Everyone is asking themselves how they can use AI to "meet business objectives" and "drive shareholder value" or whatever.

Most of us have no clue what that actually means.

Many think perhaps it means adding an Intercom chatbot to your website. Or getting a Claude subscription and learning to vibecode. Or asking ChatGPT instead of Google.

Those things are nice and yes they make a difference. But if you asked me for the one singular thing you could do to turbocharge your business with AI, it would be this:

> Record all of your teams meetings (internal + external)
> Turn each one into a markdown note file
> Organize the files neatly into a folder
> Point Claude (or your AI of choice) at that folder
> Oo it has a self-updating model of your business to use as context.

We call ours the business brain.

And it works extremely well - largely because once its set up, keeping it running takes almost zero effort beyond business-as-usual. Here's the problem it solves for us (and what I suspect it will solve for you too):

Most of what our business knows never gets written down:
the pricing logic someone explained on a Tuesday call
the reason you passed on a partnership or canceled a subscription
the client who hates long emails, or who doesnt show without a reminder email.

And writing it down never really happens (we're an early-stage startup, by the time we write SOPs they are stale within a week). We're also remote, which means everything important does get talked about in a meeting at some point, else on Slack. The knowledge ~exists~ but it’s mainly just trapped in conversations.

And, most importantly, AI tools are only as smart as the context you give them. Ask Claude "how should we pitch this client" with nothing else and you'll get generic advice that could apply to a pet groomer or a PE firm. This is not an intelligence problem, its a context problem, as my friends at Permute AI would say.

So TLDR, we connected the two: every meeting becomes notes → the notes become files → and the files become a shared brain that Claude reads. Now anyone on our team can ask "whats our pricing model" or "what did we promise that client in March" and get the real answer, because Claude has the context, sourced from our actual conversations

Here is our exact whole setup, please steal it

1) Record every meeting, internal and external.
We use Notion's AI meeting notes that get auto saved to a Notion table like this (any tool that produces a transcript works)¹. This is the only behavior change required, and it quickly stops feeling weird [see below]

2) Have Zapier turn each new note into a markdown file.
Saved into a shared folder (ours lives in Google Drive). Markdown is just plain text, which is the format AI reads best. Again you can use pretty much any automation tool; we use a Zapier automation (check Notion table for new files → turn new ones into .md format, save to a Google shared folder). You can also use n8n or Claude cowork scheduled tasks for this.

3) Organize that folder like a company wiki.
We copied Andrej Karpathy's layout: a few top-level folders (team, product, sales, decisions) plus an index file explaining whats where. Every few weeks we ask Claude to reorganize and clean it up itself - so the wiki stays tidy without anyone owning "wiki maintenance" as a job. You can literally give Claude the link above and it will do the rest

4) Mount the wiki for everyone.
Each person on the team points Claude (we use the desktop app with a Cowork project) at that shared folder. Same brain, every seat. We have a little instructional set up for this in everybody’s onboarding doc that looks like the following (for anyone who is brand new to Claude this should help)

And that’s it.

First, what’s cool is that the way Karpathy’s wiki skill is setup, Claude automatically takes the meeting notes etc and is instructed to shape them into ‘concepts’ (ex: your competitive landscape, pricing model) and ‘entities’ (customers, employees, partners) that it can reference and build upon over time. This is incredibly useful for context management bc it means Claude knows all the stuff floating around in your head too.

Second, the last step (the wiki being mounted for everyone) is where it compounds. Its not just me the founder with a smart assistant - its the whole team asking questions against the same living record of the business. New employees can get up to speed 10x faster, async. Our advisors can use it to ask questions that would have previously taken a few 30min sessions per week.

I mean it’s now literally step 1 of our onboarding doc. A new hire installs Claude, syncs the Drive folder, mounts the wiki, and tests it with real questions: "who is on our team", "how do we sell", "whats our pricing model". When the answers come back right, they have context it used to take months of meetings to absorb. It’s so cool.

Meetings are the base layer, but you can stack more on top. I've started writing internal memos every few weeks explaining how im thinking about things for our remote team. And, you guessed it, those flow into the brain the same way. You can also go further and connect your project management system (Linear), internal comms (Slack), CRM (Attio), even accounting (QuickBooks) - with Claude these integrations are straightforward via connectors and MCPs.

The best part is the maintenance cost: basically zero. We dont write documentation. We just talk, like we already did, and the pipe does the rest. A customer interview gets doc’s up instantly into feature requests and bug fixes. Advisor meeting gets piped into sales deck updates. The meetings were always full of this institutional knowledge - we (like most businesses) were just letting some of it leak away bc we didnt have the system to capture and organize it faithfully

The whole thing takes an afternoon to set up (the only "code" is one copy-paste step inside the Zap). If you are thinking about implementing something like this for your business, seriously LMK and I am happy to help you set it up.

What a world.
Ramsey