Microsoft Copilot usage limits: How to be efficient in your tasks

Nov 11, 2026
Copilot Chat usage limits, with two documents showing a checked change and the Copilot logo

Your prompting is poor.
You’re hitting Copilot Chat limits.
Here’s how to finish tasks with fewer requests.

If you use the Copilot Chat included with your Microsoft 365 work account, this guide is for you. Whether you work in HR, Procurement, Sales, Legal, Operations or Marketing, the aim is to finish everyday tasks faster and check each result against the same standard.

In a nutshell

  • Only stay in the same chat if your next task needs something from it. Start a new chat for a separate task so Copilot doesn’t pick up details from the previous one.
  • Give Copilot the details it needs in your first request. Include facts you already know and say what the answer must include. That saves you from adding them later.
  • Give Copilot only the text or files it needs. If you’ve already attached the right file in the same chat, tell it to use that file and save yourself another upload.
  • List every step you already know you need in your first request. Tell Copilot where to stop if you need to check its answer before it continues.
  • If you know the exact change and can make it in seconds, make it yourself. That saves a request to Copilot.
  • Keep each chat focused. If Copilot keeps picking up old details, start a new chat with the latest version and the facts it still needs.
  • Finish the task in as few steps as you need. Ask for steps you can check together in one request. Keep a separate check where a mistake would mean doing the work again.


What Microsoft says about Copilot Chat usage limits

Microsoft began adding limits to more Copilot Chat features in November 2026. The rollout started on 2 November for organisations with up to 2,000 Microsoft 365 licences and on 9 November for larger organisations. Microsoft says access can change during the day and does not publish the numbers. 

These limits apply to people using Copilot Chat with a work account who don’t have a paid Microsoft 365 Copilot licence.

At Koshima, we looked for a public timetable showing when Copilot Chat is less busy in different countries. We couldn’t find one. You can control how many times you ask Copilot to finish the same task. 


Use fewer requests (prompts) on your tasks

When you ask Copilot to do something, Microsoft calls that a prompt. At Koshima, we call it a request. Use the same words you’d use to ask a colleague to do the task.

1. Hard rule. Only stay in the current chat if your next task needs information from it

You’ve asked Copilot to compare two documents. Now you need an email about what changed. Stay in the same chat and ask it to write the email using the comparison it just made. You can skip asking for the comparison again. When you have a different pair of documents, open a new chat and attach the new pair. That keeps the old files out of the new chat and cuts the chance you’ll need to correct an answer based on the wrong files. 

2. Hard rule. Put all the details Copilot needs in your first request

Say what you want back and give Copilot the facts it needs. If you ask it to draft a meeting email, include the date, the reason for the meeting and what you want people to do before it. Give those details now and you can avoid follow-up requests to add them. 

3. Hard rule. Attach only the files needed for the task

If you want Copilot to rewrite one paragraph, paste the paragraph. If you want it to check whether that paragraph agrees with the rest of the document, give it the whole document. If the right file is already in the chat, tell Copilot to use it by name. You can avoid another upload. 

4. Hard rule. Set out everything you want Copilot to do in the first request

If you need a summary of some notes and a list of actions, ask for both in the first request. Tell Copilot to summarise the notes first and list the actions underneath. You can get both in one answer instead of coming back to ask for the list.

5. Hard rule. Use the right tool for the job

You’re reading a draft and see a date you already know is wrong. Change it directly. If you ask Copilot to fix one date, you send another request and may need to read the draft again to see what else changed.


Tell Copilot when to stop and keep chats focused

6. Tell Copilot where to stop and wait for your approval

If an email depends on a comparison, check the comparison first. Say Copilot missed a change in one of the documents. You can fix the comparison before Copilot writes an email based on it. That saves you from asking it to rewrite the email too.

“Compare these two documents and show me what changed in a table. Give me the page number or section for each change. Stop after the table and wait for me to check it. Once I approve it, write a short email explaining the changes I confirm.”

If all you need is a short email and you’ve already given Copilot the facts, ask for the finished draft straight away. Asking it to confirm the plan first would add a request.

7. Start a fresh chat if old details keep getting in the way

Say you’ve made several drafts in one chat and Copilot keeps pulling in a date you changed earlier. Start a new chat with the latest draft and the correct date. You leave the old drafts behind and can avoid asking Copilot to fix the same date again.

Earlier messages can add to the text Copilot has to process. AI tools count text in small pieces called tokens. A token might be a word or part of a word. Microsoft hasn’t published how much of a chat Copilot reads for each answer or how those tokens count towards its limits. 


Before you start, decide how many steps you need

Ask yourself what you need at the end and which parts you need to check along the way. If you can check two results together, ask for them together. If one result could change what you ask for next, check it first. You can finish in fewer requests and still check the parts that matter.


What to do in ChatGPT, ChatGPT Work and Cowork


ChatGPT

If you want ChatGPT to rewrite two paragraphs, paste those paragraphs and say what you want changed. That can save you uploading the whole file. OpenAI has separate limits for some models and tools, including file uploads. Where ChatGPT counts messages, a shorter answer still uses one message. 

ChatGPT Work

Tell ChatGPT Work which files to read and what you need back. If you want a one-page summary of a document, give it the document and ask for one page. OpenAI says more text, more steps and settings that tell the AI to spend longer working through the task can use more of your allowance. Check Settings → Usage before a large task. 

Microsoft Copilot Cowork

If all you need is a short email draft, ask for it in Copilot Chat. When you give Cowork a task, name the files to use and say what you want it to produce. Microsoft says tasks with more steps generally use more Copilot Credits. If your organisation uses credits, type /cost to see how many credits the open task has used so far. 

Five questions before you send a request

  1. Does this task need anything from the current chat?
  2. Have I told Copilot what I want back and given it the facts it needs?
  3. Does Copilot need the whole file, or can I paste the part it needs?
  4. Can Copilot do the steps together, or do I need it to stop for a check?
  5. Can I make this change myself in the time it takes to ask?

 

Help your team work effectively with Copilot and ChatGPT

We offer Microsoft Copilot training and ChatGPT training face-to-face and online. Your team practises on tasks they already do, such as comparing documents, and learns what to give the tool and what to check. We cover free and paid plans, plus Microsoft Copilot Cowork and ChatGPT Work.

Our enablement programmes give your team time to learn, practise on real tasks between sessions and return to learn and practise again. They learn to choose the right tool and avoid wasting requests or credits. Every programme ends with an AI clinic where we review progress, work through challenges together and guide them through ideas they want to try.

Discuss training for your team

AI consulting