People
Generate revenue.
Sell. Open doors. Close.
Operator Brief · No. 01
How to Run a Business in the AI Era
AI is good at writing emails, reviewing contracts, and making presentations. Nobody argues with that.
But that is AI as a tool. The real question is what AI looks like as leverage in business. What do you do with Claude Code, agents, and all the talk about agentic AI? What does any of that mean for a cabinet shop, a plumbing outfit, or a Fortune 500 division running the same plays it has been running for thirty years?
That is what this article is about: the simple answer to the question, how do I use AI right now for maximum leverage? The hard part is not the technology. The hard part is getting a business owner to see their own company differently. So let's start there.
I.The Pyramid
The pyramid
Every business is built the same way. Founder, owner, CEO, or board at the top. Producers, the people who actually make the thing or sell the thing, somewhere in the body. In between sits a layer that filters information up and decisions down. We call it middle management.
Small businesses do not think of it that way. They call it the bookkeeper, the office manager, or the dispatcher. Big businesses call it data analysts, financial analysts, and ops planners. Different titles, same job. Sort the information, input the information, pass it along, build the reports.
Top
Founder · CEO · BoardDecisions. Capital. Direction.
Middle
Bookkeeper · Analyst · Dispatcher · Ops PlannerSorts info, passes it along, builds reports. The layer that has to come out.
Body
ProducersThe people who make or sell the thing.
This layer was always a poor substitute for company memory. A client gets upset and the account manager handles it, but the lesson dies in their head. The mistake repeats ten more times. A vendor flakes and the bookkeeper still writes the check because nothing changes in the system. A PM solves a real problem in the field, but that solution never makes it back into the company. The work gets done. The company forgets.
Sure, there was some automation in this, but what passed as automation in most companies still rested on the shoulders of good or great middle management.
The big change is that in the AI era, that layer is done. Not shrinking. Not shifting. Done.
It does not matter whether the business builds the greatest cabinets on the planet or installs plumbing in tract homes. The middle layer has to come out. Refusing to remove it out of stubbornness or sentiment is just throwing money down the drain. Middle managers, even great ones, must become decision-makers or producers in order to remain necessary.
If your company goal is to stay a "team of humans," the competition will eat your lunch.
II.AI-Native
What an AI-native business looks like
AI-native sounds like jargon, but it means something real. It means AI is not an accessory. It is part of the operating system of the company.
Once that is true, the middle layer loses its purpose. The business no longer needs people whose main job is to collect information, clean it up, pass it along, and remind everyone what happened. The model gets simpler.
Solid green: information inbound, filtered by AI. Dashed: direction sent back out the same way.
People
Produce the product.
Make it. Ship it. Deliver the service.
People
Make decisions.
Set direction. Allocate capital. Own the rules.
Everything else becomes a system.
Information flows directly to the decision-maker in real time. No more waiting a month for a P&L. No more waiting a quarter to find out where the business stands. If a decision-maker needs to see something, it should already be in front of them.
Anything done repeatedly gets automated. That is what agents are for. That is why every business owner, whether they run a Fortune 500 division or a two-person plumbing outfit, needs to pay attention to the agent conversation.
The Fortune 500 director looks at the work that hits their desk every week and asks which parts can be automated. The plumber looks at invoicing, email triage, estimating, scheduling, follow-ups, and asks the same question. Same exercise. Same answer.
Most of it can be automated today. Not all of it, but easily seventy percent. What cannot be automated this year will be automatable next year. The direction is set.
So the next question is the one that matters: how do you build an agent, and why?
III.Tasks vs Practices
Tasks versus practices
The answer starts with a distinction that sounds small but is not.
Task
Done once.
- Held together by a person
- Breaks when they leave, get tired, or get busy
- Knowledge dies in the doer's head
- Cost stays the same — or grows
Practice
Done over and over. Gets better every time.
- Capture, rule, tool, oversight, learning
- Survives turnover
- Knowledge lives in artifacts
- Cost falls, quality rises
Companies are organized around tasks. They should be organized around practices. That switch is what makes a business AI-native.
Every practice has six pieces.
01
See
Somebody or something notices what happened.
02
Save
It gets written down in a usable form.
03
Decide
There is a rule for what to do about it.
04
Do
A tool performs the action.
05
Check
Somebody catches it if it is wrong.
06
Learn
The rule gets better next time.
That is the whole game. If the work does not have those six pieces, it is not a practice. It is a task held together by a person.
Tasks break when people leave, get tired, or get busy. Practices do not.
IV.Five Rules
Five rules
1
The practice is the unit of work.
Stop asking, "Can AI do this task?" Ask, "Can this work become a practice?" If the answer is no, the business is still running on people. If the answer is yes, the business is building something that runs without them.
2
AI cannot fix what it cannot see.
Capture comes first. Every call, every decision, every mistake, every weird thing a vendor did has to be written down somewhere a machine can read. Not a Post-it. Not somebody's head. Not a folder of scanned PDFs nobody opens. Something searchable. Something usable. If it matters, capture it. If you do not, it did not happen.
3
AI does the thinking. Tools do the work. People set the rules.
AI reads, sorts, drafts, compares, and recommends. That is what it is for. Tools do the actual action: QuickBooks, the CRM, the calendar, the project software. They have a record. They do what they are told.
People decide what good looks like and where the line is. AI does not know what a business is for. It will not figure out what matters. That part does not change. Do not let AI act without going through a tool. Do not let a tool run without a rule. Do not let a rule exist without somebody owning it.
4
Never waste a mistake.
A mistake fixed once costs money. A mistake turned into a rule pays back forever. Bad companies hide mistakes. Most companies fix mistakes. A few companies make a mistake once and then change the system so it cannot happen again. That last one is the whole game.
5
The point is compounding, not automation.
Automation saves time today. Compounding makes a business better tomorrow. A team writing faster emails saves a few hours. A team that captures every client objection, updates the sales script, and tightens the qualification rule every week is building something nobody can catch. Same AI. Different game.
V.Why This Compounds
Why this compounds
Here is the part most people miss.
A company that runs on tasks is exactly as smart as the people in the chairs that day. Tired person, dumber company. Somebody quits, the company gets dumber.
A company that runs on practices gets a little smarter every time the work runs. The artifact stays. The rule improves. The next time costs less and works better.
Task-based company — flat. Same plays, same people.
Practice-based company — sharper every week.
After a year, the company running on practices and the company running on tasks are not in the same business anymore. One is still running the same plays with the same headcount. The other is running a sharper version every single week.
That is what AI is for. Not cheaper emails. A company that gets smarter without anybody having to push it.
VI.Not Bureaucracy
Don't turn this into bureaucracy
The point is not to write everything down. The point is to write down what makes the next decision better.
If something is being captured that nobody ever uses, stop. If a rule has not changed in a year, it is probably wrong now. Kill it.
The test is not, "Is it all documented?" The test is, "Is the company getting smarter?"
VII.Six Questions
The six questions
For any work that happens more than once, ask:
What happens, and how does anybody know?
Where is it written down?
What is the rule?
What tool does the action?
What catches it if it is wrong?
What is different next time?
If those six questions do not have answers, there is no practice. There is a workflow held together by a person.
That can work for a while. It will not compound. And when the competition figures this out, the gap closes fast.
VIII.The Shift
The shift
| Tasks | → | Practices |
| Memory in people's heads | → | Artifacts |
| A team carrying everything | → | A system that remembers |
| Saving time | → | Getting smarter |
| Doing the work | → | The work doing something back |
This is how a business runs in the AI era.
Anything short of this and the business is in trouble.