.md file pattern that show up in the demo.
Clear trigger conditions. The When to use section names the trigger phrases ("cross-report analysis", "across weeks", "summary across reports"). When a user request matches these, AI in SharePoint auto-loads the Skill.
Defined output structure. The five sections of the cross-week summary are named explicitly in the SKILL.md. AI in SharePoint cannot drift to a different structure because the structure is in the file.
Rules per section. The Skill defines what counts as a "recurring theme" (an issue appearing in two or more of the weeks in scope) and what counts as an "outstanding action" (open more than 14 days). These rules apply consistently every run.
The anatomy of a SKILL.md file in detail is in the Skills deep-dive. The cross-report analysis Skill is one specific shape; the same anatomy works for contract review, status reports, audits, and most other team workflows.
What to do this week
Three actions if you want to ship a Skill like this on one of your team's sites.
-
Make sure Copilot in SharePoint is enabled on the site. The admin guide for the opt-out rollout covers the PowerShell at the admin guide article.
-
Activate the Agent Assets feature on the site if you have not already. Site settings, Site collection features, find Agent Assets, activate.
-
Write the SHAREPOINT.md context file first, then write the Skill. The Skill works much better when AI in SharePoint already knows the site's library map, naming conventions, and rules. The template gets you there fast. Detail on the file is in the SharePoint.md article.
Once the context file is in place, pick the one workflow your team runs most repeatedly that has a consistent output shape. That is your first Skill. Open Copilot in SharePoint chat, describe the Skill in natural language, let AI in SharePoint generate the SKILL.md file, iterate over the first three runs.
If you want to build three Skills in your tenant in a single day with the SharePoint.md context file already in place and an adoption plan written for your team, the Copilot in SharePoint Immersive is the workshop format. We work in your real tenant with your real content, ship three Skills, and write the playbook before we close out the day.
