How to Build a Second Brain in Copilot

To build a second brain in Microsoft Copilot, organize your source documents, supply the relevant material to a conversation, and save checked findings back into a durable notes system. Treat Copilot as the assistant working with that collection rather than assuming every conversation has access to everything you have saved.

fenbs can keep the reviewed project decisions and progress you want supported assistants to share. Microsoft Copilot is different from GitHub Copilot: fenbs has a GitHub Copilot connection, but does not connect to Microsoft Copilot. For Microsoft Copilot, transfer selected, permitted context yourself.

What are the steps for a useful first setup?

1. Pick one recurring goal, such as preparing a community workshop. Write its audience, intended result, and the questions you need to answer.
2. Create a source collection in your existing document or notes system. Give the current outline, references, and working notes recognizable names.
3. Make an index identifying the current files, their purpose, and the questions they can answer.
4. Provide only the relevant files or permitted text to Copilot. Ask it to identify the evidence it used and flag missing information.
5. Check the result, save a reviewed note, and write the next step before leaving the conversation.

Microsoft documents file uploads in Copilot↗ (opens in a new tab), including how to attach a file and ask follow-up questions. Use that page for current supported formats and upload limits. The organizing system below does not depend on a particular allowance.

Keep your first trial small enough to inspect. An approved workshop outline and two relevant references are a better test than an entire document archive. You can expand once Copilot consistently identifies the correct source and you have a routine for reviewing its answers.

Which Copilot experience are you using?

Check the account and app before choosing a workflow. A personal Copilot conversation and Microsoft 365 Copilot Chat can expose different source controls. Use the instructions for your experience and ask your organization which material you may provide.

For Microsoft 365, Microsoft explains how to add content to Copilot Chat prompts↗ (opens in a new tab), including uploaded files and available work sources. Follow the source picker your account actually offers. Do not assume a file somewhere in your organization is already included in an answer.

A practical opening request is: “List the documents you used for this answer, and say what you could not access.” Compare that list with the material you intended to supply. If a source is missing, resolve access or provide an allowed excerpt before asking for a conclusion that depends on it.

Our Copilot and ChatGPT comparison covers the broader product choice. Here the aim is narrower: establish a repeatable process for using your documents in the Microsoft Copilot experience you already have.

How should you organize documents before asking questions?

Separate original sources, your own understanding, and the current project state. Each serves a different purpose. A reference supports a claim; a working note explains your interpretation; a current-state record tells you what action follows.

Use names that distinguish source and version without opening every file. In a hypothetical workshop project, “Current audience brief” and “Prior audience brief, replaced” are clearer than two files called “Brief final.” Write the actual review date inside the record when it affects whether the information is still useful.

A notes tool can hold the full material while fenbs holds the active work. Its AI context workflow lets a supported connected assistant read the rules, current progress, and earlier notes before starting. That record still needs human review and source access.

What should you ask Copilot to produce?

Ask for an answer you can check in one pass. Specify the question, the supplied evidence, and the output you need. When comparing alternatives, require separate evidence for each rather than accepting a smooth combined summary.

Use the attached current workshop outline and reference notes.
Question: Which examples fit an introductory audience?
For each recommendation, name the source and section.
Separate source statements from your suggestions.
List any missing information that could change the recommendation.
Do not treat a proposed change as approved.
Finish with one next review action.

Keep your own question visible during review. If the task is choosing examples for beginners, a long overview of presentation design may be interesting without answering it. Ask for a revision that addresses the decision you actually need to make.

For reusable knowledge, ask a second question after the result is checked: “What short note should I save for the next session?” Review that note too. It should preserve the source, the conclusion, and its limits without importing speculative claims from the conversation.

How do you check an answer before saving it?

Open the source at the cited location. Check whether it supports the exact statement and whether a nearby qualification changes the meaning. A reference title alone is not enough: the content has to justify the conclusion.

Then check the project state. A recommendation is not an approval, and a draft is not a completed task. In the workshop example, you might save that two examples fit the outline while organizer approval remains outstanding. That is a useful handoff even though the work is unfinished.

Keep a visible distinction between reviewed and unreviewed notes. The AI second brain guide explains the larger knowledge workflow. Your immediate goal is to prevent a plausible answer from quietly becoming the authority for the next answer.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add your continuing project.
2. Add the reviewed source links and brief context notes. Keep full documents in their existing home and provide their contents separately when needed.
3. Record choices a person has approved in Decisions and rules. Create tasks for checks that still need doing.
4. For Microsoft Copilot, copy only the permitted project context into your conversation and bring reviewed updates back yourself. For a supported assistant, follow the integration guide to connect it over MCP, a standard outside-tool connection.
5. Save completed checks, unresolved questions, and the next action in Where we left off before stopping.

fenbs does not import Microsoft Copilot conversations or connect directly to Microsoft Copilot. If you want everything to stay inside your Microsoft workspace, use your existing notes and document processes. fenbs fits when a separate work record helps you continue with other supported assistants.

What should you carry into the next session?

Bring the current question, checked checkpoint, and the sources needed for that question. Ask the assistant to explain what is approved, what is unresolved, and where it will begin. Correct discrepancies before it creates another layer of work.

The useful measure is whether you can resume without rereading every chat. If you instead want to organize the underlying notebook, follow the OneNote workflow. For more approaches beyond this setup, browse the second-brain guides.