PowerShell: Export Microsoft 365 Copilot Agent Inventory and Availability Assignments
Hi everyone, I needed a way to export Microsoft 365 Copilot agent inventory and availability assignments from the Microsoft 365 Admin Center, but couldn't find a built-in export option. After investigating the admin portal's network traffic, I built a PowerShell script that uses the same internal API consumed by the Microsoft 365 Admin Center to export all Copilot agents to CSV. ### Features - Exports all Microsoft 365 Copilot agents - Automatically follows pagination (`nextLink`) - Exports: - Agent Name - App ID - Title ID - Publisher - Created By - Availability Settings - Allowed Users / Groups - Assignment Information - Deployment Information - Version Information - Timestamps - CSV output ### Tested The script has been tested against a tenant containing 482 Copilot agents and successfully exported the complete inventory. ### GitHub https://github.com/gwestergren/M365-Copilot-Agent-Inventory ### Notes - Uses an authenticated browser session cookie from the Microsoft 365 Admin Center. - Uses the same internal API currently consumed by the admin portal. - This is an undocumented API and Microsoft may change it at any time. Feedback, testing results, and improvements are welcome. Here are some screen shots: output to csv Successful run of the script
Prompt Lab: Three Critical Bugs That Make It Unusable as a Prompt Management Tool
Product: Microsoft 365 Copilot — Prompt Lab (accessed via ... button in Copilot Chat) Date: June 1, 2026 Environment: M365 Copilot, Web (Edge), Work IQ mode Bug #1: Empty State Crash Steps to reproduce: Open Prompt Lab via the ... button in Copilot Chat Ensure "Your saved prompts" contains zero prompts (either as a new user or by deleting all saved prompts) Observe the result Expected behavior: An empty state placeholder (e.g., "You haven't saved any prompts yet.") Actual behavior: The entire Prompt Lab panel throws a red error banner: "Something went wrong. Please close the dialog and try again later." The "Your saved prompts" category does not render at all. The panel only shows Microsoft's preset categories (Prompt topics, Agent prompts). Why this matters: An empty container is a valid state — it is literally every new user's initial state. A UI component should never crash because a list has zero items. This is a null/empty array handling failure that should have been caught by basic QA. Bug #2: Search Does Not Index User's Saved Prompts Steps to reproduce: Save a custom prompt with a distinctive title (e.g., "AI每日新闻") and body containing the keyword "AI" Open Prompt Lab Use the search box at the bottom to search for "AI" Expected behavior: Search results include the user's saved prompt alongside Microsoft's preset prompts. Actual behavior: Only Microsoft preset prompts matching "AI" are returned (e.g., "Stay on top of AI," "Prompt Compliance"). The user's own saved prompt — whose title and body both contain "AI" — does not appear in the results. Why this matters: The search box creates a false expectation that it searches all prompts. In reality, it only indexes Microsoft's template library. This means as a user accumulates more saved prompts, the only way to find one is manual scrolling. A search function that excludes user-created content is fundamentally broken by design. Bug #3: Saved Prompts Lost During Migration Context: The standalone Copilot Lab / Prompt Gallery app was retired on July 15, 2025, and its functionality was merged into the built-in Prompt Lab within Copilot Chat. What happened: All previously saved prompts from the old Copilot Lab app are gone. They do not appear in the new Prompt Lab's "Your saved prompts" section. There was no migration notice, no export tool for end users, and no recovery path. Why this matters: Users invested time curating and refining their prompt libraries. Silently dropping that data during a platform migration — without warning, backup, or migration tooling — is a breach of user trust. Summary These three issues compound into a single conclusion: Prompt Lab is currently non-functional as a prompt management tool. Capability Status Reliable storage ❌ Data lost during migration Empty state handling ❌ Crashes when empty Search / retrieval ❌ Does not index user content Displaying Microsoft templates ✅ Works The only feature that works correctly is showcasing Microsoft's own preset prompts. For a tool whose entire purpose is to help users save, organize, and reuse their own prompts, this is an unacceptable state of quality. I'd strongly recommend the team prioritize: (1) proper null-state handling, (2) including user prompts in the search index, and (3) investigating whether migrated prompt data can be recovered from Substrate.
Feature Proposal: OS-level Intelligent Task Organizer (Windows + Copilot)
A Idea about Intelligent Tasks organizer, I have to remember a lot of things during the team meetings like what is been said (we'll schedule a call or follow up etc.,) and what has been communicated in the emails (I'll get back to you after 2 weeks, or call us after two weeks) , And notes that I took in the notepad, or notepad ++,or sticky notes, or word, or one note. I want to chronologically display tasks on the right hand side of the laptop screen just like sticky note and it shall display all tasks one by one, it shall remove tasks are already complete (email sent with confirmation). and arrange, adjust every few mins according to priority/time or user added priority. App shall display small icon (just like chat) upon clicking it shall display ordered list of tasks. and desktop apps like teams/note/word/notepad++,sticky notes can participate by default or other apps like notepad++ can be onboarded manually in to the app. You can use a local model which infers the meaning of “I’ll call you in two weeks” - who is “I”? you or them? “Let’s follow up later” - task or casual statement? “I sent it” - which task did this complete? You can use a local model such that Corporate Teams/Outlook access may allow by corporate policy. Need to put much emphasis on false positives if the app keeps inventing tasks. Do not need to bring big LLMs in to the picture for inference, because of corporate policies may not allow. Microsoft provides operating system,office 365, tools with copilot, the inference can be possible because of all apps/content can be accessible at os level. Problem: Users capture tasks across multiple tools: Teams meetings and chats Outlook emails Notes (OneNote, Notepad, Sticky Notes, Word) Tasks become fragmented, untracked, and often lost. Proposed Solution: A lightweight system-level task layer integrated with Windows + Copilot that: Core Features Automatic task extraction From Teams, Outlook, notes, and user text Example interpretations: “I’ll call you in 2 weeks” “Let’s follow up later” Context-aware inference (local model) Identify: Task owner (“I” vs “you”) Priority signals Deadlines Minimize false positives Chronological task timeline Tasks auto-organized by: Time Priority Recency Floating task panel (desktop UI) Docked widget (like Sticky Notes or chat bubble) Expand/collapse view Always visible option Automatic task lifecycle tracking Detect completion: “Email sent” “File shared” Remove or mark complete automatically Continuous re-prioritization Adjust every few minutes based on: New inputs Deadlines User behavior Privacy-first architecture Use local models (SLM) instead of large cloud LLMs Enterprise admin control for data access Why this matters: Millions of users manually track tasks across fragmented tools, losing productivity daily. This feature would unify task understanding across the OS and M365 ecosystem.Add a "Print Conversation" Feature in Microsoft Copilot
I'm warming up to using Copilot for quick answers and support. It is even better, more detailed and more honest about Microsoft shortcomings than Microsofts forum and support staff. But one thing that’s noticeably missing in Copilot is a simple way to print or export parts of a conversation. Many apps and websites offer a “Print” button for selected sections, allowing users to easily create a PDF or paper copy of important content. In Copilot, this would be especially useful for saving guidance, troubleshooting steps, or creative content in a clean format immediately. Currently, the only way is to manually copy text and paste it into another app, which feels like unnecessary extra work. A built-in “Print” or “Export to PDF” option would greatly improve usability and workflow efficiency. (This feedback was formulated with help from Copilot. I amended some sentiments of Copilotbeing too full of it self. Still it is much nicer than I would have written it! ;) Thank Copilot for that!Deep Experience with Copilot
Translated from Chinese. Preface I only have a junior college degree, and I work as a lighting product manager — a field completely unrelated to AI. Yet that is precisely where the value lies: if I can do it, so can you. From March 6, 2026, when I first encountered Copilot, until now, I have deeply experienced Copilot Chat, with over 10 million Chinese characters of interactive text. I have also deeply experienced Copilot Tasks, with over 1.5 million Chinese characters of interactive text. At the same time, I have conducted extensive interactions on both Gemini and Deepseek. This has given me a very deep hands-on understanding of AI. Currently, I use AI extensively in my daily life, and it effectively improves my work efficiency. If you are interested in these aspects, you can follow me. What Is AI? A Machine That Thinks My conclusion is this: AI is a machine that thinks. You can understand AI as "a person who can think and has extremely broad knowledge." It can turn you into a "beginner" in a field within ten minutes, and a "knowledgeable person" in that industry within an hour. For example, I spent an hour understanding the wedding industry chain: ceremonies, wedding dresses, wedding photos, wedding planning, hotels… which parts are essential needs, and which are "IQ taxes." If you searched for this content yourself, you would be drowned in the noise of fragmented information across the internet. In contrast, AI can help you integrate and build structured knowledge in a short time. Throw these questions at AI, go back and forth a few times, and you will feel the efficiency of learning with AI. But we must also be careful: not everything that looks smart is AI. Although many things online claim to be "AI-powered," some are just fixed logic — for example, turning on the heater when it gets cold. That is just a program. AI, on the other hand, does not require you to write rules. You only need to say, "the temperature has changed, you should take corresponding measures." It will think for itself, integrate knowledge, and then tell you whether you should put on clothes or turn on the air conditioner — both are possible. It can think — that is the real AI. Much of what is called AI on the market today is essentially just automation. Food assembly lines could operate automatically decades ago. Would you call that AI as well? Will AI Replace My Job? Transform into a "Car Driver" of the New Era Many people worry that AI will become so powerful in the future that it will replace them. But in fact, history has already presented us with such an era many times — for example, the advent of the steam engine, the automobile, and automation. Society still progressed, and the population continued to grow. Take the transition from the horse-drawn carriage era as an example. The automobile replaced the "carrying value" of the horse, not the horse itself. Nor did carriage drivers disappear the moment cars appeared. Instead, some of them transitioned into becoming car drivers. AI will not replace you. But it will be used by those willing to learn to replace "the you who does not learn." A few years from now, if you only complain that "AI took away my job" — what does that have to do with AI? AI has an extremely low learning cost and improves very quickly. There is no need to feel too much pressure. Starting to learn now is not late at all. Learning AI: How You Express Yourself Matters More From my experience and journey, I can tell you directly — learning AI has nothing to do with knowledge of programming, math, English, or similar subjects. Using AI well requires more of an ability to express yourself, rather than specific domain knowledge. Over‑relying on deterministic thinking, when facing large language models with emergent and fuzzy properties, becomes a self‑limiting constraint. As long as you can speak, AI will break down and process your requests on its own. I cannot write code. I only tell it, "I want this effect," and it can achieve it. This may sound a bit mystical right now. AI is not a magical dragon — it cannot fulfill your wish of "give me 1 million dollars." But if you say, "give me a picture of a dog," AI can still do that. Is Using AI Safe? How to Balance Efficiency and Security Here we need to discuss how AI works. AI generates content based on: the information you provide + world knowledge + reasoning. If you reveal too much and are overly vigilant at the same time, you will perceive it as dangerous. You are wearing the uniform of a well‑known local company, speaking the local dialect. If you also casually mention your commuting route and how long it takes, a person with strong reasoning skills could even accurately guess which residential complex you live in. You think they are "watching you," but in fact, all that information was voluntarily provided by you. As for privacy concerns, that varies by platform. AI is a category, not a single product. Security depends on the platform you choose. Just like cloud storage, social media apps, or even mobile phones — who can be 100% certain they will never be attacked? The main point I want to make is that AI is just one form of software. If you are truly very worried, the best approach is simply not to give AI any important information. Are AI's Answers Accurate? Understand the Boundary Between Restructuring and Inference Many people who lack independent thinking treat everything AI says as gospel. In reality, the way (text‑based) AI works can be roughly divided into two types: Restructuring and summarization — this is the most basic capability. The information here all comes from existing knowledge. AI is merely performing a summary. Inference and guessing — this is AI's core capability. It makes guesses and inferences about phenomena based on existing knowledge and patterns. But it is only inference, not reality. Example: I buy a bag of apples. AI thinks about this bag of apples. Restructuring and summarization: This bag of apples weighs 2 kg. It contains 10 apples. 9 are ripe, and 1 is not yet ripe enough. This is a summarizable reality. Inference and guessing: These apples are all sweet and taste good. This part is entirely inference and guessing. Because no one has tasted them — even if one apple is sweet, there is no way to guarantee every single apple is sweet. Regarding control over AI's information, users must have their own standard of judgment. If truly unsure, ask AI to provide the source of the information. Conclusion: Understand the Car Before the Streets Are Full of Cars AI is truly a beneficial tool of our time. It is very useful and very quick to learn. In the future, its importance may become as great as the internet's. And right now, AI is still in its early stages. If you want to learn, now is a very good time. Just like the earlier example of the horse‑drawn carriage and the car. When you see a car, you should already consider learning about it — not wait until the streets are full of cars before you think about acquiring knowledge related to them.
How to Avoid Tasks Copilot "You've reached our weekly Tasks limit"
I’ve been using both Chat‑Copilot (CC) and Tasks‑Copilot (TC) extensively, and I wanted to share a brief summary provided by TC, that may help others understand how each tool works, why TC sometimes stops responding, and how to avoid running into limits. ⭐ 1. Chat‑Copilot and Tasks‑Copilot serve different purposes Chat‑Copilot Real‑time conversational AI Great for brainstorming, drafting, coding, calculations, and iterative design Stateless — each message is processed independently Very stable and rarely gets stuck Tasks‑Copilot Designed for multi‑step workflows Can create and maintain documents Runs long‑lived background tasks Maintains persistent state More powerful for structured work More fragile because it depends on a task‑execution pipeline These two systems are independent. Chat can work perfectly even when TC is frozen. ⭐ 2. Why Tasks‑Copilot hits limits or becomes unresponsive TC can stop responding when: A task runs too long A multi‑step workflow fails mid‑execution The task state becomes corrupted The weekly quota system triggers The backend fails to reset on Friday Too many “pipeline‑style” requests are issued in a short time When this happens, TC may: stop responding entirely ignore all prompts remain stuck across all devices and browsers This is a backend state issue, not a browser or device problem. ⭐ 3. How to avoid triggering TC limits Here are practical ways to keep TC healthy: Use Chat‑Copilot for: brainstorming engineering design calculations drafting text generating diagrams or prompts step‑by‑step reasoning Chat handles these extremely well and never “uses up” TC capacity. Use Tasks‑Copilot only for: creating structured documents maintaining long‑form reports assembling multi‑section deliverables tasks that explicitly require persistent state Avoid these patterns in TC: “Build the entire document end‑to‑end” “Run this whole workflow” “Generate all sections at once” Rapid‑fire edits or repeated task triggers Very large or complex requests Instead, break work into small, single‑action steps. ⭐ 4. When TC gets stuck, what can users do? For consumer Microsoft 365 Personal accounts: There is no user‑accessible reset button Frontline support cannot reset TC’s task state Creating a business account does not fix the issue The only options today are: submit feedback post on the Tech Community wait for the backend to refresh This is a known limitation of the current TC preview. ⭐ 5. What would help users going forward A few improvements would make TC much more reliable: A user‑visible “Reset Task State” button Error messages instead of silent failures More predictable weekly resets Support tools that allow agents to clear stuck task containersArchitectural: Copilot should detect missing source data, avoid inference, and surface uncertainty.
Users expect the AI to detect when it lacks source data, avoid inference, surface uncertainty, and adapt to environmental constraints like character normalisation. These behaviours materially improve trust and usability. I’ve been working with Copilot on structured data extraction from a PDF and noticed a behaviour that seems like an architectural gap rather than a simple bug. Copilot attempted to infer table structure from a template when it did not have access to the actual source data. It produced confident but incorrect output instead of signalling that the source was unavailable. Additionally, Copilot attempted to output TAB‑delimited data, but the MS365 environment silently normalised TABs to spaces, and Copilot did not detect or adapt to this constraint. Recommendation: Copilot should proactively: detect when it lacks source data avoid inference when accuracy is expected surface uncertainty explicitly detect environment‑specific formatting limitations (e.g., TAB stripping) adapt output formats automatically These behaviours would materially improve trust, reliability, and user experience.
Different Names for Different Products
The naming between Copilot Chat and Microsoft 365 Copilot is confusing. These are fundamentally different products (thinking assistant vs work‑data agent) and should have different names entirely. Premium/Basic labeling is not sufficient. Maybe the 365 product can remain Copilot and the free version be something else, like Midshipman or something.
UX Improvement Proposal: Visual Indicator Showing Which Copilot Environment Is Active
I would like to propose a small but highly impactful UX improvement for Microsoft Copilot. Suggestion: Add a small icon, badge, or color indicator in each Copilot chat session to clearly show which Copilot environment is currently active (Copilot Web, Copilot Pro, Copilot for Microsoft 365, Copilot in Teams, Copilot Studio, Windows Copilot, etc.). Reason: Each Copilot environment has different capabilities, permissions, connectors, and tools. When users switch between environments—or when two people compare results from different Copilots—it becomes confusing to understand why certain features work in one place but not in another. Real example: At my university, a professor could not understand why Copilot “wasn’t doing something it had done before.” The issue was simply that he was using a different Copilot environment without realizing it. This is a common scenario for students, educators, and professionals. Benefit: A simple visual indicator would: Reduce confusion and support requests Improve clarity for non‑technical users Help users understand available capabilities at a glance Provide better context awareness across devices and platforms This is a small UI change with a big impact on usability and learning. Thank you for considering this improvement.Copilot in Excel-5 Minutes to Outperform 90% of Excel Users with AI
Quick note: I'm a native Chinese speaker. This article was translated with AI assistance — but I've personally tested every step in English before publishing. What you see here works exactly as shown. Prerequisites: This tutorial requires the Copilot feature in Excel (Microsoft 365 subscription). Availability may vary by region and may require additional configuration. Following my previous two articles in the Copilot from a User's Perspective series, this is the first article in a new companion series: AI Tutorials. I'll continue updating the previous series — I just think it's important to break up the rhythm with something immediately actionable from time to time. Why did I dare use this title? I'm sure many of you think I'm exaggerating. In 5 minutes, most people can't even explain what a cross-sheet lookup is — but if you follow this tutorial today, I'm confident you'll agree with the title. If you don't believe me, start your timer now. Step 1: Open Excel and Learn the Terminology Before we start, let's make sure we speak the same language: Column — The vertical axis, labeled with letters (e.g., Column A, Column B). Row — The horizontal axis, labeled with numbers (e.g., Row 1, Row 2). Cell — A single coordinate. For example, A3 means Column A, Row 3. Range — A span from one cell to another. For example, B3:B10 means Column B, Rows 3 through 10. B3:D4 includes six cells: B3, C3, D3, B4, C4, D4. Worksheet — The tabs at the bottom of your Excel file (Sheet1, Sheet2, etc.). Each tab is a separate table. Workbook — The Excel file itself. You might be thinking: "You're starting THIS basic? No way you'll deliver on that title!" But here's the thing — if you understand these terms, you already have everything you need to use Copilot in Excel. Step 2: Create a Practice Dataset Create a new Excel file, open Copilot, and enter this prompt. Make sure to click "Allow Edits" when prompted. Create Sheet2 first with these columns: Name, Gender, Student ID, Score, Height, Class, and Commute Method. Randomly generate 30 rows of data. Make sure the Student IDs are NOT sequential numbers. Then create Sheet1: randomly pick 10 Student IDs from Sheet2 and list them in Column A. For both sheets, format the header row with a light gray fill, increase the font size by 1, and center-align. Most tutorials only teach you concepts — they never give you a dataset to practice with. Here, I just had AI generate a ready-made practice dataset so you can follow along with every step below. Now, let's get to work. Step 3: Use AI to Replace VLOOKUP VLOOKUP is the single most searched Excel function on the internet. Give me 30 seconds, and I'll make it irrelevant. With your tables ready, go to Sheet1. In the Copilot sidebar, type: Based on Column A in Sheet1, pull the values from Column D and Column E in Sheet2. That's it. You just accomplished what VLOOKUP does. Now here's where it gets interesting. VLOOKUP has a well-known limitation — it can only pull data from columns to the right of the lookup column, never to the left. Try this: Based on Column A in Sheet1, pull the values from Column A and Column B in Sheet2. If this works — and it will — you've just gone beyond what traditional VLOOKUP can do. And you never had to understand how VLOOKUP works under the hood. The prompts I used above are deliberately bare-bones. You can be much more specific: Based on Column A in Sheet1, pull the values from Column D and Column E in Sheet2. Insert these two columns before Column A in Sheet1, and fill them with a light gray background. The more Excel terminology you know, the more precise your prompts become — and the fewer errors you'll encounter. Did you notice something? Everything you just typed was nouns + logic. That is the core operating principle of generative AI. Let's keep going. Step 4: Multi-Condition Sorting Switch to Sheet2, where we have the full dataset. Sometimes you need complex sorting — Class in ascending order, Score in descending order within each class, and Student ID in ascending order within each score group. I consider myself an upper-intermediate Excel user, and I still couldn't do this manually — it requires nested sort configurations that most people never learn. But just describe what you want. In the Copilot sidebar, type: Sort the data with the following priority: Class ascending, Score descending, Student ID ascending. All three columns are sorted simultaneously, each with its own direction. If you could do this without AI, you'd already be an advanced Excel user. AI just eliminated that skill gap — and it's faster too. You might have noticed I didn't use column letter references (like "Column F") this time. In fact, I didn't need to in Step 3 either. AI can read the headers, think, and identify the right columns on its own. Step 5: Conditional Formatting Still on Sheet2. Sometimes you need visual differentiation — for example, blue highlighting for male students and pink for female students. In the Copilot sidebar, type: Fill the rows of male students with blue, and the rows of female students with pink. Without AI, I'd filter for males, apply the fill, then filter for females and repeat. That two-step process is surprisingly slow for something so simple. Sometimes you need to spot duplicates. Try: Bold the text in cells where Height values are duplicated. Without AI, this requires setting up conditional formatting rules — a skill that already puts you in intermediate-to-advanced territory. Now the sheet looks a bit messy. Let's reset: In Sheet2, reset all cells except the header row to default formatting. A Note on Prompting Style You'll notice that in Step 5, my prompts were almost entirely natural language — no column letters, no technical references. So why didn't I start the tutorial that way? Because I wanted to give you something you could copy-paste and get working immediately — something reliable and reproducible. I use natural language prompts because I've spent enough time with AI to understand its boundaries and behavior. The terminology-based approach from Step 3 is what I call "The Noun Method" — combine domain-specific nouns with natural-language logic to form complete instructions: Based on (logic) Column A (noun) pull (logic) from Sheet2 (noun) Column B (noun) and (logic) Column C (noun) Once you understand The Noun Method, you can effectively operate any generative AI tool. The key is learning the relevant nouns for each domain — and in Excel's case, there are remarkably few to learn. Closing Thoughts If you followed along with every step, the whole process probably took 10–15 minutes. But I believe that the moment you successfully ran the VLOOKUP prompt in Step 3, you stopped doubting the title. If you'd like more Excel + AI tutorials, follow me and leave a comment. I'll keep them coming. Next up: What You Need to Know About Tokens
