Copilot Memory: What to Check and What to Record Yourself

Copilot memory is Microsoft Copilot’s ability to retain useful details and preferences for more personalized responses. Check the controls in your actual account, because the consumer app and work experiences do not always show the same options. For project rules that must reach other supported assistants, fenbs provides a separate record; fenbs does not connect to consumer Microsoft Copilot or automatically copy its memories.

A remembered preference can save time, but it does not tell a colleague whether a task is finished or a proposal was approved. Treat personalization and project continuity as two practical needs. You can use Copilot’s controls for the first while deliberately maintaining an explicit work record for the second.

What does Copilot memory help with?

Microsoft describes memory as a way to remember details such as communication preferences, interests, goals, and recurring tasks. Its personalization explanation↗ (opens in a new tab) also distinguishes saved memories, information inferred from earlier chats, custom instructions, and eligible work insights. Which sources are relevant depends on the Copilot experience and account.

Think of a preference such as keeping routine updates concise. If that preference is available to Copilot, a later draft may need less correction. That is different from a requirement that a specific client must approve a particular document before publication. The approval needs a clear project, scope, and person responsible for the choice.

The general concept is covered in what assistant memory means. Here, the useful question is narrower: what has Microsoft Copilot saved or inferred, can you review it, and where will you put work information you need to inspect directly?

How do you check the memory controls in your account?

1. Confirm whether you are using the Copilot app with a personal Microsoft account or a work or school experience.
2. Open Settings and look under Personalization in the current consumer app. Review the memory controls actually shown rather than relying on an old screenshot.
3. If Saved memories is available, open Manage to inspect the saved entries and decide whether they are accurate and useful.
4. Review the other personalization controls separately. A memory choice is not a substitute for checking the settings for other uses of your information.
5. Before changing anything, note the outcome you want: correcting one fact, removing saved details, or stopping future memory use.

Microsoft’s current privacy controls for individuals↗ (opens in a new tab) describe Saved memories and Manage under Settings, Personalization. That page applies to the current Copilot app with a personal Microsoft account and points to separate guidance for other account types. Use it as the source for the labels on your screen.

If the screen differs, pause before choosing a vaguely similar switch. Write down the app, account type, and control name. That small check prevents you from changing a general privacy setting when you meant to correct one remembered preference.

How are saved memories different from conversation history?

A conversation is the fuller exchange: your request, the response, revisions, and follow-up questions. A saved memory is a selected detail used for personalization. A project record is something you maintain intentionally so the next person or assistant can see what is currently true about the work.

Say you explore several tones for a customer email. A preference for short messages could be helpful memory. The instruction to wait for a manager’s approval belongs with the actual email task. Keeping these apart prevents a general writing preference from hiding a specific responsibility.

The Copilot and ChatGPT comparison covers the broader product choice. These are also separate from GitHub Copilot’s app, a development product; a similar brand name does not make its session history or memory controls interchangeable.

What changes when you turn memory off or remove a detail?

Choose the control that matches your intended result. Microsoft says turning off Saved memories in the current consumer app does not automatically delete memories already saved; Manage provides deletion options. A switch that stops new saving and an action that removes existing entries serve different purposes.

Other Copilot experiences may also use information inferred from chat history. Microsoft’s personalization guidance explains separate controls for saved memories and those inferences. Follow the instructions for your account before assuming that removing a single saved entry removes every source of the same information.

After a change, review the displayed entries again. If you ask a follow-up question to check the result, remember that details you supply in that question are available in the current conversation. A test that repeats the old information cannot show whether the assistant recalled it independently.

This article explains the controls rather than diagnosing a failed recall. For that separate problem, use Copilot memory troubleshooting; development-specific behavior belongs in GitHub Copilot memory.

Which project details should you record yourself?

Record information that changes another person’s next action. Useful project context states an agreed fact, its scope, and what follows from it. It should remain understandable without requiring someone to infer your intentions from several personal conversations.

• Decisions: the option a person chose, the reason, and the project it affects.
• Rules: an agreed instruction that continues to apply, with a clear boundary.
• Progress: what changed, what was checked, and what remains unfinished.
• Open questions: what needs a person’s answer before work continues.
• Sources: the current approved document or other reference needed to check the record.

For example, a studio might record that the client approved the shorter registration form, the delivery test passed, and the remaining task is an accessibility review. Those are inspectable work facts. A memory that you often design forms does not convey them.

Keep the record concise enough to review. Remove unnecessary personal details from the summary and follow the policies that apply to the original material. The useful result is a reliable starting point, not a duplicate of every interaction.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add one current project.
2. Manually record the approved work outcome in a task. Add its source reference, what you verified, and the remaining action.
3. Put a person’s continuing instructions in Decisions and rules. Leave unanswered choices open instead of presenting them as approved.
4. Connect a supported assistant using Integrations. MCP is a standard way for an AI app to use an outside tool, with a browser sign-in for fenbs.
5. Ask the assistant to read the project context first and save Where we left off before stopping. Review that recap when the work changes.

fenbs is not the right fit if you want to manage Copilot’s own memory or need an automatic consumer Copilot connection. It does not connect to Microsoft Teams. GitHub Copilot is a separate supported assistant, alongside apps such as Claude and ChatGPT.

How can you tell whether your setup is useful?

Test a harmless preference separately from an important project handoff. For the preference, check whether the response needs less editing. For the project, ask the next supported assistant to identify the approved choice, the unfinished check, and the first action from the saved record.

If the handoff misses something, improve the record rather than assuming more personal memory will fix it. The AI memory hub explains the wider distinction. A successful setup reduces repetition while keeping the important work visible enough for a person to correct.