A Second Brain for Business: What It Is and How to Set One Up
One private place your team can ask what you quoted or promised. What a business second brain is, what goes in first and how to keep it private.
On this page
A second brain for business is one private, searchable store of what your business knows. It pulls in your email, calendar, meeting notes, documents and CRM notes on its own, and keeps them current. Staff and AI assistants can then ask it plain questions, like "what did we quote them last time?", and get an answer with a link to where it came from.
In a business with 5 to 50 staff, that knowledge usually sits in the owner's head, old email threads, meeting notes and a shared drive nobody trusts. New starters ask the same questions, and the AI tools you pay for know none of it.
The short version: start with the shared inboxes and the shared drive, then the CRM. Copy each source's permissions so nobody sees more than they can open today. Keep HR files, tax file numbers and health details out. Try the built-in tools in Microsoft 365, Claude or ChatGPT first, and build your own when your knowledge is spread across several systems.
What is a second brain for business?
The term is popular in personal productivity, where Tiago Forte's book Building a Second Brain describes a system for organising your own knowledge.
The business version works for a whole team. Nobody has to file things into it, because it reads the tools you already use, and it copies those tools' rules about who can see what. It has four parts:
- Connectors to your email, calendar, meeting notes, shared drive and CRM. A connector is a link that lets one app read another.
- A private store holding a searchable copy, refreshed on a schedule.
- Permission rules, so each person only gets answers from things they could already open.
- Somewhere to ask, like Teams, Slack or an AI assistant, with every answer linking to its source.
You'll also hear "AI memory", meaning the memory features in Claude and ChatGPT. Anthropic's memory article says Claude saves memory as topics as you chat. That only covers what comes up in chats, while a second brain covers the records your business already holds.
How is it different from a shared drive, a wiki or AI knowledge search?
The four differ in what goes in and who keeps it current.
| Shared drive | Wiki or knowledge base | AI knowledge search | Business second brain | |
|---|---|---|---|---|
| What's in it | Files people saved | Pages someone wrote | Your documents, and sometimes chats and email | Email, calendar, meeting notes, documents and CRM notes |
| Who keeps it current | Whoever remembers to save the latest version | Whoever owns each page, if anyone does | It re-reads documents as they change | It pulls in new items on a schedule |
| How you find things | Folder names and file search | Search, and links between pages | Ask a question, get an answer from your documents | Ask a question, get an answer drawn from all of it, with a link to each source |
| Can an AI assistant use it? | Only through a connector | Only through a connector | Usually, inside that tool | Yes, through a connector |
A wiki is still worth having, and a second brain can read it too. What a second brain adds is everything nobody writes up, like the quote sitting in an email thread or the promise made at a site meeting.
If your main problem is finding documents, you may not need all of this. Our guide to AI knowledge search covers that narrower job.
What is a business second brain good for?
The best test is the questions your team already asks each other:
- "What did we quote them last time?" The answer comes from the quote email and the CRM note, with a link to each.
- "What did we promise at that meeting?" It finds the meeting notes and the follow-up email.
- "How do we do this here?" A new starter gets the how-to document and past emails about it, without waiting for the one person who knows.
- "Draft a reply to this client." An AI assistant connected to it can check what was said before, so the draft starts from your real history.
Atlassian's State of Teams 2025 report, published in March 2025, surveyed 12,000 knowledge workers in six countries, including Australia, and 200 Fortune 1000 executives. It found executives and teams spend a quarter of the workweek searching for information. And 56% of workers said they often find the only way to get the information they need is to ask someone or schedule a meeting.
AI tools hit the same wall. Atlassian's State of Teams 2026 report, published 27 April 2026, found 69% of knowledge workers say their data and knowledge foundations are not optimised for AI. Only 21% fully trust their AI tools to surface and contextualise relevant information. Both are broad surveys of office workers, so treat them as a rough guide.
What should go in first, and what should stay out?
Start with the two or three sources your team searches most, and add more once those answers are good.
| What goes in | How it gets there | Who can see it |
|---|---|---|
| Shared inboxes (info@, accounts@, quotes@) | A connection to Microsoft 365 or Gmail that picks up new emails on a schedule | The people who can open that inbox today |
| Calendars | The same connection, reading meeting titles, attendees and notes | Each person's own calendar, plus calendars shared with them |
| Meeting notes | Copied from wherever your team already saves them | The people in the meeting, or the team the notes are filed under |
| Documents: templates, policies, price lists, proposals, how-to guides | Synced from SharePoint, OneDrive or Google Drive, a folder or two at a time | Whoever can open that folder now |
| CRM notes and deal history | The CRM's API, which is the door one system gives another to read its data | The same people as in the CRM, or rules you set if the connection uses one admin login |
| Job or practice management notes | Its API, or a scheduled export if it has none | The same people as in the job system |
If your CRM notes are thin because nobody logs their emails, fix that first. Our guide to email to CRM automation covers the options.
Leave these out, at least to start:
- HR files: contracts, pay, leave, performance notes and medical certificates.
- Tax file numbers, bank details and scans of licences or passports.
- Health, disability or NDIS details about the people you support, until you've worked through the privacy side.
- Anyone's own inbox, until you've asked them.
- Old versions. Archive superseded price lists, templates and policies, or the second brain will quote them back to you.
If you'd like a second opinion on what to pull in first, book a free call and I'll go through your systems with you.
Who can see what? Permissions and privacy
A second brain should never show someone more than they can open today. Anthropic's connectors article says "Claude inherits each person's permissions from the connected service." Microsoft's Copilot privacy page says it "only surfaces organizational data to which individual users have at least view permissions." A custom build should copy the same rules from each source.
The catch is loose sharing. AI search will surface a badly shared folder, like staff contracts open to everyone. So check your sharing before anything connects. Microsoft's connector overview also notes that deployed connectors are tenant-wide (open to your whole Microsoft 365 organisation) unless external item security is restricted.
Client details will be in there, so the store has to be private to your business. The OAIC's guidance on commercially available AI products (published 21 October 2024, updated 17 January 2025) recommends, as best practice, that organisations "do not enter personal information, and particularly sensitive information, into publicly available generative AI tools". It also warns that some products have terms or settings that let the vendor collect what you enter for further training. So use business plans and check the training settings:
- Anthropic says it doesn't train its models on Gmail, Drive or Calendar connector data.
- OpenAI says it doesn't train on Business, Enterprise or Edu workspace content by default.
- Microsoft says Copilot prompts, responses and Microsoft Graph data aren't used to train foundation models.
Check whether the Privacy Act covers you. The OAIC says most small businesses are not covered, but some are, whatever their size, including health service providers and businesses that trade in personal information. Its small business page has the full list. For businesses that are covered, the OAIC's AI guidance says to "take active steps to ensure that the use of an AI system will not expose your organisation's personal information holdings to additional security risks."
Ask where the store is kept and where the AI that reads it runs. They can be different places. Microsoft's 3 April 2026 update says it expects local processing of Microsoft 365 Copilot interactions to become available in Australia by the end of 2026.
Tell staff what the second brain holds and who can see what, and write it down. Our AI policy template gives you wording to start from.
Built-in tools or a custom build?
You may not need a build at all. As of October 2026, Claude, ChatGPT and Microsoft 365 Copilot each cover part of the job. Microsoft's own docs now call Microsoft 365 Copilot "Microsoft Copilot".
| Option | What it can draw on | Does it keep its own copy? | Worth knowing |
|---|---|---|---|
| Claude (Projects, memory, connectors) | Files you add to a project, plus connected apps like Gmail, Google Drive and Microsoft 365 | Project files, yes. Google Docs added to a private project stay synced, but that option is off for shared projects. Anthropic says its Google connectors only access your data when you ask | Projects can be shared on Team and Enterprise plans. Each project has its own memory |
| ChatGPT (Projects, Company Knowledge) | Project files, plus connected apps through the Company Knowledge plugin on Business, Enterprise and Edu | Project files, yes. OpenAI's 16 October 2026 Enterprise and Edu release notes say administrator-managed sync connectors have been retired, in favour of real-time apps | Company Knowledge respects each source's permissions. OpenAI says to review source links before relying on an answer |
| Microsoft 365 Copilot | The emails, chats, documents, calendar, meetings and contacts each person can access, plus outside systems through Copilot connectors | It works through Microsoft Graph. Synced connectors copy outside content into Graph, and federated ones fetch it live | It's an add-on to Microsoft 365. An AI administrator sets up connectors. It shows whatever each person can already open |
| Custom build | Whatever you choose, like inboxes, meeting notes, documents, CRM and job notes | Yes, refreshed on a schedule you set | You set the rules for what goes in and who sees it. Claude and ChatGPT both accept custom connectors built on MCP, an open standard for plugging data into AI tools |
If your business runs on Microsoft 365 and most of what you know lives there, start with Copilot and tidy your sharing first. If a small team mostly needs shared templates and policies, a shared Claude or ChatGPT project gets you a long way. Our guide to Claude for small business covers setting Claude up safely.
A custom build makes sense when your knowledge is split across systems that don't talk to each other, like Gmail, a CRM and job software, or when you need rules the built-in tools don't give you. It's also what makes AI agents useful, because an agent drafting follow-ups needs to see what was said before. If agents are new to you, start with what an AI agent is, then our guide to AI agents for small business.
Worked example: an Adelaide accounting firm's first month
Say an 18-person accounting firm in Adelaide: two partners, a practice manager, twelve accountants and three admin staff. They use Microsoft 365, a practice management system and Xero. Every tax time, new staff ask the same questions, and the partners keep hearing "what did we tell this client last year?"
- Week one. The practice manager and one partner pick the first sources: the two shared inboxes, the templates and procedures folders, and client meeting notes. Partner inboxes and the practice management system can wait. Tax file numbers, bank details, ID documents and the HR folder stay out. The OAIC's guidance on the Privacy (Tax File Number) Rule 2015 says TFN recipients must "restrict access to records containing TFN information to staff who need to handle this information under taxation, personal assistance or superannuation law", so leaving TFNs out is the simple choice.
- Week two. The sources are connected with the same permissions as Microsoft 365, so each client folder stays visible only to that client's team. The practice manager tests it with an account set up like a junior's and asks about a partner-only folder. Nothing comes back, which is what she wanted.
- Week three. Six staff start using it. A graduate asks how the firm handles a client who's late with their BAS paperwork and gets the procedure and last year's reminder email, with links to both. An accountant asks what was agreed about a client's family trust and gets the meeting note and the follow-up email.
- Week four. They check the questions that got poor answers. Two came from an old engagement letter template, so they archive it. One had never been written down, so a partner writes a short procedure. They also write a half-page staff note on what's in it, what isn't, and that a person checks anything before it reaches a client.
They could do this with Microsoft 365 Copilot plus a connector to their practice management system, or with a custom build. The decisions in week one are the same either way.
How I run my own business on an AI chief of staff
I run Workvolve on an AI chief of staff I built myself on Claude Code. It uses memory files, scheduled runs and connectors to Gmail, Google Calendar, Drive and my meeting notes. Each morning it gives me a briefing from my calendar, inbox and meeting notes. It keeps a memory of every client and decision, and lines up the day's work before I sit down.
It drafts everything, and nothing sends until I approve it. I'd build that habit into any second brain, along with one rule: when two records disagree, the newest wins, and the answer says which record it used.
Our AI operating system service is where we'd build the same kind of setup for your business: one connected layer across your inbox, calendar, files and CRM, built in your own accounts so you own it.
Want a second brain for your business? Tell us where your team's knowledge lives now. We'll show you what we'd connect first, and what we'd keep out. Book a free call.
Sources
- Atlassian: The State of Teams 2025 (March 2025, PDF) and The State of Teams 2026 (27 April 2026, PDF)
- Tiago Forte, Building a Second Brain
- Claude Help Center: what are projects, chat search and memory, using connectors and Google Workspace connectors
- OpenAI Help Center: Company Knowledge in ChatGPT, Projects in ChatGPT and the Enterprise and Edu release notes (16 October 2026)
- Microsoft: data, privacy and security for Microsoft Copilot, Copilot connectors overview, connector prerequisites and in-country data processing for Microsoft 365 Copilot (4 November 2025, updated 3 April 2026)
- OAIC: privacy and the use of commercially available AI products, small business and the Privacy (Tax File Number) Rule 2015
Facts checked 24 October 2026. This is general information, not legal advice. For your own obligations, check with the OAIC.
Common questions
What is a second brain for business?
It's one private, searchable store of what your business knows, pulled in on its own from email, calendar, meeting notes, documents and CRM notes. Staff and AI assistants can ask it questions and get answers with links to the records they came from.
Is a second brain the same as a company knowledge base?
Not quite. A knowledge base is pages someone writes and keeps up to date. A second brain pulls in what already exists, like emails, meeting notes and CRM records, and answers from all of it with sources. It can read your knowledge base too.
Can Claude or ChatGPT be a second brain for my business?
Partly. Both let you share a project of files and instructions with your team (on Claude, that's the Team and Enterprise plans), and both connect to apps like Google Drive. ChatGPT's Company Knowledge plugin, on Business, Enterprise and Edu plans, answers from connected apps and respects each source's permissions. Both get harder to rely on when your knowledge sits in systems they don't connect to.
Does Microsoft 365 Copilot already do this?
For content inside Microsoft 365, largely yes. Microsoft says Copilot works from the emails, chats, documents, calendar, meetings and contacts each person has permission to access. Other systems, like a CRM, need Copilot connectors set up by your admin. Tidy your sharing first, because it shows whatever each person can open.
Is it safe to put client information in a second brain?
It can be, with care. Use a business plan whose vendor doesn't train on your data, and copy each source's permissions. Keep tax file numbers, bank details and health information out. The OAIC recommends, as best practice, that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools.
How do you stop a second brain giving out-of-date answers?
Archive old price lists, templates and policies so they stop turning up. Set it so the newest record wins when two disagree, and make every answer show its source and date.
