Lot of talk about how founders are trying to use AI. I see 3 buckets:

People who jump on every trend (complex N8N workflows, Claude Code) then move on to the next thing

People who always have Claude and ChatGPT tabs open, constantly copy-pasting to do random tasks

People who want to use AI but aren't sure how, want their teams to use it more but don't know where to start, not sure if they should tell clients they use it

Most fall into bucket 2.

They open ChatGPT, copy-paste their brand voice, paste the email they're replying to, paste the SOP for how they handle this situation, then ask it to help.


Heres what I do:

I stopped treating AI like a tool we go to when we need help and started treating it like a layer that sits on top of our operations.

Everything about our business lives in one place (Notion): strategy, client pipeline, SOPs, roles, meeting notes, marketing campaigns, brand voice, financials, all of it.

If its not in notion it doesnt exist.

All my business context is organized in one system, so AI isn't something I feed information to, it's something that already knows my business.


How it actually works:

1. AI agents that already have context

Built agents inside Notion that handle stuff we used to do manually.

One agent takes sales call transcripts, cross-references our sales process SOP, pulls prospect data, and writes a custom follow-up email in my voice within 10 minutes of the call ending.

Another agent watches our weekly metrics review meeting, extracts every number we mentioned (cost per lead, ad spend, appointments, pipeline value), and automatically updates our dashboard. No spreadsheets, no manual data entry.

These only work because the agents have access to everything: our SOPs, our meeting transcripts, our client data.

If that stuff was scattered across Google Drive and Slack we'd need some complex Zapier workflow that breaks every other week. Don't have the time or patience for that.


2. Infinite context without switching tabs

In Chat/Claude you're constantly reintroducing yourself to the AI. Here's my brand voice again, here's what we do again, here's how we handle this situation again.

Yes there are context windows. And yes, those windows run out.

In our setup I just ask: "Pull our LinkedIn SOP and draft a post about our new offer in my voice."

It references the SOP page, looks at past posts for voice, and drafts it. One query, full context.

Or: "What did we discuss in last week's leadership meeting about Q2 hiring?"

It pulls the answer from meeting notes without me telling it where to look.

I can use any AI model (Claude, GPT, Gemini) all inside the same system without switching tabs, no context window limits, no copying and pasting.


3. The system gets smarter automatically

Every meeting we run gets transcribed and saved: leadership meetings, client calls, team standups.

The more we use the system, the more it learns about us. Six months from now it'll know more about our business than it does today, automatically.

Why am i telling you this?

Because most companies chase the next shiny AI tool thinking that's the answer.

But if your business data is scattered across ten different places, AI will always feel like extra work.

The companies winning with AI aren't using the latest trend, they're the ones who built a foundation first. They organized their business into one system, then layered AI on top.


The hard part:

This requires actually organizing your business first. You can't skip to the AI layer if your operations are chaos.

But once it's built, you stop being the human who explains context to AI fifty times a day and AI becomes something that actually knows how your business works.

I broke down the full setup (how the agents work, how the context system is structured, how it learns over time) in this video if you want to see exactly how it's built.

I thought AI would save me time. It did. But I almost lost $8,000 because of one mistake.

I spend about 3 hours a day on email. Some days more. Reading, Replying, Searching for that message, I know exists but can’t find. Deleting spam, Apologizing for late responses.

It’s exhausting and it never ends. So last week I tried something. I use AI that handle my entire inbox for 7 days. Every reply, Every decision, and Everything.

What I Learned the Hard Way

By the end of the week I figured out what AI is actually good for and what it absolutely cannot handle.

AI is genuinely useful for:

Summarizing long threads so you don’t have to read everything

Drafting simple replies like “Thanks for letting me know” or “I’ll look into this”

Organizing your inbox so you see important stuff first

Handling routine responses that don’t really matter

AI will destroy you if you trust it with:

Anything involving money or pricing

Scheduling without checking your calendar first

Sensitive topics where one wrong word causes problems

Any situation where being wrong has real consequences