Recent Discussions
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!The Difference Between Declarative Agents and Autonomous Agents in Copilot Studio
Artificial intelligence is rapidly transforming the way businesses automate tasks, improve customer experiences, and increase productivity. One of the most exciting platforms leading this transformation is Microsoft’s Copilot Studio, a low-code environment that enables organizations to create intelligent AI-powered assistants and workflows. https://dellenny.com/the-difference-between-declarative-agents-and-autonomous-agents-in-copilot-studio/Topics and Nodes in Microsoft Copilot Studio: The Foundation of Intelligent Conversations
Artificial intelligence is changing the way businesses communicate with customers and employees. From automating support requests to handling internal workflows, conversational AI tools are becoming an essential part of modern digital transformation. One platform leading this evolution is Microsoft Copilot Studio. https://dellenny.com/topics-and-nodes-in-microsoft-copilot-studio-the-foundation-of-intelligent-conversations/Enterprise AI Agent Architecture Patterns Using Copilot Studio
Artificial Intelligence is no longer limited to chatbots answering customer queries or virtual assistants scheduling meetings. Enterprises are now building intelligent AI agents capable of automating workflows, orchestrating business processes, integrating enterprise systems, and assisting employees in real-time. As organizations scale their AI adoption, architecture becomes the defining factor between isolated experiments and enterprise-wide transformation. https://dellenny.com/enterprise-ai-agent-architecture-patterns-using-copilot-studio/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 are you connecting OKR tracking to Copilot workflows?
My manager asked me to figure out how we can use Copilot to help with our OKR process. Right now we track objectives in a SharePoint list and its a mess. People forget to update progress, key results are disconnected from daily work, and nobody looks at it until the end of quarter scramble. Has anyone found a way to bring OKR tracking closer to where people actually work in Teams/Copilot? Or is there a Teams app that handles this better than a spreadsheet?
Reached weekly tasks limit
Friday at around 2pm I got locked out of using my projects under tasks in copilot. It says "You've reached your weekly Tasks limit. Check back Friday for more Tasks." I have 365 premium. After spending a day trying everything copilot could think to try, I gave up and talked to Microsoft support. They told me, after hours of being transferred and explaining what happened over and over, to try here. Copilot thought it was a corrupted webview2 and support thought it was just the back end needed to catch up. I'm a construction worker who does not know what he is doing. Please help.
How to show Copilot custom agent description for every new conversation?
We have a subscription to Copilot 365 (I work with Copilot Studio but this question is about Copilot 365 agents). I use the Copilot 365 app, I don't use it via the web. When I make a custom agent, and do "Try it" it shows the Description I entered in the config screen for every new conversation. This Description is important to show the user what info the agent has and does not have. When I do "Go to agent" which would be more like seeing the agent in production, the Description is never shown. I looked in Settings and Config for the agent and don't see any settings for this. I'm not an admin for Copilot. Is this an admin setting that must be changed? How can I show the Description each time a New Conversation is clicked? Thank you!Copilot in Edge needs direct export integration
While using Copilot in Microsoft Edge, I noticed a key limitation: there is no option to directly save or export text snippets or summaries to OneDrive, Word, or OneNote. Users must copy manually, which breaks the reading flow and reduces productivity. Competitor comparison: Claude → PDF export, integration with Notion and Drive ChatGPT → PDF/DOCX export, integration with Google Drive Gemini → Automatic export, integration with Google Drive Copilot (Edge) → Only copy/paste or browser PDF, no native integration10 Real-World Copilot Studio Use Cases That Save Teams Hundreds of Hours
Artificial intelligence is no longer a future concept reserved for enterprise tech giants. Businesses of all sizes are now using AI copilots to automate repetitive work, improve customer experiences, and help employees become more productive. One platform that is quickly gaining attention in this space is Microsoft Copilot Studio. Organizations across industries are exploring practical copilot studio use cases to reduce manual work and streamline operations. From customer support automation to internal HR assistance, Copilot Studio allows companies to build intelligent AI agents without requiring advanced coding skills. https://dellenny.com/10-real-world-copilot-studio-use-cases-that-save-teams-hundreds-of-hours/Copilot Cowork should not be prompted like Copilot Chat...
I keep seeing the same pattern with clients: they talk to Copilot Cowork the same way they talk to Copilot Chat. They write short, vague, immediate requests. The result is predictable: they get an answer, not real work executed across Microsoft 365... With Cowork, the approach has to change. You should not ask for quick help. You should describe the result you want produced. Cowork is designed to handle multi-step work, create deliverables, act across Microsoft 365, and ask for approval before sensitive actions. The right habits are simple: start with the deliverable you expect: a document, an email, a meeting, a Teams message give the relevant context: project, timeframe, people, and sources to use; define the output clearly: format, recipient, destination; add constraints: tone, length, language, deadline. In practical terms, do not say: “prepare my meeting.” Say: “analyze the last 15 days of emails and Teams messages, create a one-page Word summary with the risks and decisions to make, then send it to the participants.” That is the real difference: with Copilot Chat, you get an answer. With Copilot Cowork, you need to ask for an outcome.
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 containers
Best practices using CoPilot with Data from NetSuite and using NetSuite AI reports
At a small business that has access to readonly reports from NetSuite for retail inventory etc. Is there an existing MCP server that is available that can be easily integrated (no-code). Alternately, within CoWork, is there a plug-in that has skills and connectors for NetSuite that I can potentially search for and re-use through community or third party licenses.Using AI to prep managers before performance reviews. What are you doing?
We're a 50 person company and review season is coming up. Our managers keep saying they dont have enough context on their direct reports when its time to write reviews. Someone suggested using Copilot to summarize Teams chats and emails but that feels like it would miss the actual performance data like goals and feedback. Whats everyone else doing to help managers prep for reviews? Is there some AI-powered tool that pulls together the actual relevant stuff?Errors in source file retrieval from the knowledge base in copilot agents?
Hi all, I understand that copilot agents are connected to MS Graph, which maps the relationships between all the data stored in your MS 365 tenancy (sharepoint, onedrive files, emails etc). Recently, I created an agent and assigned a specific folder to the knowledge base and turned off the "use web content" toggle, because I wanted the responses to be very directly tailored to my folder (inclu. sub-folders with multiple files). I then tested if/how well the agent retrieved specific files using this prompt: "Can you please tell me how many files are in this folder and list the files in the folder? [Insert link to sub-folder in from the main folder in the knowledge base]" The agent responded with (1) an incorrect count and (2) listed a few files that were not in the sub-folder but in another part of the knowledge base. As I understand it, it is a counting error in (1) and retrieval+indexing error in (2). I'm more concerned about (2) because I'm worried the agent isn't retrieving (and therefore, using the info in) all the files in an important folder (when specifically linked to it even). Questions: (a) Where is this error happening in the indexing process within MS graph? Am I misunderstanding where the error lies? Any ideas on why an agent is naming the wrong files in a folder within its own knowledge base?? (b) Do agents created within the copilot agents web interface use Azure AI Search for semantic indexing or is that only for more custom RAG solutions created "from scratch" using foundry, SDK, etc? Do copilot agents use Microsoft Search to query and index files used in a response? Thanks!
Location of custom skills - cannot find them in OneDrive
I'm using Copilot Cowork to create skills, and Cowork insists the skills are being saved in OneDrive>Documents>Cowork>skills but I cannot find them in there. I cannot find the raw MD files anywhere in OneDrive, and Cowork can only surface them when I ask to see them. Anyone else experienced this and have the answer? I need to be able to create Cowork Skills and store them in a central repository for the team to use. #coworkHow to use agents response as an input to a topic?
Hi I would like to use the agents reponse to store it in a word or excel file. For that, I would like to use the response provided by the agent, pass it on to my custom topic which triggers an agentic workflow to store the agents response in a word or excel file. I cannot sort out how to pick the agents response and pass it on to my custom topic, any help will be highly appreciated, thanks!
Events
Microsoft 365 Copilot in Excel- transform your spreadsheet workflows - a live session exploring the latest innovations and powerful new capabilities in Excel.
Through Microsoft 365 Copilot and Exce...
Online
Recent Blogs
- Welcome to the May 2026 edition of What's New in Microsoft 365 Copilot! Every month, we highlight new features and enhancements to keep Microsoft 365 admins up to date with Copilot features that...May 29, 2026
- Expanding model choice in Microsoft 365 Copilot: Anthropic's Claude Opus 4.8 is rolling out today in Copilot Cowork.May 28, 2026
