Part 1 of 3 · 4 min read
Personal agents had a busy fortnight.
OpenAI's dots, always-on agents inside ChatGPT that connect to more than 4,000 apps, got a new model behind them on 3 Nov.
Perplexity launched Teammates, agents that work around the clock and message your colleagues in Slack (5 Nov).
Meta is giving its Muse personal agent a voice and a face.
DeepSeek's agent app reached 1.54 million daily users in under three months (Bloomberg, 8 Nov).
This is part 1 of a three-part series:
Today: the second brain every one of these agents depends on.
Part 2: how the routines and skills run day to day.
Part 3: AI agents on top, and when they make sense for a company.
A shameless plug before we continue:
We have two AI Assessment slots left this quarter, to help you plan your AI budget for 2027. It starts with a 20-minute call.
What a second brain is
A place where your knowledge lives in a form your AI can read and keep up to date.
The difference from a shared drive is who does the upkeep. The AI reads new material, updates the pages that already exist and links them together. Every call note, proposal or decision makes the next answer better.
Andrej Karpathy, one of the best-known AI researchers, described the pattern in April. It has spread fast among builders and has barely reached leadership teams.
Anthropic has now put numbers on it. In research published on 6 Nov, Claude answered better with a wiki it had written plus search than with search alone: 61% against 53% of the graded criteria.
Where it lives
A second brain is a collection of plain text files. Where you keep them depends on what your AI tool can reach and what your team already works in:
a folder on your laptop
a shared SharePoint site or Google Drive
GitHub, which keeps a history of every change
For a company, this becomes a knowledge foundation: a curated layer every AI tool can read. Customers, pricing, approved claims and case studies, tone, product facts, and decisions with their reasons.
In practice:
Sales: proposals use this quarter's pricing and approved case studies.
Marketing: drafts come out in the company voice, with only the claims legal has approved.
New joiners: ask ‘how do we handle a discount request?’ and get today’s answer, with its source.”.
Finance: month-end commentary builds on how last quarter was explained.
Without it, every person explains the company to the AI again and again. With it, that happens once.
And Obsidian? It turns the folder into your own wiki, plus a map of how everything connects.
This is mine.
Every dot is a page: a client, a person, a decision, an idea. The lines are the links the AI made between them.
When I start a proposal, the AI already knows the client, the last call and what we charged. The files are mine, so if I change AI tools, the knowledge stays.
What your AI brain needs
A foundation file. In Claude it is called CLAUDE.md: a short page the AI reads first, every time. It says what this is for, where the truth lives, how to work and what never to do.
Connectors. Access to where your knowledge already lives (email, calendar, SharePoint, CRM, call recordings).
Skills. Written instructions for jobs you repeat, like writing a proposal or checking a draft. Improve a skill once and everyone gets the better version.
Routines on a schedule. Best run first thing in the morning or at the end of the workday. I run two every day:
Harvest sweeps my email, call recordings, Notion and AI chats, and pulls out what is worth keeping.
Ingest files into the second brain, updates the right pages in the right folder and flags anything that contradicts what is already there.
Why this is a leadership decision
Your edge: every company can buy the same AI models. Nobody else can buy your context.
Vendor exposure: when your knowledge lives in files you own, switching AI tools is a contract decision. When it lives in one vendor's chat history, it leaves with the vendor.
Continuity: knowledge stays when people leave. For a portfolio company changing leadership, or a fund preparing an exit, that has real value.
One question for your team: if we switched AI tools tomorrow, what would our AI still know about us?
The first step
Before you build a second brain, you need to know where your knowledge lives and where AI pays off. That is what our AI Assessment does. In 2-3 weeks, we map business workflows, systems and documents, and hand you a roadmap ranked by impact.
From there, we build the second brain with your team. It becomes the foundation for your agentic teammates: AI agents that know your business and work alongside your people.
Next in part 2: how the daily routines and skills actually work, and how to set up your first one.
See you soon,
Pooja
PS: If this was useful, please share it with a colleague who keeps re-explaining the company to their AI.
