How it works

Put it in copy. It writes the knowledge base.

Add one email address to the conversations you already have. The agent reads them, writes the project knowledge base on SharePoint, inside your Microsoft 365 — and tells you when two things you wrote contradict each other.

One address, multiple places it listens

The agent has a company email like anyone else. Copy it on a thread, invite it to a meeting, add it to a Teams group, share a document with it — wherever you bring it in, what gets said there becomes a page in the knowledge base, on your SharePoint.

Scattered

“go-live is 15 September”
“migration test passed, but only on staging”
“we need a second run on production”
“clarification window closes 11 days after”
“we need the ACME data by August”

The agent

Structured

Project Ariel

Decisions /

Project timeline

linked

Go-live set for 15 September, agreed at kickoff.

Dependency: ACME data by August.

Decisions /

Test results

Migration test passed on staging.

A second run on production is required before go-live.

Decisions /

Requirements

linked

Export and user permissions, from the kickoff thread.

Decisions /

Deadlines

Clarification window: 11 days from publication.

Decisions /

Open decisions

Which go-live date holds — waiting on the team.

Everything already said in Outlook, Teams and SharePoint, turned into a knowledge base that links itself.

The agent has three jobs

Two it does on its own — one while you work, one when nobody asked. The third is yours: you decide the rules it follows.

It writes down what is said

Emails, meetings, chats and documents become knowledge base pages on SharePoint, written the way you asked for them.

It speaks up when things stop adding up

When two pieces of content contradict each other, it posts on Teams and asks which one holds. Nobody had to go looking.

You change the rules yourself

How it writes and what it watches live in a plain-language file you edit. No ticket to us, no consultant, no waiting.

While you work

It writes down what is said

You do not write anything, and you do not move anything. You bring it into the conversations you were having anyway.

New message
ToMarco Realli; Giulia Ferri
Ccagent@yourcompany.comagent@yourcompany.com

Company directory

ABAnna Bianchi
MRMarco Realli
SecondBrain 365agent

Recap of the kickoff: go-live is 15 September…

Send
No new tool, no migration. One address in Cc, and it is in the loop from then on.

Email

Put it in CC: it reads the whole thread, not just your message.

Meeting

Invite it: if the meeting is transcribed, it pulls the content out of it.

Chat

Add it to the Teams group and it follows the conversation.

Documents

Share them in the workspace and they become part of the knowledge base.

You decide when to bring it in — exactly as you would with a colleague.

When nobody asked

It speaks up when things stop adding up

This is the part no colleague actually does: re-reading what was decided months ago, and noticing that today contradicts it.

February says one thing. May says another.

The team agrees the go-live in February. In May somebody mentions a different date, in a meeting nobody connects to the first one. The agent notices and posts an alert on Teams — it is up to you to decide which one holds.

It is the check most knowledge bases never get, which is why most internal knowledge bases go out of date.

Project Ariel
A

Anna Bianchi · 12 Feb · Project kickoff

Kickoff done. Go-live confirmed for 15 September.

M

Marco Realli · 9 May · Progress meeting

We'll slip to November anyway, let's not stress the team.

{·}

SecondBrain 365Agent

Inconsistency on the go-live date

The timeline says 15 September. Today's meeting mentioned November. Which one holds?

Nobody was re-reading February's minutes. The agent was — and it asks the team instead of deciding for them.

Inconsistencies

What you just saw: both sources attached, and the decision left to you.

Tasks

The activities mentioned in a meeting end up in a list, with who owns them.

Deadlines

Reminders as they approach — and a check that they were actually met.

Whenever you want

And you change the rules yourself

How it writes and what it watches are not settings we configure for you: they are a page of plain language inside the workspace. A software project and a product team want different pages out of the same conversations — what changes is not the software, it is the file.

Two workspaces, two sets of rules:
context.mdSoftware project

Keep one folder for meeting recaps, written as decisions and next steps.

Keep user requirements separate from test results.

Maintain a timeline page, and flag it when a date changes.

Plain language. Rewrite it whenever the project changes.

Knowledge baseon SharePoint
  • Meeting recaps

    • Kickoff — 12 Feb
    • Progress — 9 May
  • Requirements

    • Export — user requirements
  • Test results

    • Migration test — staging
  • Timeline

    • Project timeline

Same agent, same sources. The structure changes because the rules changed — and you rewrite them the day the project changes, not by filing a ticket.

No ticket, no consultant

You do not open a request with us and wait for a quote. You edit the file in SharePoint, like any other page, and the next thing it writes follows the new rules.

Plain language, not configuration

“One folder for meeting recaps, one for requirements.” There is no schema to learn and nothing to redeploy — if you can brief a new colleague, you can write this file.

One agent, as many use cases as you need

The same agent runs a software project, an HR knowledge base or a tender response. The file is what tells it which job it is doing here.

The workspace

A Teams group, a SharePoint site, an agent

When you create a workspace you get three things together. You already know the first two: they are the tools your team uses every day.

A workspace is three things. Two of them you already have.

Where your people already collaborate.

  • The channels the team already uses
  • The people you decide belong to this workspace
  • The files they already share there
  • The agent, added as one more member

Nothing moves and nothing changes. Your team keeps working where it works.

Where documents and the knowledge base live.

The only new thing: an identity and an email address.

Two of the three parts are the ones you already pay Microsoft for. The third is an email address. That is what “nothing to integrate” means in practice.

The KB as a source

And then you talk to it. From Copilot too

The knowledge base is not an archive you visit: it is the place your questions get answered, whether you ask the agent directly or ask Microsoft 365 Copilot.

EmailCC it inMeetinginvite itChatadd it to the groupDocumentsshare themThe agentKnowledge baseon SharePointInconsistencyalert on Teams
From Copilot
The content is in SharePoint, so Microsoft 365 Copilot sees it like anything else in your tenant — you are already paying for it. Copilot does the answering and the agent the writing: here is why an AI knowledge base needs both halves.
From the agent
Ask it directly and it answers on the project knowledge base, citing the page it took the answer from.
From the dashboard
Progress and generated documents, built from the knowledge base itself.

Viva Topics used to do a narrower version of this — its AI compiled topic pages, but only from what was already in SharePoint, never from email or meetings, and Microsoft has retired it. What replaces Viva Topics on Microsoft 365 goes through the comparison row by row.

Nothing gets in without someone saying yes

The approval flow is optional, but it is there: the project manager — or the team — validates what the agent proposes before it enters the knowledge base. And everything stays editable by hand, like any SharePoint page.

The agent proposes. People decide. That order never changes.

{·}THE AGENTproposes a pageApproveDiscardA PERSONreviews itKNOWLEDGE BASEthe page goes instill editable by hand

Security and governance

Your data doesn't leave. Ever.

SecondBrain 365 lives inside your Microsoft 365 tenant. There is no “our server” your conversations travel through: AI meeting assistants process your conversations on their own infrastructure, we have nowhere to put them.

Identity
Where it usually livesAn account in the vendor's console
SecondBrain 365Your Entra ID, like any employee
Access policies
Where it usually livesWhatever the vendor supports
SecondBrain 365Your Conditional Access, unchanged
Compliance and audit
Where it usually livesTheir dashboard, if they have one
SecondBrain 365Purview, where you already look
Classification
Where it usually livesReinvented per tool
SecondBrain 365Sensitivity labels inherited from SharePoint

FAQ

The questions your IT department asks

In your Microsoft 365 tenant, in the region you already chose for it. The knowledge base is a SharePoint site you own: if you close your account with us tomorrow, the content stays where it is and remains yours.

No. Your content is used to answer your questions and to write your knowledge base, not to train models. It is not pooled with other customers' material.

What is shared with it inside a workspace: the email threads it is copied on, the meetings it is invited to, the Teams group it belongs to, the documents placed there. It has no standing permission to search the rest of your tenant.

You disable the agent's identity in Entra ID, like any other account. The knowledge base it produced stays in SharePoint and remains readable and editable — nothing of yours is held hostage.

The agent operates as an identity in your tenant, so its activity appears in your Microsoft 365 audit logs alongside everyone else's. You inspect it with Purview, in the place you already look.

Let's see it on your next project