Every client question routes to one person
What we agreed, what we charge, how we run this account. It all waits until the owner is free.
The company brain for small agencies
Out of your head. Into the company brain.
Patricia learns every client's context, your processes and the decisions from your tools, chats and calls. Then she answers your account managers in Slack or Teams, so every question about a client stops ending up with you.
# marketingSaturday
Goke14:37
@Ricardo gm! can you send me the Patricia context, so I can make the ads before Monday?
PatriciaApp14:38
gm! I pulled this from the live site and our persona notes. Here is the ad-ready context:
- Positioning
- The proactive AI coworker for small agencies, inside Slack or Teams
- Difference
- A chatbot waits for a prompt. Patricia spots the work and asks before anything goes out
- Pains
- Scattered client context, forgotten follow-ups, the owner as bottleneck
- Careful
- Use the BetterPic numbers only as a case study. Check pricing first.
patricia.apppersona-note.md
Ricardo15:37
@Goke seems she got it all
Where your clients' context lives today
In heads, DMs and client calls. So every question finds you.
The client call decided it. Nobody wrote it down.
Decisions live in client calls no system can read, so the team works from last month's version.
Every new account manager starts from zero
The client history sits with whoever ran the account. When they leave, it leaves with them.
Everyone has an AI seat. None of them know your clients.
Each person connects the same client tools and pastes the brief in again, one private chat at a time.
42%
of what a company knows is known by one person only
47%
of digital workers struggle to find the information they need to do their job
58%
of the workday goes to work about work, not the work itself
Sources: Panopto with YouGov, Workplace Knowledge and Productivity Report, 2018. Gartner Digital Worker Survey, 2023. Asana Anatomy of Work Global Index, 2023.
What is a company brain?
A living record of what your company knows, decides and does.
It is the shared layer of facts, decisions, processes and permissions that your people and your AI both work from. For an agency that is every client in one place: what was agreed, what was promised, and how each one likes to work. Patricia builds it from what you already use, and keeps it where your team already asks questions.
01
Context
Every client: their processes, their tone of voice, their pricing, and the decisions from every call she joins.
02
Connections
1,000+ tools, from Google Drive and Notion to HubSpot, Google Ads and Meta. You choose what she reads.
03
Routines
The client work she runs every week: the brief before a call, the Monday report, the follow-up nobody sent.
- Slack channels
- Teams chats
- Your calls
Google Drive
Notion
HubSpot
Company brain
facts · decisions
processes · permissions
- Account managers
- New joiners
- The agency owner
Claude and ChatGPT
Wiki, AI seat or company brain
Why the internal knowledge base never became the brain
Notion, Confluence and a ChatGPT seat each hold a piece. None of them were in the client call where the decision changed, and all of them wait for someone to do the work.
Knows your clients
Patricia
Every client account you connect
- A ChatGPT or Claude seat
- What each person pastes in
- A Notion or Confluence wiki
- What someone wrote down
Tool connections
Patricia
Connect a client account once, the whole agency has it
- A ChatGPT or Claude seat
- Every person connects every client tool
- A Notion or Confluence wiki
- Imports kept up by hand
Hears client calls and chats
Patricia
The calls and client channels you add her to
- A ChatGPT or Claude seat
- No
- A Notion or Confluence wiki
- No
Stays current
Patricia
Built from live tools, not a page to maintain
- A ChatGPT or Claude seat
- Starts from zero every chat
- A Notion or Confluence wiki
- Stale until someone edits the page
Shared by the team
Patricia
One brain, answers in the thread for everyone
- A ChatGPT or Claude seat
- A private memory per person, no shared brain
- A Notion or Confluence wiki
- If people go and look
What one person works out
Patricia
Lands in a shared thread the team can see
- A ChatGPT or Claude seat
- Stays in their own chat
- A Notion or Confluence wiki
- Only if they write it up
AI models
Patricia
Not tied to one AI supplier: the best model for each task
- A ChatGPT or Claude seat
- One supplier's models only
- A Notion or Confluence wiki
- The vendor's own AI add-on
Who sees what
Patricia
Per person and per channel, sensitive topics left out
- A ChatGPT or Claude seat
- Whatever got pasted
- A Notion or Confluence wiki
- Page permissions
Does the work
Patricia
Answers, reminds, drafts and reports, with your approval
- A ChatGPT or Claude seat
- Suggests
- A Notion or Confluence wiki
- Waits to be read
The fair question
Could we build this ourselves with Claude Code?
You could. A company brain that holds up needs every part on this list, and someone to keep them working. We have a full-time engineering team on it, and it is still hard. A small agency has client work to do instead.
- Connectors that do not break
- Search across every source
- Permissions per person
- Meeting capture and recaps
- Old facts retired, new ones kept
- Approvals before any draft goes out
What changes on Monday
First ask Patricia. Then ask the owner.
A brain only matters when it moves the work. These are the jobs an agency hands to it first, shown with fictional clients.
The owner is out. The answer is not.
Anyone can ask about a client, a price or a past decision, and get the answer with its source.
Maya10:02
What did we agree with Lumen on the retainer?
PatriciaApp10:02
Kickoff call, 3 Sept: four campaigns a month, reporting every Monday, no TikTok before Q4. Confirmed by email on 5 Sept.
Lumen kickoff callGmail
Day one, with every client's history
A new account manager asks the brain, not the busiest person on the team. When someone leaves, the client context stays.
Tom09:15
I take over Northwind this week. What should I know first?
PatriciaApp09:15
Three things: they approve copy by Thursday, their CEO rejects stock photos, and the Q3 report slipped twice. The full brief is in your DMs.
The client call ends. The to‑dos are already out.
After each client call, every attendee gets their own items, and one thing she can take off their plate.
PatriciaApp11:31
Fernbrook weekly, 3 items for you: the revised media plan by Thursday, the shoot date, the Q4 budget. Want me to draft the media plan email?
The brief lands before the client call
Before a call, whoever runs the account gets what changed since the last one. No digging through threads.
PatriciaApp13:30
Kepler call in 30 minutes. Last time they asked for a CPA under $40. Since then: CPA is $36, one ad was disapproved, and the invoice was paid on Tuesday.
Google AdsQuickBookslast call
It only speaks when something moved
The daily check on a client account stays quiet until a number or a deadline needs a person.
PatriciaApp08:05
Google Ads: spend is up 38% since yesterday, CPA flat. One live ad still says "2025". Want me to fix the copy?
Your own Claude knows the clients too
Connect Claude or ChatGPT through MCP, and every session starts with your agency's client context.
Using Patricia, draft the Lumen Q4 proposal in our tone, with last quarter's results.
patricia · company context
Who sees what
Everyone gets the ICP. Nobody gets the salaries.
A company brain is only useful when your team, and your clients, trust what goes in and who gets it back. You decide both.
Private or shared, per connection
Your mailbox stays yours unless you share it. A client ad account you share answers the whole agency.
Sensitive topics never enter the brain
Salaries, HR and anything you mark as off limits are left out of meeting memory.
Answers fit the person asking
Private conversations and personal memory stay with their owner. A freelancer in one client channel only sees that space.
No draft goes out without approval
Emails and posts she drafts wait for a yes, and by default so do spend and changes. Every action records who asked for it and why.
# meetingsMonday
PatriciaApp12:04
Leadership sync
Some parts of this meeting were left out because of your workspace's redaction settings.
Part 1 · Client capacity
- Two new retainers fit the current team until December.
- Lumen moves to reporting every two weeks.
Part 2 · Team
- redacted: off-limits topic
- redacted: off-limits topic
Next actions
- Maya: update the Lumen reporting schedule.
- Tom: send the Northwind brief to the new designer.
- SOC 2 Type II, audited
- GDPR Compliant
- ISO 27001 Certified
- Never used to train models
We run our own companies on it
Built inside a small team that runs two companies on one brain.
Patricia started as the system our own team needed. BetterPic and Patricia both run on it today.
14
jobs that run without anyone asking
16,884
things the brain remembers about our companies
600+
action items captured from our calls
What it runs for us every week
- The Monday revenue table
- A Google Ads watch that only speaks when something moved
- The weekly SEO recap against competitors
- 130 client security questions answered from our trust docs
“The first moment as a founder that I was not needed for an important decision.”
Ricardo GhekiereCo-founder, on the Saturday thread aboveSet up with you in 20 minutes
From scattered to shared in four steps
No migration and no new app. The brain grows every time she joins a client conversation.
- 01
Add her to Slack or Teams
One install gives the whole agency the same brain, in the place they already work.
- 02
Connect what she should read
Pick the client accounts and channels. Each connection is private or shared. Your call.
- 03
Invite her to your client calls
Decisions and owners land in the brain, and everyone gets their own to-dos.
- 04
Switch on the first routines
Start with one job you repeat for every client every week. Add the next one when it earns trust.
Pricing
One brain for every client. One price.
Founding plan
$45/ month
For the whole agency, every seat included.
- 20,000 credits included every month
- 50,000 free credits with your onboarding call, no card
- Spend cap on by default
- Works in Slack, and Microsoft Teams in beta
Done with you
Want us to build the brain with you?
In a setup sprint we connect your client accounts, load your processes and switch on the first routines together with your team.
FAQ
Questions, answered.
What is a company brain?
A company brain is a living record of what your company knows, decides and does: for an agency, that is every client's context, the decisions, the processes and the permissions your team and your AI work from. Patricia builds it from your tools, chats and meetings, and answers inside Slack, or Microsoft Teams in beta.
How is this different from Notion, Confluence or an internal wiki?
A wiki holds what someone wrote down, and it goes stale when nobody updates it. Patricia also reads the places where client decisions happen, the channels and calls you add her to. You can keep your wiki: she reads that too.
We already pay for ChatGPT or Claude. Why add this?
A chat seat is a private brain per person: every account manager connects their own client tools, what one person works out stays in their chat, and you are tied to one supplier's models. Patricia is one shared brain. You connect a client account once and the whole agency has it, answers land in shared threads with access scoped per person, and she picks the best model for each task from more than one supplier. Through MCP, your Claude and ChatGPT sessions can use the brain as well.
Who can see what?
You choose per connection whether it is private or shared, so a client account can answer the whole agency or stay with the person who runs it. Private conversations and personal memory stay with their owner, and off-limits meeting topics are left out. Every email, post or message she drafts waits for approval, and by default so does every spend.
Does she listen to every meeting and every channel?
No. She reads the channels she is added to and joins the calls she is invited to. You can block topics so they never enter the brain.
Is it for an agency like ours?
Patricia is built for small marketing agencies that run on Slack or Teams and have no engineers free to build internal tooling. Performance and SEO agencies fit best today, because Google Ads, Meta, GA4 and Search Console are the platforms she reads best. Web and development agencies are a worse fit.
How long does setup take?
About 20 minutes on a call with us. After that the brain keeps growing as she joins client conversations and you connect more accounts.
What does it cost?
The Founding plan is $45 a month for the whole agency, every seat included, with 20,000 credits included each month. You also get 50,000 free credits with your onboarding call, with no card needed.
Give every account manager the context only you have.
One brain for every client, in Slack or Teams, set up with you in 20 minutes.