AI(l) or nothing · Second brain

A second brain for your organisation

Give your AI a memory,
not a better prompt

AI only knows what it can read about you. A second brain is where that lives, so every conversation starts from what you already know.

Memory Rules Hands
Speaker notes

Everyone is trying to make AI smarter with better prompts. Longer instructions, clever templates, a list of tricks. It helps a little. But the real problem is that AI knows nothing about you.

That is what this is about: a second brain for your organisation. We work with one ourselves every day, and we will show you how.

Three words to hold on to: memory, rules, hands.

The main point
Every chat starts from zero.
A second brain does not. A second brain is a shared folder of plain text files about your organisation: who you are, how you work, what was decided. AI reads that folder at every start, follows the rules written in it, and writes back what happened.
Speaker notes

If you remember one sentence: every chat starts from zero, a second brain does not.

It is not a new system. It is a shared folder of text files. The difference is in how you use it: AI reads it every time, follows the rules in it, and writes back.

The problem · how it usually goes

One week of AI in a team of fourclick to follow

Monday · colleague 1"We are an Odoo partner, our tone is..."ten lines of explanation first
Monday · colleague 2Own prompt, different tonedifferent answer to the same question
Tuesday · colleague 1New chat, start overyesterday is gone
Wednesday · colleague 3"How do we do a quote?"AI guesses, the process is written nowhere
Thursday · meetingDecision madesits in someone's notes
Friday · new colleague"Where is that written?"in a head, an email, a folder
Friday · colleague 2Corrects the AIsame mistake next week
Week afterAll over againnothing stuck
Knowledge lost after the conversationevery week again
All gone againknowledge lives in heads, email and loose prompts

Explaining it all over again

What happens

Everyone explains to AI, in their own words, who you are. Every conversation stands alone. What was explained or corrected yesterday, AI no longer knows today.

What it costs

Time per conversation, and answers that differ per colleague. Processes and decisions that were never written down, AI fills in by itself.

AI is only as good as what it can read about you.

Speaker notes

Sound familiar? One week of AI in a team of four. Everyone first types out who we are. Everyone in their own words, so everyone gets a different answer.

Click through the week. A new chat is a blank page. A process that was never written down, AI makes up. A decision sits in someone's notes. A correction is forgotten next week.

The meter fills up again every week. That is not the prompt's fault. AI is only as good as what it can read about you.

01 · Memory · what AI can read

Inside the second brainclick an element

One shared foldersecond-brain/ ├─ Context/ # who we are, what we sell, our tone ├─ Processes/ # how we work, step by step ├─ Projects/ # one folder per client or project ├─ Decisions/ # what we decided, and why ├─ Team/ # tasks and daily notes per person ├─ Daily/ # every morning, where things stand └─ CLAUDE.md # the house rules, see chapter 02
Plain text files
Readable for peopleReadable for AINo proprietary formatLinks between notes
Shared and synced
Same version for allRights per folderVersion historyOffline too

One folder, a fixed place for everything

What it is

A folder of text files about your organisation. Context, processes, projects, decisions and, per person, their own tasks.

Why text

People and AI read exactly the same thing. No export, no integration needed to get at it, and still readable ten years from now.

The agreement that makes it work

Every type of information has one fixed place. No "miscellaneous" folder. That way AI knows where to look, and where to put something.

Click the folder or the blocks for an explanation.

Speaker notes

The memory is simpler than you think. A folder of text files. At the top: who we are and how we sound. Then the processes, the projects, the decisions, and per person their own tasks and daily notes.

Plain text, so people and AI read the same thing. Shared, so everyone has the same version, with rights per folder.

And one agreement makes the difference: every type of information has one fixed place. Then AI knows where to look and where to file something.

01 · Memory · demo with our own process navigator

Our own processes as a clickable mapwhat is the navigator?

Management→ Primary / chain→ Supporting
2.2.1 Introducing our process→2.2.2 Qualify and write a proposal
Process 2.2.2 · Qualify and write a proposal Input: lead, from Odoo CRM Output: proposal or quote with price ladder Steps, roles and rules: as in the navigator Relies on: the proposition and the pricing logic
↳"What do we need before we write a proposal?"
AI reads the process note and names the input, the steps and where it is written.

Your processes: first mapped, then in the brain

The demo

Our own process navigator shows how we work, from management through the primary chain to supporting processes. Every process has input, output, steps, roles and systems.

From map to note

The same process lives as a note in the brain. So the analysis that fills the map also fills the memory of AI.

What you get back

A question about how you work gets the answer you wrote down yourselves, with the place where it lives.

Speaker notes

An example from our own house. This is our process navigator: all our processes as a clickable map, from management through the primary chain to supporting processes. We build such a map for clients too.

Click into the chain, then into qualify and write a proposal. Every process has input, output, steps, roles and systems. And exactly that also lives as a note in the brain.

So if you ask AI what we need before we write a proposal, it does not guess. It reads the note and names the input, the steps and where it is written.

02 · Rules · how AI behaves

The house rules of the brainclick to follow

CLAUDE.md · read by AI at every start## At the start 1. Read today's daily note. 2. Ask who this session is for. ## Rules - Every piece of information has one fixed place. - Never a message to a client without approval. - No personal data of clients in the brain. - A correction becomes a permanent rule. # why: see Decisions/2026-10-30-...
↺"That's not how we do it." Saying it once is enough.

House rules that come along every session

What it is

One file with the agreements for AI: where it starts, what it never does, how it writes. It reads it at every start, the same for every colleague.

A correction becomes a rule

Say "that's not how we do it", and it goes into the file as a rule. Next week AI still knows, and so does your colleagues' AI.

Short, with the why next to it

The rule itself is one sentence. Why it exists lives in a decision with a date and a reason. That keeps the rulebook readable.

With us: over seventy rules, each with a decision behind it.

Speaker notes

The second word: rules. The brain holds one file with the house rules for AI. Where it starts, what it never does. It reads it at every start, and it is the same for everyone.

The best part: a correction becomes a rule. Say "that's not how we do it" once, and it is in there. For your colleagues too.

We have over seventy now. Each is one sentence, and the why lives separately in a decision. Otherwise the rulebook turns into a novel.

02 · Rules · who sees what

A brain that knows a lot must also know what it may not share

1 · Rights

Per folder

Whoever has no access to a folder does not get the files. So neither does their AI.

  • A private folder per person
  • A confidential folder for management
  • Rights in the app, not in the prompt
Why in the app?
2 · Data

No case files

Personal data of clients and their customers stays out.

  • No care or client records
  • A scan before sharing
  • When in doubt: leave it out
How do you check?
3 · Traceable

Everything can be found

What changes can be seen and rolled back.

  • Version history per note
  • Decisions with date and reason
  • Changes in Odoo in the chatter
What is a decision?

What is in the prompt is a request. What is in the rights is a lock.

Speaker notes

The more the brain knows, the more the boundaries matter. Three things.

Rights per folder. Whoever has no access does not get the files on their computer, so their AI cannot reach them either. Data: personal data of clients and their customers stays out, with a scan beforehand. And everything is traceable: every note has a history, every decision a date.

Remember the difference: what is in the prompt is a request. What is in the rights is a lock.

03 · Hands · what AI can do

The brain, connected to your systemsvia MCP

OdooTasks, tickets, ordersread; write with a note in the chatter
Mail and calendarWhat is going onread; prepare drafts
Meeting recordingsTranscriptbecomes minutes with actions and owner
Website and socialDraftsout through one fixed route
Back in the brain
MinutesDecisionTask proposalDaily noteNew rule

Connections: read and write back

What it is

The brain knows how you work. The connections show AI what is happening now: in Odoo, in the mail, in the calendar.

Read first

Reading is allowed. Writing leaves a trail, and a message to a client never goes out without a person's approval.

The loop

What AI does comes back into the brain. Today's minutes are tomorrow's context for the next question.

More on connections: our knowledge page "Let AI code, not call".

Speaker notes

The third word: hands. The brain knows how you work. Connections show AI what is happening now. In Odoo, in the mail, in the calendar, in the recordings of meetings.

Reading is allowed. Writing leaves a trail: in Odoo every change is in the chatter. And nothing ever goes to a client without a person approving it.

And what AI does comes back into the brain. So the memory grows a little every day.

In practice · an ordinary day with us

A day with the second brain

05:50 · morning roundAI sets up the dayreads calendar, recordings and tasks, writes the daily note
09:00 · startA colleague opens AI in the folderdaily note, profile and tasks already read
10:30 · question"Where do we stand with this client?"project folder plus Odoo, answer with source
12:00 · after a meetingRecording becomes minutesdecisions on top, actions with owner and date
14:00 · correction"That's not how we do it"becomes a rule plus a decision
16:00 · sharing knowledgeWay of working changedpresentation and web page follow
17:30 · end of dayOwn daily notewhat was done, and what comes tomorrow
Weekly · maintenanceClean-up roundduplicates, outdated notes, broken links
Speaker notes

This is what an ordinary day looks like for us. At ten to six AI does a morning round: calendar, recordings of meetings, open tasks. It writes the daily note of the organisation.

Whoever starts at nine opens AI in the brain's folder. It has already read the daily note, your profile and your tasks. A question about a client gets an answer from the project folder and from Odoo, with the source. A meeting becomes minutes with actions.

Correct something and it becomes a rule. Change a way of working and the presentations and web pages follow. And every week a maintenance round cleans up.

Who does what

Three roles, each with its own guide

RoleWhat they do
UserEvery dayInstalls the app, always opens AI in the brain's folder, lets AI read and write back, and reports what is wrong.
Administrator on your sideAt the startDecides who sees which folder, switches on the connections for the team and manages the keys.
NewwayOngoingSets up the folders and rules, fills the first knowledge from the process analysis, runs the maintenance round and adjusts rules after corrections.

A brain without an owner goes stale. So every role has a fixed task.

Speaker notes

Who does what? Three roles, and each role gets its own guide.

The user: install, always start in the brain's folder, and report what is wrong. The administrator on your side: rights per folder, switching on connections, managing keys. And us: the set-up, the first content from the process analysis, and the maintenance.

A brain without an owner goes stale. So every role has a fixed task.

The counter-check
What we often hear"Isn't this just a wiki? We already have SharePoint."
  • You read a wiki when you searchAI reads this brain at every start, for every colleague. The knowledge comes to the work.
  • It writes backMinutes, decisions and tasks are added along the way. A wiki only grows when someone remembers to.
  • But it only works if someone keeps it upAn outdated note gives a confident, wrong answer. So: an owner, one fixed place per topic and a weekly maintenance round.
Speaker notes

The strongest counter-argument: isn't this just a wiki? Partly true. The difference is that a wiki waits until someone searches. AI reads this brain at every start. And it writes back, so it grows by itself.

But say it honestly: it only works if someone keeps it up. An outdated note gives an answer that sounds certain and is wrong. That is why there is an owner, and a maintenance round.

Remember

Memory, rules, hands

Three images

  • MemoryA filing cabinet that AI reads through every morning.
  • RulesThe house rules on the inside of the door.
  • HandsA key ring with only the keys it needs.

Three moments to recall it

  • Day 1Retell it to one colleague in three sentences.
  • Day 3Why is a better prompt not enough? Answer without this deck.
  • Day 10Pick one process that now only lives in someone's head. Where would it go in the brain?
Speaker notes

Three images to hold on to. Memory: a filing cabinet that AI reads through every morning. Rules: the house rules on the inside of the door. Hands: a key ring with only the keys it needs.

And recall it rather than rereading. Retell it tomorrow, answer the recall question in three days, and in ten days give one process from someone's head a place.

To close
Give your AI a memory,
not a better prompt.
Memory Rules Hands

Our advice: start with one process and one folder. Let AI read it and write back every day, and turn every correction into a rule.

Speaker notes

Back to the sentence from the start: give your AI a memory, not a better prompt. And the three words: memory, rules, hands.

Our advice: start small. One process, one folder. Let AI read it and write back every day, and turn every correction into a rule. After a few weeks you notice you explain less and get more done.

→ next · ← back · click an element for an explanation · I contents · N notes · F full screen
Contents