<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Peafowl IT Solution on Medium]]></title>
        <description><![CDATA[Stories by Peafowl IT Solution on Medium]]></description>
        <link>https://medium.com/@peafowlit?source=rss-668919ceb248------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*U6TajrtOm5H1Xm8RJS_tzA.png</url>
            <title>Stories by Peafowl IT Solution on Medium</title>
            <link>https://medium.com/@peafowlit?source=rss-668919ceb248------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 00:42:03 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@peafowlit/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[SharePoint Tools in Work IQ: Build AI Agents That Search, Create and Manage SharePoint Content]]></title>
            <link>https://medium.com/@peafowlit/sharepoint-tools-in-work-iq-build-ai-agents-that-search-create-and-manage-sharepoint-content-d5990698eee5?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/d5990698eee5</guid>
            <category><![CDATA[sharepoint-development]]></category>
            <category><![CDATA[iq-work]]></category>
            <category><![CDATA[sharepoint-work-iq]]></category>
            <category><![CDATA[sharepoint]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 19 Nov 2026 04:44:30 GMT</pubDate>
            <atom:updated>2026-11-19T04:44:30.222Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oHo8ZqUT8rBmHXms_LMfyg.png" /><figcaption><a href="https://peafowlit.com/blog/sharepoint-tools-work-iq-ai-agents/">SharePoint Tools in Work IQ: Build AI Agents That Search, Create and Manage SharePoint Content</a></figcaption></figure><p>Microsoft has opened a public preview of <a href="https://techcommunity.microsoft.com/blog/spblog/introducing-sharepoint-tools-in-work-iq/4569917">SharePoint tools in Work IQ</a>, giving AI agents more than 60 native SharePoint operations to search, create and manage content. For teams whose agents have mostly answered questions about documents, this moves agents from reading SharePoint to working in it.</p><p>Until now, many Microsoft 365 Copilot scenarios centred on retrieval: find a document, summarize it, answer a question. That is useful, but real SharePoint work also involves updating a tracker, filing a document in the right library, routing an approval or checking a version. This post covers what the preview includes, how agents reach it, what you could build, and how to read the performance figures Microsoft has shared.</p><p><em>Product scope: SharePoint Online, Microsoft 365 Copilot and Copilot Studio.</em></p><h3><strong>What’s new: SharePoint tools in Work IQ</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xlZcqz1fTFDeDUECL_HjvQ.png" /></figure><p>SharePoint tools in Work IQ are in public preview. They expose SharePoint capabilities to AI agents as native operations, so an agent can call them directly rather than as something each team has to wire up itself.</p><p>Access runs through the Work IQ MCP server, which means any agent that can talk to MCP can use the same set of SharePoint operations. Microsoft describes integration scenarios for both Copilot Studio agents and agents built by developers.</p><h3><strong>What agents can do with 60+ native operations</strong></h3><p>The preview covers more than 60 SharePoint operations. Based on Microsoft’s announcement, they span these areas:</p><p>An agent can perform different tasks in SharePoint. It can search SharePoint sites and find documents along with their details. It can create, update, and delete items in SharePoint lists. It can also manage files and folders, work with document libraries, and handle approval processes. In addition, the agent can manage document versions, allowing users to work with different versions of documents easily.</p><p>The list-item operations matter most for tracker-style scenarios, because an agent can write changes back rather than only report them.</p><h3><strong>How agents reach SharePoint: the Work IQ MCP server</strong></h3><p>The Model Context Protocol (MCP) is an open standard for giving AI agents access to tools and data. In this preview, the SharePoint operations are served through the Work IQ MCP server, so an agent connects once and can discover and call the SharePoint tools it needs.</p><p>Two build paths are in scope:</p><ul><li><strong>Copilot Studio agents:</strong> makers can bring SharePoint operations into low-code agents.</li><li><strong>Developer-built agents:</strong> teams building their own agents can connect to the same MCP server from code.</li></ul><p>In both cases the agent works with what the signed-in context is allowed to reach, so existing SharePoint permissions and governance remain an important part of the design. Confirm the exact permission model in Microsoft’s documentation before you build.</p><h3><strong>What you could build</strong></h3><p>These are example scenarios that follow from the operations above, not Microsoft-announced products.</p><ul><li><strong>Policy retrieval agent:</strong> searches policy libraries, returns the right document with its metadata, and can confirm which version is current.</li><li><strong>Tracker update agent:</strong> reads and updates list items for project, issue or request trackers when someone reports a status change in chat.</li><li><strong>Document filing agent:</strong> places files in the correct library and folder and keeps the structure consistent.</li><li><strong>Approval assistant:</strong> checks where a document sits in an approval and helps move it along.</li><li><strong>Version-aware reviewer:</strong> works from document versioning to compare or restore earlier versions.</li></ul><p>The common thread is that the agent finishes a task in SharePoint instead of handing the person a to-do list.</p><h3><strong>Microsoft’s reported results, and how to read them</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f6f5M1cyo7muPQYfhXnKIw.png" /></figure><p>In its own internal evaluation, Microsoft reported that the task pass rate rose from 70.1% to 84.6%, and that testing used 16% fewer tokens.</p><p>Treat these as Microsoft-reported internal evaluation results, not guaranteed customer outcomes. Your tenant’s content, permissions, prompts and task mix will differ. Before committing, run your own evaluation on representative tasks, such as your real trackers and libraries, and measure both success rate and token cost.</p><h3><strong>Where to start, by role</strong></h3><h4>SharePoint administrators</h4><ul><li>Review site permissions and sensitivity labels before agents can reach a library; agents amplify whatever access already exists.</li><li>Pick a low-risk pilot site first, with a clear owner.</li><li>Decide which operations, such as deleting list items, should stay out of early pilots.</li></ul><h4>Microsoft 365 and Copilot Studio developers</h4><ul><li>Connect to the Work IQ MCP server in a test tenant and inspect the available SharePoint tools.</li><li>Start with read and search tasks, then add create and update operations.</li><li>Log every write an agent makes so changes can be audited and reversed.</li></ul><h4>AI solution architects</h4><ul><li>Define which tasks need human approval before an agent writes to SharePoint.</li><li>Build an evaluation set from real tasks so you can track pass rate and token use against your own baseline.</li><li>Plan for preview limits: features and behavior can change before general availability.</li></ul><h3><strong>How Peafowl IT can help</strong></h3><p>We’re a Microsoft Solutions focused on Power Platform. If you have more agent ideas than time to build them safely, our team can take you from a first SharePoint pilot to a governed, working agent:</p><ul><li>A review of your SharePoint sites, libraries and lists to find the best candidates for agent automation, with owners and permissions checked</li><li>A pilot scope for each candidate: which operations the agent may use, and where human approval is required</li><li>Hands-on Copilot Studio agent builds that connect to SharePoint tools in Work IQ</li><li>Developer-built agent support for business-critical scenarios, with solutions and Git for version control</li><li>An evaluation plan using your own tasks, so you can measure pass rate and token cost against your baseline</li><li>Licensing and Copilot Credits setup so users aren’t blocked after go-live</li></ul><p><a href="https://peafowlit.com/get-in-touch/">Book a SharePoint agent readiness review</a></p><h3><strong>Conclusion</strong></h3><p>SharePoint tools in Work IQ turn SharePoint into something agents can act on: more than 60 native operations, reachable through the Work IQ MCP server, for both Copilot Studio and custom-built agents. The reported gains in pass rate and token use are encouraging, but they come from Microsoft’s own testing. Start with a contained pilot, measure against your own tasks, and expand as the preview matures.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d5990698eee5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Migrate your Vibe and App Builder apps before November 30]]></title>
            <link>https://medium.com/@peafowlit/migrate-your-vibe-and-app-builder-apps-before-november-30-d9a931acaa60?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/d9a931acaa60</guid>
            <category><![CDATA[vibe-code]]></category>
            <category><![CDATA[app-builder]]></category>
            <category><![CDATA[no-code-app-builder]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[low-code]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 02:31:01 GMT</pubDate>
            <atom:updated>2026-11-12T02:31:01.803Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M-2DIK2dnhsiXmSDrNpt7A.png" /><figcaption><strong>Migrate your Vibe and App Builder apps before November 30</strong></figcaption></figure><p>Microsoft is retiring the App Builder agent in Microsoft 365 Copilot and vibe.powerapps.com on <strong>November 30, 2026</strong>. Apps that aren’t migrated by then won’t be accessible. Creating new apps in both tools stopped in mid-October.</p><p>Migration doesn’t move your app. In every path, a new app is built from your old app’s source code and reconnected to the same data. Plan for a manual rebuild and don’t wait for an automatic option.</p><h3><strong>Step 1: Pick your path</strong></h3><p><strong>A. Copilot Cowork:</strong> Copilot Cowork is available within Microsoft 365 Copilot and is rebuilt using the /app skill. It is best suited for creating personal and team applications. To use this option, you need a Microsoft 365 Copilot licence, access to the Frontier program, and Copilot Credits.</p><p><strong>B. Apps in Copilot Studio (Public Preview):</strong> These apps are created in Copilot Studio and can be shared with specific people or groups. This option is best for developing department-level applications, especially for organisations that do not have access to the Frontier program. To use this option, you need Copilot Studio access and Copilot Credits.</p><p><strong>C. Power Apps Code App (Vibe Apps Only): </strong>This option allows you to create applications using your own React code, edit them in Visual Studio Code, and deploy them through solutions. It is best suited for business-critical applications that require Application Lifecycle Management (ALM). To use this option, you need Code Apps enabled, Node.js, Power Platform CLI (pa CLI), and Premium users.</p><p>A fourth option is to let the app retire. If nobody opened it last month, export what you need and move on.</p><p><strong>1. If your app was built in App Builder and stores data in SharePoint:</strong> Choose option A (Copilot Cowork) or option B (Apps in Copilot Studio). Both options are suitable for applications that use SharePoint to store data.</p><p><strong>2. If your app uses Dataverse tables in the default environment: </strong>You must move the Dataverse tables to the appropriate environment first, or choose option C (Power Apps Code App).</p><p><strong>3. If your app needs development, testing, and production environments or source control: </strong>Choose option C (Power Apps Code App), as it supports application lifecycle management (ALM) and better control over the development process.</p><p><strong>4. If your app is used by people outside your Copilot rollout: </strong>Choose option B (Apps in Copilot Studio) or option C (Power Apps Code App), as these options are more suitable for sharing apps with users beyond your Copilot rollout.</p><p><strong>5. If your app holds sensitive data: </strong>Choose option B (Apps in Copilot Studio) or option C (Power Apps Code App), with group-based sharing to ensure that only authorised users can access the application and its data.</p><h3><strong>Step 2: Prepare the app (all paths)</strong></h3><ol><li><strong>Inventory makers and apps.</strong> List everyone who used App Builder or vibe.powerapps.com and tell them the deadline.</li><li><strong>Publish draft tables (Vibe).</strong> Open the app, select Publish and pick a data environment. Unpublished draft tables aren’t migrated.</li><li><strong>Collect data details.</strong> For Dataverse, you need the environment ID and each table’s logical name (Tables → open table → Tools → Copy logical name). For SharePoint, you need the site URL and exact list names.</li><li><strong>Check the environment ID.</strong> If it starts with Default-, Paths A and B can’t reach those tables.</li><li><strong>Download the source.</strong> In Vibe, open the app → Edit → Download source. In App Builder, open the app → App Builder settings → Debug → Download app. Keep the .zip as is for Paths A and B.</li><li><strong>Screenshot every screen</strong> so you can compare the rebuilt app. The old app disappears on November 30.</li></ol><h3><strong>Step 3: Write the migration prompt</strong></h3><p>The rebuild agent treats your .zip as reference material, so the prompt must say “copy, don’t redesign.” Avoid “similar to” or “inspired by,” which invite a redesign. Use logical names, not display names, and attach the .zip in the same message.</p><pre>I&#39;m migrating my app, &lt;App name&gt;, which I built. The attached .zip is its <br><br>complete source code. Name the new app &lt;App name&gt;. <br><br> <br><br>It helps &lt;who&gt; to &lt;do what&gt;. Screens: &lt;screen names&gt;. <br><br> <br><br>Data: connect to my existing Dataverse tables in environment &lt;environment ID&gt;: <br><br>&lt;logical names&gt;. Don&#39;t create new tables and don&#39;t use sample data. <br><br>(For SharePoint: use the site URL and list names instead.) <br><br> <br><br>Extract the .zip and read every file before you start. Treat the source as the <br><br>specification: same screens, navigation, labels (word for word), fields, rules, <br><br>filters and create/edit/delete actions. Don&#39;t add, remove or rename anything. <br><br>If you build in stages, list what&#39;s still missing after each stage. </pre><h3><strong>Step 4</strong></h3><h4>A: Rebuild in Copilot Cowork</h4><ol><li>Go to copilot.cloud.microsoft and select <strong>Cowork</strong>.</li><li>Type /app, a space, then paste your prompt.</li><li>Select <strong>+ → Upload images and files</strong>, attach the .zip and send.</li><li>Approve data-access requests. If the agent builds in stages, reply “Continue with the deferred items” until it finishes.</li><li>Select <strong>Publish</strong>, then <strong>Share</strong> to create a link.</li></ol><p><strong>Warning:</strong> Anyone with a Cowork app link can open the app and see all its data. For anything beyond team-level data, use Path B or C.</p><h4>B: Rebuild in Apps in Copilot Studio</h4><ol><li>Sign in to copilotstudio.microsoft.com and turn on <strong>New experience</strong>.</li><li>Select the <strong>App</strong> tile, or <strong>Apps (Preview) → New app</strong>.</li><li>Paste the prompt, attach the .zip and submit.</li><li>Answer follow-up questions and test in the preview pane.</li><li>Open the app’s menu and check that <strong>Data</strong> and <strong>Connections</strong> point to your real tables or lists.</li><li>Select <strong>Publish</strong>, then share with named people or groups.</li></ol><p>If the .zip is over 10 MB, delete the node_modules and dist folders and compress it again. Finish your prep before opening the builder, because credits are billed while you build.</p><h4>C: Convert a Vibe app to a code app</h4><p>This path keeps your actual code, and you get VS Code, Git and solution-based deployment. It needs someone comfortable with a terminal. Before you start, enable code apps in the target environment, install Node.js LTS and the Power Apps CLI (pa), and pick the environment that holds the app’s Dataverse tables.</p><pre># 1. Extract the source and open apps/&lt;your-app&gt; in VS Code. <br><br># 2. In power.config.json, copy connectionReferences / databaseReferences <br><br>#    (and everything after them) somewhere safe, then delete the file. <br><br> <br><br>npm install --save-dev &quot;oxlint-tsgolint@^7.0.2001&quot; --allow-remote=all <br><br>pa app init                 # choose environment + display name <br><br> <br><br># 3. Paste the saved properties into the new power.config.json. <br><br> <br><br>pa app run                  # test with real data at the Local Play URL <br><br>npm run build <br><br>pa app push                 # returns the Power Apps URL of the code app </pre><p><strong>Things to know:</strong></p><ul><li>Images from Microsoft’s Vibe content hosts may not load until you add those hosts to the environment’s Content Security Policy for code apps.</li><li>Users need a Power Apps Premium licence (or pay-as-you-go).</li><li>The code app is separate from the Vibe app, so later changes don’t flow back.</li><li>Add the app to a solution and commit it to your own repository on day one.</li></ul><h3><strong>Step 5: Verify before switching users over</strong></h3><p>Rebuilt apps often look right but still read sample data. Check each app before sending the new link.</p><h4><strong>Data</strong></h4><ul><li>The app shows existing records, not sample data.</li><li>A test record was created, edited and deleted, and the change shows in the table or list.</li><li>Data and Connections (Copilot Studio) or power.config.json (code app) point to the right environment.</li></ul><h4><strong>Behaviour</strong></h4><ul><li>Every screen from your screenshots exists, with the same labels and fields.</li><li>Validation rules, filters, sorting and calculations match the original.</li><li>You asked the agent to compare the new app with the source, and fixed the differences.</li></ul><h4><strong>Access and cost</strong></h4><ul><li>The app is shared with the right people or groups, and the old sharing list is recreated.</li><li>Each user has a Power Apps Premium licence or is covered by a Copilot Credits spending policy.</li></ul><h4><strong>Cut-over</strong></h4><ul><li>Users have the new link, and bookmarks and Teams tabs are updated.</li><li>The original app stays untouched until the new one has run in real use for a few days.</li></ul><h3><strong>What carries over and what doesn’t</strong></h3><p><strong>1. Screens, Logic, and Components: </strong>In Paths A and B, screens, logic, and components are regenerated during the migration process. In Path C, they are retained as code, allowing developers to maintain and modify the original application code.</p><p><strong>2. Dataverse / SharePoint Data:</strong> In both Paths A and B and Path C, existing Dataverse and SharePoint data stays in place. The migration does not remove or move this data.</p><p><strong>3. Unpublished Draft Tables:</strong> In all three paths, unpublished draft tables are lost during the migration process.</p><p><strong>4. Connections: </strong>In Paths A and B, connections are recreated by the agent. In Path C, connections are restored using configuration settings.</p><p><strong>5. App Link and Sharing:</strong> In all paths, the existing app link and sharing settings are not carried over. They need to be configured again after migration.</p><p><strong>6. Default-Environment Dataverse:</strong> In Paths A and B, default-environment Dataverse is not supported. In Path C, support depends on the target environment configuration.</p><p><strong>7. ALM Across Environments:</strong> Paths A and B do not currently support Application Lifecycle Management (ALM) across multiple environments. Path C supports ALM, making it suitable for applications that require development, testing, and production environments.</p><p><strong>8. Exact Look and Fonts:</strong> In Paths A and B, the application uses the closest matching theme, so the original appearance and fonts may change slightly. In Path C, the original look, design, and fonts remain unchanged because the application is retained as code.</p><h3><strong>Cost and limits</strong></h3><p>Apps built in Cowork and Copilot Studio run on Copilot Managed Runtime (preview). Users without Power Apps Premium consume Copilot Credits each time the app calls an API, and during preview, users without credits are blocked after a short trial window. Set up a spending policy in the Microsoft 365 admin center before go-live.</p><p>Both runtimes are in preview, so labels and limits may change. Check Microsoft’s migration page before each batch. For apps with dozens of screens or heavy business rules, expect several refinement rounds or go straight to Path C. If a team now relies on an app daily, consider rebuilding it as a canvas, model-driven or code app with proper governance.</p><h3><strong>How Peafowl IT can help</strong></h3><p>We’re a Microsoft Solutions Partner focused on Power Platform. If you have more apps than time, our <strong>Vibe App Migration Review</strong> gets you through the deadline:</p><ul><li>A tenant-wide inventory of Vibe and App Builder apps, with owners and usage</li><li>A retire / rebuild / convert decision for each app</li><li>Hands-on migration and verification for the apps you keep</li><li>Code app conversion with solutions and Git for business-critical apps</li><li>Licensing and Copilot Credits setup so users aren’t blocked after go-live</li></ul><p><a href="https://peafowlit.com/get-in-touch/"><strong>Book a Vibe App Migration Review</strong></a></p><h3><strong>FAQ</strong></h3><p><strong>Will my data be deleted on November 30?</strong></p><p>No. Data in Dataverse tables or SharePoint lists isn’t stored in the app. You lose access to the app itself, and unpublished draft tables are lost.</p><p><strong>Can I migrate after November 30?</strong></p><p>Not through the retiring tools. Download the source of every app you might want now, because without it a rebuild starts from scratch.</p><p><strong>Do my users need Microsoft 365 Copilot licences?</strong></p><p>Not to run the apps. They need Power Apps Premium or Copilot Credits through a spending policy. Makers using Cowork do need a Microsoft 365 Copilot licence.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d9a931acaa60" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Your Power Pages Portal Leaking Data? A 2026 Security Checklist]]></title>
            <link>https://medium.com/@peafowlit/is-your-power-pages-portal-leaking-data-a-2026-security-checklist-a01ce4be5e42?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/a01ce4be5e42</guid>
            <category><![CDATA[security-audit]]></category>
            <category><![CDATA[microsoft-power-pages]]></category>
            <category><![CDATA[power-pages]]></category>
            <category><![CDATA[dataverse]]></category>
            <category><![CDATA[security]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 05:30:45 GMT</pubDate>
            <atom:updated>2026-11-05T05:31:28.636Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6eh4yIKktQHluGXly3HJEA.png" /><figcaption><strong>Is Your Power Pages Portal Leaking Data? A 2026 Security Checklist</strong></figcaption></figure><p>If any table behind your Power Pages site grants read access to the Anonymous Users web role, anyone on the internet may be able to pull those records with a plain browser request. No malware, no stolen password, no software vulnerability is required. In the 2026 ExfilSquad campaign, that is the leading explanation for roughly 27 million leaked records. This guide explains what happened, how to check your own portal in about 15 minutes, and how to fix and monitor it.</p><h3><strong>What happened: the ExfilSquad campaign in plain terms</strong></h3><p>ExfilSquad is a data-extortion group that surfaced in late July 2026. According to Fortra’s research team (FIRE), it claimed data from 15 organizations, and by August 7 it had published dumps for 13 of them: about 382.64 GB and roughly 27 million records, distributed via torrent. Reported victims span government, education, financial services and manufacturing, and include the City of Atlanta, the UK Department for Education and District of Columbia Public Schools.</p><p>The key finding is <em>how</em> the data was most likely obtained. Researchers found no evidence that malware, lateral movement, credential theft or an exploited software flaw was needed. Instead, the data looks like exports from Microsoft Dataverse (the data layer behind Dynamics 365) taken through public Power Pages portals whose permissions allowed unauthenticated reads. Fortra identified more than 10,000 potentially exposed Power Pages instances and noted that a publicly known tool, Power Pwn, has been used to scan for them.</p><p><strong>Two honest caveats, because accuracy matters more than drama:</strong></p><ul><li><strong>This is the leading theory, not a confirmed fact for every victim.</strong> At least one analyst write-up states plainly that the evidence does not yet confirm every organization was breached through the same configuration issue.</li><li><strong>It is not a flaw in Dynamics 365 or Power Pages itself.</strong> Multiple researchers describe it as a configuration problem: the platform did what it was told.</li></ul><p>Regulators have noticed. FINRA issued a cybersecurity alert to member firms, and the California Department of Financial Protection and Innovation (DFPI) warned registered investment advisers and broker-dealers to audit and remediate their Power Pages and Dataverse configurations, pointing to Microsoft’s Power Pages security documentation. FINRA’s alert also advises that, after remediation, firms confirm unauthenticated requests return an authorization error.</p><h3><strong>Root cause: how “Anonymous Users” access works</strong></h3><p>Power Pages builds external-facing websites on top of Dataverse. Access to that data is governed by several layers working together:</p><ul><li><strong>Web roles</strong> group users. Every visitor who is not signed in belongs to the built-in <strong>Anonymous Users</strong> role; every signed-in user gets <strong>Authenticated Users</strong>.</li><li><strong>Table permissions</strong> define which Dataverse rows a web role can create, read, update or delete, and with what scope (Global, Contact, Account, Parent/child, Self).</li><li><strong>Web API site settings</strong> decide whether a table is reachable at https://&lt;your-portal&gt;/_api/&lt;table&gt; at all, and which columns it returns. Microsoft documents that the Web API is disabled per table by default (Webapi/&lt;table&gt;/enabled defaults to False) and that it follows the table permissions granted through web roles.</li></ul><p>The dangerous combination is simple: <strong>a table with the Web API enabled + a table permission with Read + the Anonymous Users role attached + a broad scope such as Global.</strong> Together they turn a marketing-friendly portal into a bulk export endpoint. Microsoft’s own guidance says to treat anonymous table permissions as public data access, and to avoid global table permissions on sensitive business data because the risk is amplified when the table is exposed through the Web API.</p><p>Why does this keep happening? Because nothing breaks. A developer grants anonymous read to make a public “services” or “locations” page work, a column list is left wide, the table later gains sensitive fields, and no alert fires. There is no patch to apply and no CVE to track. It is configuration drift, and it is invisible until someone scans for it.</p><h3><strong>Remediation checklist</strong></h3><p>Work top to bottom; the first items remove the most risk.</p><ol><li><strong>Remove Anonymous Users from any table permission that doesn’t strictly need public data.</strong> If the business case is a public lookup (store locations, public notices), keep it, but scope it to the minimum and expose only non-sensitive columns.</li><li><strong>Replace Global scope with relationship-based scopes</strong> (Contact, Account, Self, Parent/child) so the server enforces which rows each user may see.</li><li><strong>Disable the Web API for tables that don’t need it</strong> (Webapi/&lt;table&gt;/enabled = false), and for the rest, list explicit columns instead of *.</li><li><strong>Apply column-level protection.</strong> Use Power Pages column permissions for the Web API authorization model, or Dataverse column security profiles if enhanced authorization is enabled. Microsoft warns that Web API column permissions do not protect the same data when it is exposed through search, Liquid, FetchXML, forms or lists, so review those paths separately.</li><li><strong>Minimize sensitive data in portal-facing tables.</strong> Field masking and data minimization are only as good as what the portal can reach; if a column isn’t needed on the site, don’t expose it.</li><li><strong>Lock down page and file access.</strong> Hiding a page from navigation doesn’t secure it. Use page permissions, and test direct file-download URLs as an anonymous user.</li><li><strong>Add layered controls:</strong> disable open registration where access should be controlled, enable CAPTCHA on anonymous forms, and put a Web Application Firewall with rate limiting in front of production sites.</li><li><strong>Preserve evidence before you change anything.</strong> If you suspect exposure, keep Dataverse audit logs and web server logs, then review who requested /_api/ endpoints. Involve legal and your regulator-notification process early.</li></ol><h3><strong>Test from an unauthenticated session</strong></h3><p>After each change, re-verify rather than trusting the admin console’s own summary:</p><ul><li>Open a private/incognito window with no cookies and browse to /_api/&lt;table&gt;?$top=5 for every exposed table. You want an authorization error, not JSON.</li><li>Run the script above again and archive the output as evidence.</li><li>Test as an <strong>authenticated but low-privilege</strong> user too. Some leaks happen when any signed-in account, including a self-registered one, can read everyone else’s rows.</li><li>Check public forms still submit. A good fix blocks unauthenticated reads while leaving legitimate public create flows working.</li></ul><h3><strong>Ongoing monitoring, not a one-time fix</strong></h3><p>Microsoft’s guidance is explicit that security reviews aren’t limited to launch. Build these into routine operations:</p><ul><li><strong>Power Pages Security Scan</strong> on a schedule and after significant changes, with findings tracked to closure.</li><li><strong>Security Hub</strong> in the Power Platform admin center to see which sites have been scanned, which haven’t, and what’s open.</li><li><strong>Dataverse auditing</strong> on sensitive tables and columns, reviewed regularly.</li><li><strong>Change control:</strong> deploy Power Pages configuration through solutions and reviewed ALM processes, and restrict who can edit web roles, table permissions and site settings in production.</li><li><strong>Re-review after change:</strong> new table, new column, new integration or new site means the permissions need another look.</li><li><strong>Retire the unused:</strong> delete obsolete sites, permissions and APIs rather than leaving them dormant.</li></ul><h3><strong>How Peafowl IT can help</strong></h3><p>We build and support <a href="https://peafowlit.com/power-pages/">Power Pages</a> and Dataverse solutions every week, so we know where these settings hide in real projects. Our <strong>Power Pages Security Audit</strong> is a fixed-scope review that gives you:</p><ul><li>An inventory of every Power Pages site in your tenant, including forgotten pilots</li><li>A full report of anonymous and Global-scope permissions, Web API exposure and other read paths</li><li>Signed-out test evidence for each site, before and after remediation</li><li>A prioritized fix plan, or hands-on remediation by our team</li><li>A governance setup so the problem doesn’t come back</li></ul><p><a href="https://peafowlit.com/get-in-touch/"><strong>Book a Power Pages Security Audit</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a01ce4be5e42" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Create, Edit, and Publish HTML Pages in Microsoft SharePoint with Copilot]]></title>
            <link>https://medium.com/@peafowlit/how-to-create-edit-and-publish-html-pages-in-microsoft-sharepoint-with-copilot-6f317a9d5e8e?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/6f317a9d5e8e</guid>
            <category><![CDATA[microsoft-copilot]]></category>
            <category><![CDATA[co-pilot]]></category>
            <category><![CDATA[sharepoint-development]]></category>
            <category><![CDATA[sharepoint]]></category>
            <category><![CDATA[sharepoint-copilot]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 05:00:08 GMT</pubDate>
            <atom:updated>2026-10-29T05:00:08.260Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l5_aJcrn8dIBPQx6m_yGkw.jpeg" /><figcaption><strong>How to Create, Edit, and Publish HTML Pages in Microsoft SharePoint with Copilot</strong></figcaption></figure><p>Every business has a backlog of web content: campaign microsites, KPI dashboards, onboarding hubs, event landing pages. Most of it waits for a developer. SharePoint’s new HTML pages capability, powered by Copilot, lets business teams build that content themselves, inside the platform they already use.</p><p>This guide covers how it works, how to get started, and what to put in place before you roll it out.</p><h3><strong>What Are SharePoint HTML Pages?</strong></h3><p>In SharePoint, you can create an HTML page with Copilot or upload an existing HTML file. The page is stored in the Site Pages library, where you can edit, publish and share it alongside your existing ASPX pages. HTML pages are an additional content format and don’t replace your existing ASPX pages.</p><p>For marketing, comms and operations teams, that means richer layouts and interactive content without leaving SharePoint or standing up separate hosting.</p><h3><strong>Why It Matters for Your Business</strong></h3><ul><li><strong>Speed to publish:</strong> Describe a page in plain English and let Copilot produce a first draft. Developer time goes to review rather than starting from scratch.</li><li><strong>Reuse existing assets:</strong> You can upload an existing .html file from the HTML page creation screen, or upload HTML pages or folders from the Site Pages library.</li><li><strong>Governance stays intact:</strong> Publishing an HTML page does not grant new permissions. Existing SharePoint access rules continue to apply.</li><li><strong>Connected data potential:</strong> Community walkthroughs show HTML pages connecting to live SharePoint and Excel data, which makes dashboards and reporting pages realistic.</li></ul><h3><strong>How to Create an HTML Page with Copilot</strong></h3><p>You can start from several places. On a SharePoint site, select + New and then HTML page, or use the floating Copilot button. From the Publish landing page, select + Create, then HTML page.</p><ol><li><strong>Open the creation screen.</strong> Use any of the entry points above.</li><li><strong>Write your prompt.</strong> Describe the page you want and attach files for extra context. Be specific about the audience, purpose, sections, tone and any brand colours (or colors, if your audience is in the US).</li><li><strong>Review the draft.</strong> Copilot creates a draft HTML page and opens it in edit mode.</li><li><strong>Refine and publish</strong> (covered below).</li></ol><p><strong>Prompt tip:</strong> Instead of “build a landing page”, try: “Create a landing page for our Q1 customer onboarding programme with a hero banner, three benefit cards, a timeline of the first 30 days and a contact section. Professional, friendly tone.” Specific prompts give you far better first drafts.</p><h3><strong>How to Edit Your HTML Page</strong></h3><p>You have two routes. You can edit text directly on the page, or use Copilot for broader content and design changes.</p><p>You also don’t need a Copilot licence (license in US spelling) for every task. Editing page content directly needs no Copilot licence, but creating or editing a page through the Copilot chat pane requires a Microsoft 365 Copilot licence. So a small group of Copilot-licensed authors can generate pages, and a wider group can maintain them.</p><h3><strong>How to Publish and Share</strong></h3><p>Once you’re happy, save and publish. HTML pages live in the Site Pages library alongside your modern pages, so you edit, publish and share them using the same flow you already know. Viewers need only permission to view the page.</p><h3><strong>Before You Roll It Out: A Quick Governance Checklist</strong></h3><ol><li><strong>Check the file type setting.</strong> In Site Pages, open File types and check that Allow HTML files is on.</li><li><strong>Define who can author.</strong> Your ability to create, edit and publish still depends on the permissions assigned to you.</li><li><strong>Set content standards.</strong> Agree on brand guidelines, accessibility requirements and a review step before anything goes live.</li><li><strong>Train your authors.</strong> Microsoft’s own rollout guidance suggests informing SharePoint authors about the new creation, upload, edit, publish and sharing capabilities before rollout.</li><li><strong>Review custom code.</strong> If teams upload HTML from outside sources, add a security review for scripts and external references.</li></ol><h3><strong>Where Power Platform Fits In</strong></h3><p>HTML pages are a strong front end, but the real efficiency gains come from what sits behind them. Pair a Copilot-built page with Power Automate approval flows, Power Apps for data capture and Power BI for reporting, and you have a joined-up solution rather than a standalone page. Examples include an HR onboarding hub that triggers IT provisioning, or a sales enablement page that feeds live pipeline data.</p><h3><strong>Where Peafowl IT Fits</strong></h3><p><a href="https://peafowlit.com/">Peafowl IT</a> works across the Microsoft 365 and SharePoint ecosystem, from initial Copilot readiness assessments and information architecture reviews to hands-on deployment and governance setup.</p><p>If you want to use Copilot in SharePoint effectively across your organisation, the difference between a useful deployment and a frustrating one usually comes down to two things: the quality of your underlying <a href="https://peafowlit.com/sharepoint-development/">SharePoint information architecture</a>, and the governance model you put in place before the first dashboard goes live.</p><p>Our SharePoint Copilot Readiness Assessment covers both. It starts with an honest audit of where your SharePoint environment stands today, not a demo of what it could look like in a best-case scenario.</p><h3><strong>Final Thoughts</strong></h3><p>SharePoint HTML pages with Copilot shorten the path from idea to published content, while keeping your existing permissions and governance in place. Start with one low-risk use case, such as an internal campaign page or team hub, measure the time saved, and scale from there.</p><h4>Ready to get more from SharePoint and Microsoft 365?</h4><p>Peafowl IT helps US and UK organisations assess, deploy, govern and automate their Microsoft 365, SharePoint and Power Platform environments. Book a free consultation to see what’s possible.</p><p><strong><em>Note:</em></strong><em> Microsoft has been rolling this feature out in stages, so availability and admin controls may vary by tenant. Check your Message Center (MC1479517) and Microsoft’s support documentation for your organisation’s status before publishing.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6f317a9d5e8e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Power Apps’ “Vibe” Experience Is Retiring: What It Means and What to Do Before November 30]]></title>
            <link>https://medium.com/@peafowlit/power-apps-vibe-experience-is-retiring-what-it-means-and-what-to-do-before-november-30-ac8dfd26025d?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/ac8dfd26025d</guid>
            <category><![CDATA[codapps]]></category>
            <category><![CDATA[microsoft-power-apps]]></category>
            <category><![CDATA[powerapps]]></category>
            <category><![CDATA[microsoft-power-platform]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 05:22:28 GMT</pubDate>
            <atom:updated>2026-10-22T05:22:28.678Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fYMcH9HjhkimU6Z8SiZNPg.jpeg" /><figcaption>Power Apps’ “Vibe” Experience Is Retiring: What It Means and What to Do Before November 30</figcaption></figure><p>Microsoft is not abandoning AI app building. It is moving it. Here is the plan, the risks, and a checklist for IT managers and makers.</p><p>If your team is currently prototyping applications via `vibe.powerapps.com` or by using the ‘App Builder’ agent within Microsoft 365 Copilot, you should take note of a specific deadline. Through ‘Message Center’ notice <strong>MC1505609</strong>, Microsoft has announced that both of these features will be discontinued on <strong>November 30, 2026.</strong></p><p>The headline “Power Apps is retiring <a href="https://peafowlit.com/blog/the-impact-of-vibe-coding-on-the-tech-industry-what-you-need-to-know/">AI vibe coding</a>” is a little misleading, and the difference matters for how you plan. Let us separate what is actually ending from what is simply moving.</p><h3><strong>What is actually being retired</strong></h3><p>Per Microsoft’s notice, two things are going away:</p><ul><li><strong>The App Builder agent</strong> in Microsoft 365 Copilot</li><li><strong>vibe.powerapps.com</strong>, the standalone “vibe” experience for Power Apps</li></ul><p>Microsoft’s stated reason is consolidation: <strong>Copilot Cowork</strong> and <strong>Microsoft Copilot Studio</strong> become the primary places to build apps with natural language, giving makers one consistent path instead of several overlapping ones.</p><blockquote><strong>Key takeaway: </strong>Natural-language app building is not being removed from the Microsoft stack. The <em>entry points</em> are changing. The risk is for apps and habits built on the old entry points.</blockquote><h3><strong>The timeline</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/555/1*fw3l9umQZuiXevUneWIblA.png" /></figure><blockquote><strong>The part that bites: </strong>Microsoft states that apps not successfully migrated before November 30 will no longer be accessible through the retired experiences. That is roughly six weeks from today.</blockquote><h3><strong>Who should care</strong></h3><h4>Business users and makers</h4><p>If you built an onboarding tracker, inventory list, or project tracker through vibe, open it now and decide whether it is worth keeping. Those are the typical use cases Microsoft itself pointed to when it introduced the experience.</p><h4>IT managers and admins</h4><p>You likely do not know exactly who is using vibe. Environments, makers, and “just trying it out” apps can sprawl quietly, especially because the experience required a non-default environment and a tenant-level Copilot setting. Your job is to find them before the deadline does.</p><h4>Tech teams and architects</h4><p>The vibe experience generated real code (React-based) behind apps. Whatever you migrate, treat it as code you now own: review it, test it, and decide where it lives.</p><h3><strong>Why this is not a surprise</strong></h3><p>The vibe experience has always carried a preview label. Microsoft’s own Learn documentation states that preview features are not meant for production use and may have restricted functionality, and the same page now carries a notice recommending <a href="https://peafowlit.com/blog/microsoft-copilot-cowork-ga-governance-guide/">Copilot Cowork</a> and, soon, Copilot Studio for app building going forward.</p><p>Early community feedback said the same thing: great for learning and prototyping, not a replacement for Canvas, Model-driven, or your existing architecture. If any vibe app has quietly become something a team depends on, that is the real governance finding here, regardless of the retirement.</p><h3><strong>The cost change nobody is talking about</strong></h3><p>This is more than a change of address. Microsoft’s companion notice <strong>MC1469329</strong> (Build apps in Microsoft Copilot Studio and Copilot Cowork) tells admins to review the <strong>Copilot Credits</strong> requirements for building <em>and</em> running apps and to configure usage-based billing as needed. Microsoft’s Copilot Studio licensing guidance adds that building consumes credits based on app complexity and how much generation and iteration happens during authoring, and that running apps built in Copilot Studio follows usage-based billing by default.</p><blockquote><strong>For IT managers: </strong>The preview vibe experience did not behave like a metered workload. Moving makers to Copilot Cowork or Copilot Studio means experimentation now has a cost line. Set budgets, owners, and billing policies <em>before</em> you redirect makers, not after the first invoice.</blockquote><p>Two more points from Microsoft’s notice are worth sharing with admins: apps built in Cowork can be opened and edited further in Copilot Studio, and published apps appear under <strong>Apps &gt; All apps</strong> in the Microsoft 365 admin center, which gives you a central place to see what exists.</p><h3><strong>Your 6-week migration checklist</strong></h3><ol><li><strong>Inventory.</strong> Identify makers and teams using vibe.powerapps.com or the App Builder agent. Check non-default environments first, since vibe was not available in the default one.</li><li><strong>Triage each app.</strong> Sort into <em>retire</em> (experiments), <em>migrate</em> (genuinely useful), and <em>rebuild properly</em> (business-critical, deserving a governed Canvas or Model-driven build).</li><li><strong>Communicate and redirect.</strong> Tell affected makers the dates above and move all ongoing and new natural-language app work to Copilot Cowork (Frontier program, needs a Microsoft 365 Copilot license) or Copilot Studio (public preview rollout).</li><li><strong>Prepare billing and access.</strong> Complete Frontier onboarding, enable eligible app creators in the Microsoft 365 admin center (Apps &gt; Overview), and configure Copilot Credits usage-based billing.</li><li><strong>Use the in-product migration.</strong> Microsoft says existing apps can be opened to start migration once it becomes available around mid-October. Confirm in your tenant whether it has reached you yet, since rollouts are staged.</li><li><strong>Export what matters.</strong> Keep the data in Dataverse safe and document the logic of anything you might rebuild by hand.</li><li><strong>Finish before Nov 30.</strong> Leave buffer. A deadline on the last day of the month is a bad day to discover a migration failure.</li></ol><h3><strong>When migration is the wrong answer</strong></h3><p>Not every vibe app deserves a straight port. If an app handles sensitive data, needs formal ALM (dev, test, prod pipelines), or serves more than a small team, a one-click migration of an AI-generated prototype may just move technical debt to a new home. In those cases, treat the prototype as a requirements document and rebuild on a governed architecture.</p><p>Also keep in mind that Apps in Copilot Cowork sit in the Frontier program and Apps in Copilot Studio are in public preview. Preview-to-preview migration is fine for experiments, but it does not give you a production-grade footing by itself.</p><h3><strong>What we do not know yet</strong></h3><ul><li>How complete the in-product migration will be for apps with custom logic or generated code</li><li>Whether your tenant and region are covered yet, since the vibe experience was limited to specific regions and English</li><li>The final licensing, pricing, and GA timeline for the replacement experiences, including real-world Copilot Credits consumption per app</li></ul><p>We will update this post as Microsoft publishes more detail. Check the Message Center and the linked docs below for the latest.</p><p><strong>Not sure how many vibe apps are in your tenant?</strong></p><p>Our <strong>Power Platform Architecture Review</strong> inventories your environments, flags at-risk apps, and gives you a migrate / retire / rebuild plan before November 30.</p><p><a href="https://peafowlit.com/"><strong>Book a Power Platform Architecture Review</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac8dfd26025d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Power Platform Agent Governance: Where 2026 Wave 1 Actually Stands in October]]></title>
            <link>https://medium.com/@peafowlit/power-platform-agent-governance-where-2026-wave-1-actually-stands-in-october-3921b1cc06e4?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/3921b1cc06e4</guid>
            <category><![CDATA[microsoft-365]]></category>
            <category><![CDATA[power-platform-consulting]]></category>
            <category><![CDATA[powerplatform]]></category>
            <category><![CDATA[power-platform-developers]]></category>
            <category><![CDATA[microsoft-power-platform]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 06:30:29 GMT</pubDate>
            <atom:updated>2026-10-15T06:30:29.239Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gd2Thjb_1Rw92IrzGDOh6g.jpeg" /><figcaption>Power Platform Agent Governance: Where 2026 Wave 1 Actually Stands in October</figcaption></figure><p>Microsoft’s “Q3 2026” governance targets have come and gone, and their announced release plans have been retired. Here’s what’s shipping, what’s still in preview, and how to rank it without breaking your make’s agents.</p><p>If you’ve been tracking Power Platform Agent Governance from the 2026 Wave 1 plan, you’ve probably noticed two things. The September target dates have passed, and the plan itself will not be updated. Microsoft announced on August 25, 2026 that the Dynamics 365, Power Platform, and Dataverse roadmap content is moving to the AI ​​at Work roadmap, and the release planner is retiring on November 15, 2026.</p><p>That makes it a good time to stop reading the headlines and examine what each governance feature actually does and what state it’s in. Three things are most important to anyone running Copilot Studio agents at scale: enhanced admin controls for agent security, the Agentic Center of Enablement (Agentic CoE), and GitHub support for solution source control. They are not at the same maturity level, and treating them as a “Wave 1 governance bundle” surprises teams.</p><h3><strong>The status check: three features, three different answers</strong></h3><p><strong>Manage agent security with enhanced admin controls:</strong> This feature was in Preview from 31 May 2026, with General Availability (GA) listed for September 2026. A Message Center post also cited 2 September 2026. However, the authentication documentation reviewed still carried a Preview label. Check your tenant’s status in the admin center before confirming GA availability or making commitments to the business.</p><p><strong>Automate governance with Agentic CoE:</strong> This feature entered Preview in September 2026, with no GA date announced. You can pilot it to evaluate its governance capabilities, but avoid making it a core part of your compliance strategy until its availability and capabilities are confirmed.</p><p><strong>GitHub as a Git provider for solution source control:</strong> This feature entered Preview on 31 August 2026, according to Message Center notification MC1455050. Azure DevOps integration is already generally available (GA). GitHub integration is worth exploring in development environments and for fusion teams, but continue using established deployment pipelines for production workloads.</p><blockquote><strong>Worth fixing internally</strong></blockquote><blockquote>If a slide in your organization says all three are “GA in Q3 2026”, it’s wrong on at least one count. The Agentic CoE page lists a September 2026 preview and there is no general availability date at all.</blockquote><h3><strong>Where to track this now: AI at Work Roadmap</strong></h3><p>Existing roadmap items with preview or GA dates on or after June 1, 2026 are moving from Release Planner to the AI ​​at Work Roadmap this fall. Views saved in My Release Plans will not migrate. If your change-management process relies on those views, recreate them before November 15. Microsoft says features are still shipped on their own timeline; they’re just changing where you find them.</p><p>Actionable step: Assign a named person to review the monthly AI at Work roadmap for Power Platform, Dataverse, and Copilot Studio items, and log the change date. This is important because the dates for agent security controls have moved earlier. Previous iterations of this same admin-center capability were announced with 2025 GA dates and then postponed. Dates are a purpose, not a commitment.</p><h3><strong>Why Agents Break the Old Governance Model</strong></h3><p>Classic power platform governance assumes apps and flows: who can use which connectors, which environments touch production data. Agents add autonomy, multi-system reach, and instant delivery to a single conversation after being published to a channel like Teams. We covered this in depth in our <a href="https://peafowlit.com/blog/power-platform-copilot-agent-governance-2026/">Wave 1 Agent Governance Primer.</a> This post is a follow-up: not “what’s in Wave 1,” but “what you can trust today and in what order.”</p><h4><strong>1. Agent security controls: authentication as policy</strong></h4><p>The core of this feature is an admin-center setting called <strong>Authentication for agents</strong>. You can find it under Security, then Identity &amp; Access. You select an environment or environment group and choose how user-created agents in Copilot Studio can authenticate:</p><ul><li><strong>No authentication (publicly available):</strong> Makers can configure any option, including anonymous access.</li><li><strong>Require Microsoft authentication:</strong> The strictest option. Only “Authenticate with Microsoft” is available to makers.</li><li><strong>Require Entra authentication:</strong> Manual authentication with Microsoft authentication or Microsoft Entra ID; generic OAuth 2 and anonymous are blocked.</li><li><strong>Allow all supported authentication methods:</strong> Authentication is required, but both Entra and generic OAuth 2 manual setup are allowed. Anonymous is still blocked.</li></ul><p>The release plan adds that administrators can also set policies governing agent sharing and restrict unsupported configurations evaluated at deployment and runtime so that only compliant agents can be published and accessed. For a security reviewer, the point is simple: weak authentication paths like anonymous endpoints stop being a per-builder decision and become a tenant policy.</p><p><strong>A sequence that avoids outages</strong></p><ol><li>First inventory. List agents per environment and note which ones use anonymous or generic OAuth 2 access.</li><li>Contact owners of anything that will be breached and agree on a migration date.</li><li>Implement on new environments and development groups first using environment groups so that the rule is consistently applied as environments are added.</li><li>After owners migrate, tighten production last.</li></ol><h4><strong>2. Agentic CoE: Governance that reads your tenant for you</strong></h4><p>Agentic CoE adds three agents to the Power Platform Admin Center:</p><ul><li><strong>Highlights</strong> posts a daily snapshot of significant tenant activity, such as resource creation and capacity usage, with no configuration.</li><li><strong>Insights</strong> continuously scans the tenant and prioritizes security, compliance, performance, and adoption findings. Microsoft examples include ownerless resources and activity in the default environment.</li><li><strong>Action Plans</strong> turns findings into remediation plans that you can edit and must be approved before anything is implemented. Each agent action is logged on the Activity page.</li></ul><p>It is automatically enabled for Power Platform administrators who have at least one managed environment in the tenant. That detail is important: If you have never adopted a managed environment, this is another reason to start.</p><h4><strong>3. GitHub for Solution Source Control</strong></h4><p>Dataverse Git integration lets teams sync solutions and their objects with a Git provider, and Microsoft’s documentation now names both Azure DevOps and GitHub. An August 2026 Message Center post described GitHub support as an extension of the original integration that also brings better performance, reliability, and new public APIs to both providers.</p><p>Why is this a governance topic and not just a developer feature? Because source control gives you three things agents ask about: who changed what and when, a path to roll back, and a recursive path to production. Microsoft’s own guidance is specific about how to use it:</p><ul><li>Use Git integration with a <strong>developer environment</strong>, not test or production.</li><li>Source control holds unmanaged solutions from the producer’s environment; <strong>managed solutions are built from source</strong> and deployed downstream.</li><li>Deploy with <strong>pipelines in Power Platform</strong> using builds (Azure Pipelines or GitHub Workflow) to create artifacts.</li></ul><p>If your engineering team already lives in GitHub, this removes the long-standing excuse of keeping low-code assets outside of the common change process. Remember that the GitHub provider is in preview, so confirm with your own testing before standardizing.</p><p><strong>When not to rush</strong></p><ul><li>Don’t put a preview feature in a compliance control. Microsoft’s authentication documentation states that preview features are not for production use and may have limited functionality. If a control survives an audit, wait for a confirmed GA label.</li><li>Don’t implement first and then inventory. See the outage scenario above.</li><li>Don’t let the Agentic CoE approve its own plans. The approval step is only a safeguard if a human actually reads the plan.</li><li>Don’t ignore the basics. Environmental strategy, DLP policies limited to agent knowledge sources, and ownership of each environment still determine whether any of this works.</li></ul><h3><strong>A 30-day plan for IT managers</strong></h3><p><strong>Week 1: </strong>The Power Platform admin should confirm the status of each feature in their own tenant and rebuild roadmap tracking based on the AI at Work roadmap.</p><p><strong>Week 2: </strong>The admin and security teams should inventory existing agents and their authentication methods, as well as review managed environment coverage.</p><p><strong>Week 3:</strong> The admin should pilot Agentic CoE in a managed environment and apply the authentication policy to a development environment group.</p><p><strong>Week 4:</strong> The IT manager and development lead should pilot GitHub-backed source control in one developer environment, present the findings, and share a proposed production rollout sequence with business owners.</p><p>What business leaders should ask IT</p><ul><li>Which of our agents can be reached without signing in?</li><li>Who owns each agent, and who is told when it changes?</li><li>If an agent misbehaves, how fast can we roll it back?</li></ul><h3><strong>Not sure which of your agents would break?</strong></h3><p>Our <strong>Agent Readiness Assessment</strong> inventories your Copilot Studio agents, maps authentication and ownership gaps, and gives you a sequenced governance plan before you switch on enforcement.</p><p><a href="https://peafowlit.com/get-in-touch/"><strong>Book an Agent Readiness Assessment</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3921b1cc06e4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Copilot Studio Credits Explained: What Your Agents Will Actually Cost in 2026]]></title>
            <link>https://medium.com/@peafowlit/copilot-studio-credits-explained-what-your-agents-will-actually-cost-in-2026-44a5b1e5b265?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/44a5b1e5b265</guid>
            <category><![CDATA[microsoft-copilot-studio]]></category>
            <category><![CDATA[copilot-studio-credits]]></category>
            <category><![CDATA[copilot-studio]]></category>
            <category><![CDATA[calculator]]></category>
            <category><![CDATA[co-pilot]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 08 Oct 2026 04:01:02 GMT</pubDate>
            <atom:updated>2026-10-08T04:01:02.054Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IDjGHQow1GO9kbFQO6RgNw.jpeg" /><figcaption>Copilot Studio Credits Explained: What Your Agents Will Actually Cost in 2026</figcaption></figure><p>A plain-English cost model, a worked 500-employee example, and a calculator you can run on your own numbers before the next invoice does it for you.</p><h3>What actually changed, and what didn’t</h3><p>If you’ve read that Copilot Studio “switched to credits” this summer, here is the precise version, because it changes how you should react.</p><p>The unit changed from <em>messages</em> to <em>Copilot Credits</em> on <strong>September 1, 2025</strong>. Microsoft stated there was no change to the quantity per prepaid pack or to the pay-as-you-go rate. That was a renaming of the meter, not a price move.</p><blockquote>The 2026 change that matters</blockquote><blockquote>Microsoft Message Center item <strong>MC1461678</strong> ended the billing grace period for agents and workflows built on the <strong>GitHub Copilot harness</strong> that were created before <strong>August 3, 2026</strong>. From <strong>September 1, 2026</strong>, both maker activity (authoring, previewing, testing, building evaluations) and runtime execution consume credits. Standard-harness and Copilot-chat-harness agents were not affected.</blockquote><p>So the cost shock lands in a specific place: teams that experimented with reasoning-heavy, multi-step agents while they were effectively free. If that’s you, your first job is to find those agents. Everyone else still needs a model, because the standard-harness meter is easy to under-estimate.</p><h3>How a credit gets consumed</h3><p>A Copilot Credit is a measure of the time and effort an agent spends retrieving information, responding and acting. Microsoft publishes a rate table for standard-harness agents. These are the lines that drive most real bills:</p><p>The pricing is simple:</p><ul><li><strong>Classic answer: </strong>1 credit = $0.01</li><li><strong>Generative answer: </strong>2 credits = $0.02</li><li><strong>Agent action:</strong> 5 credits = $0.05</li><li><strong>Tenant graph grounding:</strong> 10 credits per message = $0.10</li><li><strong>Agent flow actions:</strong> 13 credits per 100 actions = $0.13</li><li><strong>Text/Generative AI tools:</strong> Basic = $0.01, Standard = $0.15, Premium = $1.00 per 10 responses</li><li><strong>Content processing:</strong> 8 credits per page = $0.08 per page</li></ul><p><strong>In short:</strong> 1 credit costs $0.01 under PAYG.</p><p>The important sentence in Microsoft’s docs is that one interaction can use several features at once. Their own example: a tenant-graph-grounded agent can use <strong>12 credits</strong> (10 for grounding plus 2 for the generative answer) to answer a single prompt. Grounding is five times the cost of the answer it supports.</p><blockquote>Reasoning models add a second meter</blockquote><blockquote>When an agent uses a reasoning-capable model, you pay the normal feature rate <em>plus</em> the premium text and generative AI tools rate of 10 credits per 1,000 tokens. Deep reasoning on a simple lookup is the most avoidable line on the bill.</blockquote><h3>The M365 Copilot exception, and its fine print</h3><p>Employee-facing use of Copilot Studio agents is included, at no credit charge, when the user holds a Microsoft 365 Copilot licence (Microsoft now brands it Microsoft Copilot) and the agent runs under that user’s authenticated identity in Copilot Chat, Teams or SharePoint. Three caveats matter for modelling: usage is subject to fair-use limits, computer-using agents are excluded, and agent flows are zero-rated only when triggered by <em>“When an agent calls the flow”</em>. Unlicensed colleagues using the same agent still consume credits.</p><h3>Pay-as-you-go vs. packs vs. pre-purchase: the break-even math</h3><ul><li>Pay-as-you-go (Azure meter): Costs $0.01 per credit. You are billed after usage and need an Azure subscription with a billing policy.</li><li>Capacity Pack: Costs $200 for 25,000 credits per month, which works out to $0.008 per credit. It is billed annually, unused credits expire each month, and overage is charged at 125%.</li><li>Copilot Credit Pre-Purchase Plan (P3): Starts at $2,850 for 300,000 credits, with an effective rate of $0.0095 per credit, which can decrease to $0.0080 with higher tiers. It is a one-year credit pool, with discounts ranging from 5% to 20% depending on the tier.</li></ul><p><strong>Packs vs. PAYG.</strong> A pack costs $200 and PAYG costs $0.01 a credit, so a pack wins only if you will actually burn more than <strong>20,000 of its 25,000 credits</strong> in the month. Below 80% utilisation you are paying for credits that vanish at month end. Because pack credits are consumed before the PAYG meter, the cheapest design is usually <em>N packs sized to your reliable floor, plus PAYG as overflow</em>, not enough packs to cover your peak.</p><p><strong>Pre-purchase is not automatically cheaper than packs.</strong> Check the tier table: Tier 1 works out to $0.0095 a credit, which is more than a pack’s $0.0080. Only Tier 9 (300 million credits) reaches a 20% discount. Microsoft’s own conversion shows Tier 1 equal to one pack for 12 months, which would cost $2,400 against $2,850. P3 earns its price through flexibility: credits don’t expire monthly, so spiky or seasonal workloads can use the pool when they need it, and it works across eligible Microsoft products. If your consumption is smooth, packs win. If it is lumpy, model both.</p><blockquote>Enforcement is a business risk, not just a cost</blockquote><blockquote>On prepaid capacity, custom agents are disabled once the tenant reaches 125% of purchased credits. Agent flows are stricter: new runs are blocked when capacity is exhausted. Keeping a PAYG meter linked to the environment removes that cliff.</blockquote><h3>A worked example: a 500-employee internal support agent</h3><p>The brief for this post imagined a “500-seat customer support agent”. Seats matter most for <em>employee-facing</em> agents because of the licence exception above, so we model an internal IT/HR help agent. These are illustrative assumptions, not client data; the calculator below lets you replace each one.</p><p>· 500 employees, 70% active in a month, 6 conversations each, giving <strong>2,100 conversations</strong></p><p>· Per conversation: 1 classic answer, 3 generative answers (2 of them tenant-graph grounded), 1 agent action</p><p>· 10,500 agent flow actions a month (ticket creation, lookups)</p><p><strong>Per conversation:</strong> (1×1) + (3×2) + (2×10) + (1×5) = <strong>32 credits</strong>. Monthly: 2,100 × 32 = 67,200, plus flows (10,500 ÷ 100 × 13 = 1,365) = <strong>68,565 credits</strong>.</p><ul><li><strong>A. As designed: </strong>68,565 billable credits would cost $685.65 with PAYG. The best option is 2 capacity packs + PAYG for the extra usage, bringing the monthly cost down to $585.65.</li><li><strong>B. 40% of staff hold M365 Copilot:</strong> 41,139 billable credits would cost $411.39 with PAYG. The best option is 1 capacity pack + PAYG for the extra usage, costing $361.39 per month.</li><li><strong>C. Tenant graph grounding turned off:</strong> 26,565 billable credits would cost $265.65 with PAYG. The best option is 1 capacity pack + PAYG for the extra usage, costing $215.65 per month.</li><li><strong>D. B and C together:</strong> 15,939 billable credits would cost $159.39 with PAYG. In this case, PAYG only is the best option, with a monthly cost of $159.39.</li></ul><p><strong>Overall:</strong> The biggest savings come from reducing M365 Copilot usage and turning off tenant graph grounding.</p><p>Two lessons. First, the buying decision is worth about $100 a month in scenario A, but the <em>design</em> decision (grounding) is worth about $370. Optimise the agent before you optimise the contract. Second, scenario D lands below the break-even for even one pack, so a team that bought “a pack to be safe” would be paying for credits it never uses. Note also that scenario C trades answer quality for cost; grounding exists because it improves relevance, so test before you switch it off.</p><p>For scale, Microsoft’s own customer-support example (four classic and two generative answers per session, 900 customers a day) comes to 7,200 credits a day, over 200,000 a month. External-facing agents, where nobody holds a licence, are where credit bills get large.</p><h3>Hidden cost drivers</h3><ul><li><strong>Grounding on by default thinking.</strong> Tenant graph grounding is optional per agent. SharePoint- and Teams-backed knowledge feels free because the content already exists, but graph grounding bills 10 credits per message.</li><li><strong>Agent flows.</strong> 13 credits per 100 actions looks tiny until a flow loops over a list. Power Automate cloud flows are licensed separately and do not consume credits, so deterministic work can often move there.</li><li><strong>The GitHub Copilot harness bills from the first build.</strong> Unlike the standard harness, which bills after publish, it consumes credits when you build, preview, test and generate evaluations. Credits cover LLM tokens, tools (including knowledge and MCP) and the harness itself.</li><li><strong>Environment sprawl.</strong> Capacity is a tenant pool. Unallocated environments draw from it, and an untested sandbox can quietly spend what production needs.</li><li><strong>Month-end expiry.</strong> Pack credits do not roll over. An over-bought pack is pure waste.</li></ul><h3>Five ways to cut consumption without cutting capability</h3><ol><li><strong>Scope grounding deliberately.</strong> Turn tenant graph grounding off for agents whose answers live in a known SharePoint site or a small document set. Keep it for agents that genuinely need tenant-wide retrieval.</li><li><strong>Promote your top ten intents to classic answers.</strong> A classic answer costs 1 credit, a generative answer 2, and a grounded generative answer 12. FAQ traffic is highly repetitive, so authoring the common paths pays back quickly.</li><li><strong>Set per-agent monthly limits.</strong> Microsoft documents a per-agent consumption cap in the Power Platform admin center (Licensing → Copilot Studio → Manage Agents). Use it so one runaway agent can’t trigger tenant enforcement.</li><li><strong>Route employees through the licensed path.</strong> Where staff already hold M365 Copilot, publish the agent into Copilot Chat, Teams or SharePoint under the user’s identity, and trigger flows with “When an agent calls the flow”.</li><li><strong>Match model and harness to the job.</strong> Don’t send a lookup through premium reasoning (10 credits per 1,000 tokens extra), and keep reasoning-heavy GitHub Copilot harness agents out of production until their credit profile is measured.</li></ol><h3>Run your own numbers</h3><p>This calculator uses the standard-harness rates above. It models one agent; run it once per agent and add the results. It is a planning aid, not a quote: Microsoft’s usage estimator and your own admin-centre data should confirm the final figures.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/916/1*QVYbKJPVNXOzair4J-Yn0Q.png" /><figcaption>Calculator Link: <a href="https://peafowlit.com/blog/copilot-studio-credits-pricing-2026/">https://peafowlit.com/blog/copilot-studio-credits-pricing-2026/</a></figcaption></figure><p><strong>The five-line checklist</strong></p><p>1. Inventory every agent and its harness. Flag anything on the GitHub Copilot harness.</p><p>2. Pull 60 days of credit consumption per agent from the Power Platform admin center.</p><p>3. Price each agent in the calculator with its real conversation mix, not an average.</p><p>4. Buy packs for the floor you reliably hit above 20,000 credits per pack; leave the rest to PAYG.</p><p>5. Cap each agent and review consumption monthly.</p><h3>Get a free Copilot Studio cost assessment</h3><p>Send us your agent inventory and we’ll model credit spend, recommend a pack/PAYG/pre-purchase mix, and flag the GitHub Copilot harness agents that are about to start billing. A short working session, a written summary, no obligation.</p><p><a href="https://peafowlit.com/contact/?topic=copilot-studio-cost-assessment"><strong>Book your Copilot Studio Cost Assessment</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=44a5b1e5b265" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Build a SharePoint Intranet 80% Faster (Without Writing Code)]]></title>
            <link>https://medium.com/@peafowlit/how-to-build-a-sharepoint-intranet-80-faster-without-writing-code-aace02385d41?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/aace02385d41</guid>
            <category><![CDATA[sharepoint-spfx]]></category>
            <category><![CDATA[spfx]]></category>
            <category><![CDATA[sharepoint-framework]]></category>
            <category><![CDATA[sharepoint-intranet]]></category>
            <category><![CDATA[intranet]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 05:48:57 GMT</pubDate>
            <atom:updated>2026-10-01T05:48:57.714Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AQUnNW535f1pCTFzMC3NKg.jpeg" /><figcaption>How to Build a SharePoint Intranet 80% Faster (Without Writing Code)</figcaption></figure><p>Most intranet projects don’t stall on planning. They stall waiting for developers to build the same homepage web parts every company needs. Here’s how to skip that part and launch in weeks, not months.</p><p>Ask any IT manager how long their last intranet took, and you’ll rarely hear “a few weeks.” Six months is common. A year isn’t unusual. And after all that effort, the result often isn’t used much: SWOOP Analytics’ 2026 SharePoint Intranet Benchmarking Report, based on usage data from over 410,000 employees across 41 organizations, found that nearly everyone opens the intranet but far less of it is actually read or trusted.</p><p>That’s two problems. Intranets take too long to build, and the time spent building leaves little for the work that drives adoption: good content, clear ownership, and features employees actually interact with.</p><p>This guide addresses both. We’ll show where the months go in a traditional SharePoint intranet project, which of those phases you can remove entirely, and a step-by-step, no-code process to launch a modern employee portal using ready-made SharePoint Framework (SPFx) web parts.</p><blockquote><strong>Who this is for</strong></blockquote><blockquote>IT managers and SharePoint admins who need a working intranet quickly; HR and internal comms leads who own the content; and business leaders deciding whether a six-figure custom build is really necessary.</blockquote><h3><strong>Where the months actually go</strong></h3><p>A typical custom SharePoint intranet for a mid-sized organisation runs through six phases. The estimates below are illustrative, based on common project patterns, and your numbers will vary with scope. The pattern is consistent, though: the biggest block of time is building and testing web parts that nearly every intranet needs.</p><figure><img alt="Phase 
 
 Custom build 
 
 With accelerator kit 
 
 Why it changes 
 
 Discovery &amp; information architecture 
 
 2 weeks 
 
 1 week 
 
 Still essential, but a known component set narrows the decisions 
 
 UX design &amp; branding 
 
 2 weeks 
 
 2–3 days 
 
 Layouts are pre-designed; you apply your theme and imagery 
 
 Web part development 
 
 5 weeks 
 
 0 
 
 Banner, links, events, polls, directory already exist 
 
 Testing &amp; QA 
 
 2 weeks 
 
 2–3 days 
 
 Components are pre-tested; you test your" src="https://cdn-images-1.medium.com/max/613/1*8eT32YTMG0Oj5BtW9irnDg.png" /></figure><p>The saving isn’t a trick. It comes almost entirely from <strong>not rebuilding commodity components</strong>. Every company intranet needs a hero banner, quick links to key apps, an events calendar, a place for policies, and some way to get employee feedback. Paying developers to rebuild those from scratch is where most intranet budgets go.</p><h3><strong>What’s in the SharePoint Productivity Accelerator Kit</strong></h3><p>The <a href="https://peafowlit.com/sharepoint-productivity-accelerator-kit/">SharePoint Productivity Accelerator Kit</a> is a set of SPFx web parts built by Peafowl IT for modern SharePoint Online pages. It runs inside your own Microsoft 365 tenant and is added to pages through the standard SharePoint editor, so site owners don’t need to learn a new tool.</p><h4><strong>Free tier</strong></h4><p>Everything you need for a working homepage.</p><ul><li><strong>Banner Slider</strong>: rotating hero for announcements and campaigns</li><li><strong>Quick Links</strong>: one-click access to HR, IT and line-of-business apps</li><li><strong>Events</strong>: meetings, celebrations and company activities</li><li><strong>Employee Corner</strong>: policies, announcements and resources</li><li><strong>Polls</strong>: quick pulse checks and feedback</li></ul><h4><strong>Premium tier</strong></h4><p>Engagement and people features.</p><ul><li><strong>Employee Directory</strong>: find colleagues, departments and contact details</li><li><strong>Employee Shoutouts</strong>: public recognition of achievements</li><li><strong>Applause</strong>: an ongoing culture-of-recognition feed</li></ul><p>Because the free web parts are fully usable on their own, you can launch with them and add premium features later, once you’ve seen how employees use the portal.</p><h3><strong>The no-code build: 6 steps from blank site to live intranet</strong></h3><h4>01 Plan your structure (days 1–5)</h4><p>Decide on a <strong>communication site</strong> as your intranet home and, if you have multiple departments, register it as a <strong>hub site</strong> so department sites share navigation and branding. List the 10–15 links employees ask for most, the recurring events you want visible, and the policies that belong in Employee Corner.</p><p>Most importantly, name a content owner for each section. This does more for long-term readership than any feature.</p><h4>02 Install the kit (under an hour)</h4><p>Get the kit from <a href="https://marketplace.microsoft.com/en-us/product/WA200011163?utm_source=peafowl-blog&amp;utm_medium=cta&amp;utm_campaign=intranet-80-faster&amp;utm_content=step2">Microsoft Marketplace</a>, or upload the .sppkg package directly to your tenant’s Apps site (the app catalog). Per <a href="https://learn.microsoft.com/en-us/sharepoint/use-app-catalog">Microsoft’s documentation</a>, adding apps there requires the <strong>SharePoint Administrator</strong> role or above.</p><h4>03 Choose tenant-wide or site-level deployment</h4><p>When you enable the app, SharePoint offers <strong>“Enable this app and add it to all sites.”</strong> Selecting it makes the web parts appear in the web part picker on every site immediately, which Microsoft calls <a href="https://learn.microsoft.com/en-us/sharepoint/dev/spfx/tenant-scoped-deployment">tenant-scoped deployment</a>. Leave it unchecked if you’d rather pilot on the intranet site first and roll out to departments later.</p><blockquote><strong>Admin tip</strong></blockquote><blockquote>Microsoft recommends testing solutions before broad deployment. A 30-minute check on a test communication site is enough for a pre-built kit.</blockquote><h4>04 Build the homepage layout (1 Day)</h4><p>Edit your home page, click <strong>+</strong>, and add the accelerator web parts. A layout that works well for most organisations:</p><ul><li>Banner Slider</li><li>Employee Corner</li><li>Quick Links</li><li>Events</li><li>Polls</li><li>Shoutouts (Premium)</li></ul><p>Put the Banner Slider at the top for high-visibility messages, Quick Links in the right column where people scan for them, and interactive elements (Events, Polls, recognition) lower on the page to encourage scrolling.</p><h4>05 Configure and brand (1–2 days)</h4><p>Each web part is configured through its <strong>property pane</strong>: titles, links, images, event sources and poll questions. No code, no developer tickets. Apply your organisation’s SharePoint theme so colours and typography match your brand across hub-connected sites.</p><h4>06 Launch, then measure (week 3 onward)</h4><p>Announce the launch with a Banner Slider slide and a first Poll (“What should we add next?”), which makes launch day interactive. After launch, check SharePoint’s built-in site usage analytics weekly for the first month and remove or rework sections nobody visits.</p><h3><strong>Start with the free web parts today</strong></h3><p>Install from Microsoft Marketplace and have your homepage running this week. Upgrade to premium only when you need it.</p><p><a href="https://marketplace.microsoft.com/en-us/product/WA200011163?utm_source=peafowl-blog&amp;utm_medium=cta&amp;utm_campaign=intranet-80-faster&amp;utm_content=mid"><strong>Get it on Microsoft Marketplace</strong></a></p><h3><strong>Why this matters more in 2026</strong></h3><p>Two developments make a fast, modern rebuild more urgent this year:</p><ul><li><strong>SharePoint add-ins are retired.</strong> Microsoft <a href="https://learn.microsoft.com/en-us/sharepoint/use-app-catalog">retired SharePoint add-ins</a> for SharePoint in Microsoft 365 starting 2 April 2026. Older intranets that relied on add-in-based components need modern SPFx replacements, and rebuilding each one by hand is exactly the slow path this guide avoids.</li><li><strong>Copilot reads your intranet.</strong> Microsoft 365 Copilot grounds its answers in SharePoint content. A well-structured, current intranet with clear policies improves the quality of Copilot’s answers, while an outdated one lowers it.</li></ul><h3><strong>When an accelerator isn’t enough</strong></h3><p>Here’s where a kit isn’t the right fit on its own:</p><ul><li><strong>Deep line-of-business integrations.</strong> If your homepage needs live data from an ERP, HR system or ticketing tool, you’ll need custom SPFx or Power Platform work alongside the kit.</li><li><strong>Complex workflows.</strong> Approvals, onboarding checklists and request forms belong in Power Automate and Power Apps, not in homepage web parts.</li><li><strong>Strict, bespoke design systems.</strong> If brand guidelines require pixel-exact custom layouts, expect some front-end customisation.</li><li><strong>Heavily regulated content.</strong> Records management, retention and sensitivity labelling need a Purview and information architecture plan, whatever web parts you use.</li></ul><p>Even in these cases, the accelerator still helps by covering the standard components, so custom development budget goes only to what’s specific to your business.</p><h3><strong>A real-world example: logistics company</strong></h3><p>A 300-person logistics company had originally planned a custom SharePoint intranet that was estimated to take around 10–12 weeks to design, develop and deploy. Using the SharePoint Productivity Accelerator Kit, the team went live in just 3 weeks.</p><p>The standard homepage components announcements, quick links, events, employee resources and polls were all delivered using ready-made SPFx web parts. The only custom work was connecting a small number of internal logistics resources and configuring the company’s branding and navigation.</p><p>The biggest lesson? <strong>Don’t spend your development budget rebuilding features that employees already expect.</strong> By starting with proven, ready-made components, the team was able to focus its time on content, ownership and the information employees actually needed to do their jobs.</p><h3><strong>Frequently asked questions</strong></h3><h4><strong>Do I need a developer to build a SharePoint intranet with the kit?</strong></h4><p>No. The web parts are pre-built SPFx components. A SharePoint admin installs the package once, and site owners add and configure the web parts through the normal page editor and property panes.</p><h4><strong>Is the SharePoint Productivity Accelerator Kit free?</strong></h4><p>It uses a freemium model. Banner Slider, Quick Links, Events, Polls and Employee Corner are free. Employee Directory, Employee Shoutouts and Applause are part of the premium package.</p><h4><strong>What permissions do I need to install it?</strong></h4><p>Adding the package to the tenant Apps site requires the SharePoint Administrator role or above. After that, any site owner can add the web parts to their pages.</p><h4><strong>Where does the 80% time saving come from?</strong></h4><p>Mostly from removing custom web part development, QA and deployment, the longest phases of a traditional build. Planning and content still take time and shouldn’t be skipped.</p><h4><strong>Can I use it on an existing SharePoint intranet?</strong></h4><p>Yes. The web parts can be added to any modern SharePoint page, so you can improve an existing portal section by section instead of rebuilding it.</p><h3><strong>Launch your intranet this month, not next year</strong></h3><p>Install the free web parts from Microsoft Marketplace and build your homepage today. If you’d like a walkthrough, or help with integrations and governance, our SharePoint team can show you the full kit in 30 minutes.</p><p><a href="https://marketplace.microsoft.com/en-us/product/WA200011163?utm_source=peafowl-blog&amp;utm_medium=cta&amp;utm_campaign=intranet-80-faster&amp;utm_content=final"><strong>Get it free on Microsoft Marketplace</strong></a></p><p><a href="https://peafowlit.com/get-in-touch/"><strong>Book a 30-min intranet demo</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aace02385d41" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GitHub Copilot now takes work straight from your Teams meetings.]]></title>
            <link>https://medium.com/@peafowlit/github-copilot-now-takes-work-straight-from-your-teams-meetings-2f264e896f4c?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f264e896f4c</guid>
            <category><![CDATA[microsoft-copilot]]></category>
            <category><![CDATA[github-copilot]]></category>
            <category><![CDATA[microsoft-teams]]></category>
            <category><![CDATA[co-pilot]]></category>
            <category><![CDATA[microsoft-power-platform]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 05:07:43 GMT</pubDate>
            <atom:updated>2026-09-24T05:07:43.442Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>GitHub Copilot now takes work straight from your Teams meetings. Is your organisation ready to govern it?</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cH2BNTf9bPjMmp29mZGR3g.jpeg" /><figcaption><strong>GitHub Copilot now takes work straight from your Teams meetings. Is your organisation ready to govern it?</strong></figcaption></figure><p>A decision made in Monday’s standup can become a pull request before the call ends. That’s a real productivity gain, and it quietly changes who can start code changes, what data the AI sees, and where your AI spend comes from.</p><h3><strong>What actually changed</strong></h3><p>In August 2026, GitHub released a new public preview that lets anyone in a Microsoft Teams channel, thread or direct message type @GitHub and start a GitHub Copilot cloud agent session.</p><p>This isn’t a notification bot. The agent investigates the problem, plans a fix, writes code and opens issues or pull requests. It runs asynchronously in a secure cloud sandbox and posts results back to the conversation. Everyone in the thread can see what it’s doing, add context and steer it. Developers can then pick up the agent’s work in their IDE, the terminal or the GitHub Copilot app.</p><p>GitHub’s own example is the one that will get leadership’s attention. You hand an action item to Copilot during a standup, and it starts resolving the issue before the meeting has ended.</p><blockquote>In plain terms</blockquote><blockquote>Teams becomes a place where software work gets started, not just discussed. For many organisations running Microsoft 365, that’s the first time a chat message can directly produce a change to production code.</blockquote><h3><strong>How it works, end to end</strong></h3><ol><li><strong>Discuss: </strong>Team talks through a bug or request in a Teams thread or meeting chat</li><li><strong>Mention: </strong>Someone with repo write access types @GitHub plus the task</li><li><strong>Work: </strong>Copilot uses the whole thread as context and works in a cloud sandbox</li><li><strong>Output: </strong>Issue or pull request opened in the channel’s default repository</li><li><strong>Gate: </strong>Human review and approval rules decide whether it merges</li></ol><p><strong>A few details matter more than they first appear:</strong></p><ul><li><strong>Repository targeting.</strong> Users can name a repository and base branch in the prompt (repo=OWNER/REPO branch=BRANCH). If they don’t, Copilot uses the channel’s default repository. If no default has been set, <strong>the first repository anyone uses in that channel becomes the default</strong>.</li><li><strong>Iteration.</strong> Mentioning @GitHub again in the same thread refines the pull request, using every message since the previous mention.</li><li><strong>Issue creation.</strong> Copilot can turn a discussion into one issue or several linked parent and child issues.</li></ul><h3><strong>The business case: why leaders will want this</strong></h3><p>For business users and product owners, the value is the shorter distance between a decision and working code.</p><ul><li><strong>Less lost context.</strong> The reasoning behind a change, captured in the Teams discussion, travels with the work instead of being retyped into a ticket.</li><li><strong>Visible progress.</strong> Non-developers can watch an investigation happen in the channel rather than waiting for a status update.</li><li><strong>Faster triage.</strong> Routine fixes, small features, logging and refactors can start while the team is still talking about them.</li><li><strong>It meets people where they already work.</strong> Organisations standardised on Microsoft 365 don’t need another tool for this; the agent comes to Teams.</li></ul><p>That’s why this feature will spread quickly once a few developers try it. It’s also why IT should set the ground rules before that happens, not after.</p><h3><strong>The four governance questions IT has to answer</strong></h3><h4>1. Whose identity is making the change?</h4><p>This depends on where Copilot is invoked, and the difference is significant.</p><figure><img alt="1. Whose identity is making the change? This depends on where Copilot is invoked, and the difference is significant." src="https://cdn-images-1.medium.com/max/632/1*WOS54X-ixbZNFR_Qc35_VA.png" /></figure><p>Only users with <strong>write</strong> access to a repository can make Copilot change it, though anyone in the conversation can add input. Guest members and outside collaborators can’t start or steer a session. That’s a sensible default, but it means your repository permissions now indirectly define who can start code changes from Teams.</p><h4>2. Is there a human gate before agent code merges?</h4><p>GitHub added a specific control for this. Because shared-context pull requests aren’t attributed to a person, repository rulesets require <strong>one additional approval</strong> before they merge. If a repository requires two approvals, Copilot-created pull requests from Teams need three. This is on by default.</p><h4>3. What data is the AI seeing, and where does it end up?</h4><p>This question is the easiest to overlook. When you mention @GitHub in a thread, Copilot <strong>captures the entire thread as context</strong>, and that context is stored in the artifacts the agent generates, such as the issue or pull request.</p><p>Think about what gets said in a typical Teams thread: customer names, commercial details, credentials pasted “just for a second”, opinions about colleagues. If that thread becomes the context for a pull request, some of it may end up in your repository history.</p><p>GitHub’s own advice is to use a direct message to the app when you want to limit the context, or to start a fresh thread for focused work.</p><blockquote>Also changing on 28 September</blockquote><blockquote>Separately, GitHub is merging Copilot Chat on github.com, Copilot Chat in GitHub Mobile and Copilot cloud agent into one experience under a single policy, <strong>enabled by default</strong>, no earlier than 28 September 2026. As part of this, chat data on github.com will be retained <strong>for the life of the account instead of 28 days</strong>. Check this against your records retention and privacy commitments.</blockquote><h4>4. Who is paying, and how much?</h4><p>Copilot cloud agent sessions started from Teams consume AI credits, which are governed by usage-based billing budgets. <strong>Cloud sandbox usage is billed separately</strong> and needs its own product-level or SKU-level budget. A popular channel with an enthusiastic team can generate agent sessions all day, and every one of them has a cost.</p><p><strong>GitHub has shipped two things that help here:</strong></p><ul><li><strong>Budget increase requests</strong> (generally available since 16 September). When members run out of AI credits, they can request more, and owners or billing managers can approve, adjust or deny the request from settings. This is available on Copilot Business and Enterprise under usage-based billing, but not for enterprises with managed users.</li><li><strong>Feature engagement in the Copilot impact dashboard</strong> (17 September). Admins can see how many active users regularly use each Copilot feature, which helps separate real adoption from one-off experiments.</li></ul><h3><strong>When not to use it (yet)</strong></h3><p>We’re positive about this feature, but it isn’t right for every team today. We’d hold back if any of these apply:</p><ul><li><strong>Your repositories don’t enforce required reviews.</strong> Fix branch protection first, or the additional-approval safeguard does nothing.</li><li><strong>The channel handles sensitive or regulated discussion.</strong> Finance, HR, legal and customer-escalation channels shouldn’t double as coding context. Keep agent work in dedicated engineering channels.</li><li><strong>You rely on it being production-ready.</strong> It’s a public preview and may change. It’s fine for pilots, but don’t build a critical process on it yet.</li><li><strong>You haven’t set budgets.</strong> Without AI credit and sandbox budgets, you’ll find out about usage from the invoice.</li><li><strong>Many people have write access to many repositories.</strong> If “everyone can write everywhere” is your current model, Teams makes that exposure more visible and easier to trigger.</li></ul><h3><strong>Dates your IT team should have in the calendar</strong></h3><p>28 Sep 2026 — <strong>Unified Copilot policy.</strong> Review the new <em>Copilot cloud agent</em> policy in Copilot settings on github.com before this date. The unified experience is on by default and extends chat retention. Opting out removes Copilot on github.com and Mobile.</p><p>28 Sep 2026 — <strong>Code review default changes from Lite to Balanced.</strong> If you want to keep Lite, select it explicitly at the organisation or repository level beforehand.</p><p>1 Oct 2026 — <strong>Upfront seat billing</strong> for existing Copilot Business and Enterprise customers paying by credit card or PayPal. Prices aren’t changing, but assigned seats are charged upfront, and revoking a seat doesn’t give a prorated refund.</p><p>19 Oct 2026 — <strong>Model deprecations.</strong> Selected models will be retired across Copilot Chat, agent modes and code completions. Check any workflows or custom agents pinned to specific models.</p><h3><strong>An 8-point governance checklist before you switch it on</strong></h3><ol><li><strong>Decide the pilot scope.</strong> Start with one or two engineering channels and named repositories, not a tenant-wide rollout.</li><li><strong>Enforce required reviews</strong> on every repository the agent can reach, so the extra approval for Copilot pull requests actually applies.</li><li><strong>Set channel default repositories deliberately</strong> using @GitHub settings, instead of letting the first session choose one.</li><li><strong>Audit write access.</strong> Trim broad write permissions, since write access now also means “can start code changes from Teams”.</li><li><strong>Publish a channel hygiene rule.</strong> No secrets or sensitive customer data in threads where @GitHub is used. Use direct messages or new threads for focused work.</li><li><strong>Set budgets for both AI credits and cloud sandboxes</strong>, and assign owners for budget increase requests.</li><li><strong>Review the unified Copilot policy and retention change</strong> before 28 September, alongside your Microsoft Purview and records retention position.</li><li><strong>Measure after 30 days</strong> using the impact dashboard’s feature engagement data, then decide whether to expand, adjust or pause.</li></ol><h3><strong>The Peafowl view</strong></h3><p>In our <a href="https://peafowlit.com/microsoft-365-development/">Microsoft 365 and development engagements</a>, the same pattern keeps coming up. Organisations put real effort into governing Microsoft 365 Copilot through sensitivity labels, SharePoint permissions and data loss prevention, while GitHub is managed separately by the engineering team. Features like this one sit between the two.</p><p>A Teams thread is Microsoft 365 content. The pull request it produces is GitHub content. The governance question, “what did the AI see and what did it change?”, crosses both. We’d recommend treating GitHub Copilot in Teams as part of your overall Copilot governance programme, with the same owners and review cadence, rather than as a developer tool that IT finds out about later.</p><p>Done well, this is one of the more practical agentic AI features we’ve seen this year: it’s visible, collaborative and gated by human review. Done carelessly, it becomes a way for chat messages to reach production code with less oversight than a normal ticket.</p><h4><strong>Get the speed without losing control</strong></h4><p>Our GitHub Copilot Governance &amp; Adoption Assessment reviews your GitHub and Microsoft 365 setup together and gives you a clear rollout plan. It covers:</p><ul><li>Repository permissions, rulesets and required-review coverage</li><li>Copilot policies, including the 28 September unified policy change</li><li>AI credit and cloud sandbox budget design</li><li>Teams channel hygiene and data exposure</li><li>A scoped pilot plan with success measures</li></ul><p><a href="https://peafowlit.com/contact-us/"><strong>Book your Copilot Governance Assessment →</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f264e896f4c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The New Copilot Harness in Copilot Studio: What It Actually Changes for the Agents You’re Building]]></title>
            <link>https://medium.com/@peafowlit/the-new-copilot-harness-in-copilot-studio-what-it-actually-changes-for-the-agents-youre-building-f5710e67559b?source=rss-668919ceb248------2</link>
            <guid isPermaLink="false">https://medium.com/p/f5710e67559b</guid>
            <category><![CDATA[copilot-studio-harness]]></category>
            <category><![CDATA[copilot-harness]]></category>
            <category><![CDATA[harness-engineering]]></category>
            <category><![CDATA[agent-harness]]></category>
            <dc:creator><![CDATA[Peafowl IT Solution]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 06:47:45 GMT</pubDate>
            <atom:updated>2026-09-17T06:47:45.736Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HWDetW_Iy1PPM5Y3z6bMYw.jpeg" /></figure><p>Microsoft has quietly changed the engine underneath Copilot Studio. Before your team builds or rebuilds another agent, here’s what a “harness” is, why it now comes in two flavours, and how to decide which one belongs under the hood.</p><h3>What is a “harness,” really?</h3><p>If you’ve opened Copilot Studio recently and noticed the home page looks different, you’re not imagining it. Microsoft has introduced a second <strong>harness,</strong> a new authoring and runtime option that sits alongside the classic experience every existing agent was built on.</p><p>A harness is the operating layer between the underlying AI model and your agent’s configuration. It’s not the model itself; it’s the layer that decides how the model receives context, when it reaches for a tool or a knowledge source, how it interprets the results it gets back, and how it moves through a task toward completion. Put simply: the model supplies the reasoning; the harness decides how that reasoning gets used.</p><p>This matters because harness selection is <strong>an architecture decision, not an interface preference</strong>. It changes how a solution interprets a request, how it recovers from a failed step, and, critically for anyone budgeting a Copilot rollout, how it gets billed.</p><blockquote><strong>The short version:</strong></blockquote><blockquote>Copilot Studio now runs on two harnesses side by side: the <strong>standard harness</strong> (the flow-and-topic model most existing agents already use) and the new <strong>GitHub Copilot harness</strong> (a natural-language-first, reasoning-driven model built for longer, messier business processes). A third option, the <strong>Copilot Chat harness</strong>, extends Microsoft 365 Copilot Chat itself rather than building a standalone agent.</blockquote><h3>Why Microsoft built a second one</h3><p>The standard harness has carried Copilot Studio well for rule-based, conversational agents the kind that ask a clarifying question, follow a defined topic, and hand off cleanly when the conversation leaves the script. What it was never designed for is <strong>ambiguous, multi-step, multi-system work</strong> where the “correct” path depends entirely on what the agent finds along the way.</p><p>The GitHub Copilot harness is Microsoft’s answer to that gap. It’s built on the same orchestration engine behind Copilot Cowork and the GitHub Copilot coding agent, and it’s designed to run against frontier reasoning models. Instead of authoring explicit topics and branching logic, you describe the agent’s goal in natural language, connect the tools and knowledge it needs, and let the orchestration runtime plan the steps itself, including replanning when a step fails or a result changes what should happen next.</p><p>For business users, this sounds abstract until you picture a real process: an accounts-payable agent that reads an invoice, tracks down the matching purchase order, compares the figures, flags a discrepancy, requests the missing evidence from the right person, and only then routes the exception, adjusting at every step based on what it actually finds. That’s the class of work the standard harness’s fixed topics were never meant to carry.</p><h3>Standard harness vs. Copilot harness</h3><p>Here’s the comparison worth pinning to your project wall before your next build decision:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/702/1*-QDwe6Hoe7zT0oSbA7MRzQ.png" /></figure><p>Worth noting: there’s technically a <strong>third</strong> harness, the Copilot Chat harness, which isn’t a standalone agent at all. It grounds Microsoft 365 Copilot Chat in your organisation’s content (SharePoint sites, policy documents, approved sources) and respects the user’s existing Microsoft 365 permissions, so employees stay inside the Copilot Chat interface they already use rather than jumping to a separate agent experience.</p><blockquote><strong>The one-way door</strong></blockquote><blockquote>You choose a harness when you create the agent, and today, an agent built on one harness <strong>cannot</strong> be transferred to the other. Standard-harness agents don’t migrate into the GitHub Copilot harness, and vice versa. This makes the choice worth deliberately slowing down for, not defaulting on.</blockquote><h3>Which one should you build on?</h3><p>Skip the feature list for a moment and ask these five questions, in order:</p><ol><li><strong>Does the work need an independent agent, or should it extend Microsoft 365 Copilot Chat?</strong> If employees should never leave their existing Copilot Chat interface, the Copilot Chat harness may be the right call before you even compare the other two.</li><li><strong>Must the agent follow explicit conversation paths, or reason dynamically toward a goal?</strong> Predictable, auditable steps point to standard. Ambiguous, judgment-call work points to GitHub Copilot.</li><li><strong>Does the work require multiple tools, memory, or recovery from mid-task failures?</strong> That’s squarely GitHub Copilot harness territory.</li><li><strong>Where do users need to access the finished agent?</strong> Both standard and GitHub Copilot harnesses publish to internal teams or external customers; Copilot Chat stays internal.</li><li><strong>Which licensing and consumption model actually fits your budget?</strong> Copilot Credits scale with usage and complexity for high-volume, simple interactions, standard licensing can still be the more predictable spend.</li></ol><p>If your honest answer to questions 2 and 3 leans toward “we can’t script every path in advance,” that’s your signal, not the marketing copy, not which harness looks newer on the home page.</p><h3>Migration gotchas nobody mentions</h3><p>A few things we’d flag before your team commits a build to either path:</p><ul><li><strong>No agent portability.</strong> As above, this is an architecture decision made at creation time, not a toggle you flip later.</li><li><strong>Copilot Credits change your cost model.</strong> Building, testing, and evaluating GitHub Copilot harness agents consumes credits, not just running them in production. Teams used to flat Copilot Studio licensing should model this before scaling a pilot.</li><li><strong>Existing standard-harness topics and agent flows remain fully supported.</strong> Nothing forces a rebuild; the standard harness isn’t being deprecated, and reusing existing topics, prompts, and flows is still a first-class path.</li><li><strong>“Natural-language-first” still needs boundaries.</strong> Describing an agent’s goal in plain language reduces the flow-authoring burden, but it doesn’t remove the need for clear instructions, scoped tools, and defined limits; an under-specified GitHub Copilot harness agent will improvise in directions you didn’t intend.</li></ul><h3>What we’re seeing with clients</h3><p>In our own Copilot Studio engagements across SharePoint-grounded knowledge agents, Power Platform automation, and M365 Copilot extensions, the pattern holds up: agents that answer from a defined knowledge base or walk a known process HR policy lookups, IT ticket intake, equipment requests are staying on the standard harness, and rightly so. The candidates we’re steering toward the GitHub Copilot harness are the ones with a genuine multi-system, multi-step shape: reconciliation-style work, document-heavy approvals, and anything that currently needs a human to chase down missing information before a process can close.</p><p>The mistake we’d caution against is defaulting to the newer harness because it’s newer. A well-built standard-harness agent that does one bounded job reliably will usually outperform and definitely outcost an over-engineered GitHub Copilot harness agent solving the same bounded problem.</p><h3>Key takeaways</h3><ul><li>A harness is the orchestration layer between the model and your agent it decides how context, tools, and results get used, and it’s an architecture decision, not a UI preference.</li><li>Copilot Studio now has three: <strong>standard</strong> (predictable, topic-driven), <strong>GitHub Copilot</strong> (reasoning-driven, multi-step, credit-billed), and <strong>Copilot Chat</strong> (extends M365 Copilot Chat with your organisation’s knowledge).</li><li>The choice is made at build time and is currently one-way; agents don’t migrate between harnesses.</li><li>Match the harness to the shape of the problem, not the age of the feature. Bounded, well-understood processes still belong on standard.</li></ul><h3>Not sure which harness your next agent needs?</h3><p>We’ll walk your use case against both architectures, model the Copilot Credits cost against standard licensing, and hand you a build recommendation before a single flow gets built.</p><p><a href="https://peafowlit.com/contact"><strong>Book a Copilot Studio Agent Architecture Review</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f5710e67559b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>