Take a labor cost running at 32% of revenue.
A generic chat session tells you that's high.
But an intelligent layer built around your business tells you why, by how long, and what's driving it.
Most people use AI all the time now.
But the output is shallow.
And it never tells them anything specific to their business.
So they assume generic responses (but still good ideas).
Owners are using AI two ways now:
Generic Chat Sessions: Open a browser, paste a prompt, take whatever the model gives back.
Intelligent AI Sessions: Open AI desktop application in sessions preloaded with full business context before the first question.
Intelligent AI sessions require a personal knowledge database which is just a stack of markdown files (.md files) that each chat session reads before it answers anything.
At first you can't tell generic sessions from intelligent. But after enough sessions you start getting answers that are alien.
Back to that 32% labor question.
Ask a generic chat session what to do about it and it tells you you're overstaffed and to cut a shift.
You're overstaffed cut a shift. [Maybe some extra explanation… AI loves to scaffold]
It sounds plausible and it's wrong.
Now ask the intelligent layer that knows your business the same question. The answer looks like this:
32% labor is above your Tuesday baseline of around 22%, but two things are driving it. A closing shift didn't clock out properly and inflated the count. And your new operations hire is two weeks in and still overlapping with the old role. Back those two out and you're at baseline (with MTD still green). Don't move on staffing this week. Let the timekeeping clean up by Thursday and look again.
Then the same system points somewhere else entirely.
Your top vendor's per-unit price is up 4% over the trailing six weeks. That's what's moving against your margin this week.
The same question produced a completely different category of answer.
The first one would have made the business worse.
The second one knows the business well enough to keep you off the wrong direction and put your attention where the money is.
What you'll notice is that you fight with AI because it lacks context all the time.
What you need to do is start building your own context database.
And I've been at this putting in 12-14 hour days for months.
I call it my AI second brain (or Personal Context Database).
A folder of markdown files where sub-agents handle individual jobs, skills pull the right context at the right time, and routing logic decides what to load.
Here’s a peek of my Obsidian Second Brain
The output I have now is solid and it's starting to get alien.
And the compounding is just starting to hit.
Another six months and the output will look like something else entirely.
I've watched many owners use AI every day and never see what it can look like when it pulls 14 targeted .md files and surfaces a small process improvement.
It catches the clock-out error before you cut the wrong person.
It flags the vendor whose prices are up 4% over the trailing six weeks.
You don't get that out of the box.
The few who build the layer underneath the tool are going to have output the rest of the market can't touch.
What's the most useful thing AI does for you in your business right now?
If you want to go deeper:
You should know exactly what your business should be paying you. Click below and let's bring it to the surface.
PS: Not ready to book? Reply GROW and I'll send you the 5-minute playbook I run with my clients to pull more profit out of their business.