Intelligence Will Be Rented. Context Will Be Owned.

The Signal #093 — Dakota’s read on the AI shift that actually matters to people running a business.

Imagine two companies hire the smartest employee in the world on the same morning.

Same intelligence. Same skills. Same work ethic.

Company A gives them a blank laptop.

Company B gives them the customer history, the operating rules, the open problems, the decisions that were made last year, the reasons behind those decisions, every promise made to customers, every correction the team has learned from, and a live view of what is happening right now.

Which employee is more useful by Friday?

That is where AI is heading.

We are spending a lot of time debating which model is smartest.

But as models get better, cheaper, and closer together, that question matters less for a growing share of business work.

The more important question becomes:

What does your AI know about your business that someone else’s AI does not?

That is the next layer of competition.

Intelligence will be rented. Context will be owned.

The model is not the bottleneck anymore

A frontier model already knows a remarkable amount about accounting, property management, sales, contracts, marketing, support, and operations.

But it does not know your business.

It does not know that the customer called twice yesterday.

It does not know your employee already promised a refund.

It does not know the “closed” work order reopened three times.

It does not know why you made an exception six months ago.

It does not know which system is right when two systems disagree.

It does not know what happened after the last recommendation it made.

The model can be brilliant and still make a bad decision because it is missing the picture around the decision.

That is the context gap.

The Context Gap

Most businesses do not have a data shortage.

They have a context problem.

The information already exists. It is just scattered across:

• email
• CRM
• accounting
• phone calls
• tasks
• calendars
• documents
• support threads
• spreadsheets
• people’s heads

Each system holds a fragment.

The operator is usually the thing connecting those fragments together.

That works when a human is making the decision.

It breaks when the goal is to let AI make or assist with the decision.

A payment is data.

Knowing which property it belongs to, why it happened, whether it was expected, whether it reconciled, and whether someone still needs to act on it is context.

A work order is data.

Knowing this is the third time the same issue appeared, what the vendor already tried, how frustrated the tenant is, and whether nearby units have the same problem is context.

A customer record is data.

Knowing what you promised that customer yesterday is context.

Data tells the system what happened. Context tells it what the event means.

The simplest way to think about the stack

I think the language becomes clearer if we stop treating “data” as one giant bucket.

There are really four layers:

Data — What happened?

Context — What does it mean right now?

Memory — What should not be forgotten?

Experience — What happened after we acted?

Then AI can do something useful with it.

Data → Context → Memory → Experience → Better decisions

That is the flywheel.

The important part is not that the base model magically becomes smarter every day.

The important part is that your system knows more about your business tomorrow than it knew today.

Most AI systems start in the wrong place

The normal conversation starts with:

Which model?

Which agent framework?

Which vector database?

Which AI vendor?

Those are implementation choices.

The real question is:

Can the system reconstruct enough of the business to make a good decision?

Can it identify who this is about?

Can it see what already happened?

Can it tell what is still unresolved?

Can it see what someone already promised?

Can it distinguish current information from stale information?

Can it trace a decision back to evidence?

Can it record the outcome afterward?

If not, changing models is like replacing a smart employee with a smarter employee while keeping both of them blindfolded.

This is why company memory suddenly matters

Businesses have always generated useful knowledge and then lost it.

An employee solves an unusual problem. The fix stays in their head.

A customer explains something important on a call. The recording gets buried.

A manager overrides a normal rule. Six months later nobody remembers why.

An AI makes a mistake. A human fixes it. The correction disappears.

A decision works. Nobody connects the outcome back to the decision.

Historically, capturing every piece of that history was hard to justify because no human could continuously read all of it.

AI changes that.

Machines can search it, connect it, summarize it, reason across it, and surface it at the moment it matters.

That means something that used to look like operational exhaust can become an asset:

organizational memory.

A better mental model: make the business machine-readable

Most businesses today are human-readable.

People know how things work because they have been there long enough.

The real rules live in habit.

The exceptions live in inboxes.

The relationships live in people’s heads.

The reasons behind decisions are rarely recorded.

Then we drop AI on top and expect it to behave like someone who has worked there for five years.

It cannot.

Not because the model is too dumb.

Because the business is invisible to it.

A machine-readable business is different.

Its important entities are identifiable.

Its events are captured.

Its relationships are connected.

Its history is preserved.

Its current state can be reconstructed.

Its decisions leave evidence.

Its outcomes feed back into the system.

That is when AI stops being a chatbot sitting beside the business.

It starts operating inside the business.

The moat is the history nobody else can download

This is the part that gets missed when people say “data is the moat.”

Raw data is not automatically valuable.

A pile of emails is not a moat.

A database full of transactions is not a moat.

What becomes valuable is the connected history of the business:

what happened, to whom, why, what decision was made, what happened next, and what the system learned from it.

A competitor can buy the same model tomorrow.

They cannot buy your last three years of decisions.

They cannot buy every customer correction your team has made.

They cannot buy your exception history.

They cannot buy the relationship between thousands of actions and their real-world outcomes.

That is why this framing matters:

The model is replaceable. The context is not.

And eventually:

The best AI company may not be the company with the smartest model. It may be the company whose AI understands the business best.

What operators should actually do

Do not start by asking what model to standardize on.

Start by asking:

What does our business know that our systems cannot currently express?

Where does context disappear?

Which promises live only in calls?

Which decisions live only in Slack?

Which fixes live only in someone’s head?

Which corrections are never captured?

Which outcomes are never connected back to the action that caused them?

Those are not housekeeping problems anymore.

They are AI infrastructure.

Because the model layer will keep improving whether you do anything or not.

Your company memory will not.

You have to build that yourself.

And when intelligence is available to everyone, the companies that did will have something much harder to copy:

a business that remembers.