Your AI Goes to Work
For the last several months, I have been teaching a transition that I believe will define personal computing over the next few years.
The app is disappearing.
In its place is the stack.
Apple captured the previous era perfectly with one of the great technology slogans: “There’s an app for that.” Need a ride? There is an app for that. Need to edit a photo? There is an app for that. Need to manage your calendar, book a hotel, pay a bill, track a package, write a document, or order dinner? There is an app for that.
Our slogan for the next era is different:
Your Stack Does That.
Your stack is the collection of capabilities accumulated around your personal artificial intelligence. It knows how you work. It knows what matters to you. It has access, with your permission, to the places where your life exists. It can use tools, operate software, manipulate files, communicate with other entities, and perform work.
You stop searching for an application capable of doing something.
You tell your AI what you want done.
That transition has enormous implications for individuals. But it may have even larger implications for the enterprise.
One Billion Direct Relationships
I believe roughly one billion people will eventually spend at least $100 per month on artificial intelligence.
More important is where that money will go.
It will go primarily to the companies producing frontier intelligence itself.
OpenAI. Anthropic. Google. xAI. And whoever joins that very small group.
There will certainly be intermediaries along the way. There always are.
AOL was an extraordinary intermediary during the early Internet. For millions of people, AOL essentially was the Internet. It made a frighteningly complicated new technology approachable.
Then people discovered the Internet itself.
They no longer needed someone to construct a carefully controlled environment around it. Search engines increasingly gave people direct access to whatever had just been published somewhere in the world.
The intermediary became the constraint.
Artificial intelligence will follow a similar path.
During 2027 and the years immediately following it, frontier providers will continually release extraordinary new capabilities. When one appears, people will want it immediately.
They will not want to wait six months for their accounting application, CRM vendor, project-management platform, property-management software, or some other intermediary to incorporate it.
The intelligence will already be available.
People will simply go to the intelligence.
That is why the durable relationship is not between a person and hundreds of AI applications.
It is between a person and their AI.
The Enterprise Is a Person Too
Now extend the same idea one level higher.
An enterprise is also an entity.
Rainbow Packaging is an entity. Boeing is an entity. The College of Charleston is an entity. A five-person landscaping company is an entity.
And just as a person develops a direct relationship with frontier intelligence, the enterprise will develop one too.
This is what I mean when I use the term Enterprise AI.
Enterprise AI is not an AI application purchased for every employee.
It is not a chatbot placed on top of the company intranet.
It is not Microsoft Copilot with the company logo added to it.
The enterprise itself has intelligence.
And that intelligence exists primarily to understand, protect, maintain, and operate the enterprise.
It knows the company's customers, products, orders, contracts, inventory, financial records, policies, procedures, intellectual property, trade secrets, communications, commitments, and operating history.
More importantly, it knows who is allowed to do what with them.
Your Company Does Not Give You an AI
This distinction becomes much easier to understand if we look at the smartphone.
Businesses conduct an astonishing amount of business through smartphones.
Yet for most workers there is no meaningful concept of a “company smartphone.”
It is your phone.
You bought it. You selected it. Your photographs are on it. Your family communicates with you through it. Your dentist sends reminders to it. Your bank is on it.
And your employer uses it too.
The company does not own your smartphone simply because you use it to conduct company business.
Instead, the company gives your smartphone permission to interact with certain company resources.
Artificial intelligence will work the same way.
You will have your AI.
Your employer will have its AI.
When you go to work, your AI goes to work with you.
That is the architecture.
Your AI Talks to Their AI
Imagine you wake up tomorrow morning and ask:
“What do I have going on today?”
Your AI might answer:
“You have a dentist appointment at 10:00. You need to call your mother. Your electric bill is due Friday. And there are six customer orders at work that require your attention this morning.”
Notice what just happened.
There was no separation between your personal calendar application and your company's order-management system from your perspective.
You asked one question.
Your AI knew how to find the answer.
The dentist appointment came from one part of your life. The six orders came from an enterprise where you happen to have permission to access certain information.
Your AI did not become the company's AI.
The company's AI did not become yours.
Two independent entities interacted.
That distinction is going to matter enormously.
The Enterprise Becomes a Knowledge Network
Today we tend to think of an enterprise as a collection of applications.
Salesforce contains this.
SAP contains that.
SharePoint contains something else.
Another database contains inventory.
Another application manages service tickets.
Another manages payroll.
Another contains contracts.
Employees spend enormous amounts of time moving between these representations of the company.
Artificial intelligence changes the abstraction.
The enterprise increasingly becomes one enormous governed knowledge network.
Some of its records describe customers.
Some describe orders.
Some describe employees.
Some describe machines.
Some describe contracts.
Some describe intellectual property.
Some describe commitments that have not yet been fulfilled.
And every person, AI, or machine interacting with that network has permissions.
Can John retrieve this record?
Can Susan modify it?
Can the warehouse AI update it?
Can this employee's personal AI see the customer's pricing history?
Can the CEO's AI retrieve the acquisition analysis?
Can an outside accountant access these financial records for thirty days without seeing anything else?
This is what much of Enterprise AI actually becomes.
Not another interface.
A permissioned intelligence layer around the enterprise.
CRUD Becomes More Important Than Chat
Database engineers have used a wonderfully boring acronym for decades:
CRUD.
Create.
Retrieve.
Update.
Delete.
Those four actions describe an enormous percentage of what happens inside a company.
Create an order.
Retrieve a customer record.
Update a shipment.
Delete an obsolete file.
The AI era does not make these operations disappear.
It makes their interfaces disappear.
An employee does not necessarily open the order-management application anymore.
She tells her AI:
“Make sure the three Charleston orders ship today.”
Her AI determines what that means, discovers what systems or enterprise capabilities are available, establishes her authorization, communicates with the enterprise AI, performs whatever actions she is permitted to perform, and reports the result.
The employee experiences intelligence.
The enterprise experiences governed transactions.
That is an important distinction.
Enterprise Software Does Not Completely Disappear
Something still has to maintain the ledger.
Something still has to enforce permissions.
Something still has to preserve transaction integrity.
Something still has to know that warehouse inventory cannot simultaneously belong to two different customer orders.
The enterprise therefore still needs systems.
What disappears is our assumption that humans must operate those systems directly.
That is why I am increasingly skeptical of categories such as “AI software for property managers” or “AI software for insurance agencies.”
Those products may be useful transitional steps.
But they misunderstand where intelligence is going.
The property-management company will have an Enterprise AI.
The property manager will have a personal AI.
The tenant may have one too.
The maintenance company will have one.
The owner of the property will have one.
Those entities will interact.
There is no particular reason they should all live inside somebody else's artificial-intelligence application.
The New Enterprise Boundary
For fifty years, information technology departments worried about devices.
Which computers belong on the corporate network?
Which phones can access email?
Which applications are approved?
Which databases can this employee see?
The next enterprise boundary is different.
It is an intelligence boundary.
The important questions become:
Who is this entity?
What does it know?
What is it allowed to know?
What is it allowed to change?
What actions may it perform?
What information may leave the enterprise?
What information must remain inside it?
What requires human authorization?
What can happen autonomously?
Enterprise architecture moves away from controlling interfaces and toward governing intelligence.
That may be one of the largest architectural changes in the history of business computing.
Bring Your Own Intelligence
We once had Bring Your Own Device.
The next version is much more consequential.
Bring Your Own Intelligence.
A talented employee arriving at a company in 2030 may have spent five years developing an extraordinary personal stack.
Her AI understands how she thinks.
It understands how she researches.
It knows how she prepares presentations.
It knows how she communicates.
It knows the people in her professional network.
It knows her preferred analytical methods.
It may have thousands of capabilities accumulated through years of work.
She is not going to abandon that because she walked through the door of a corporation.
Any more than she would abandon her brain.
The company will not give her a replacement intelligence.
Instead, the company will grant her intelligence appropriate access to the enterprise.
That changes what employment itself means.
The worker is no longer just bringing skills.
She is bringing an accumulated operating environment.
Two AIs, One Job
This is the part I want my students to see clearly.
There are two entities.
There is you.
And there is the enterprise.
You have an AI.
The enterprise has an AI.
Your AI represents your interests, remembers your world, operates your stack, and accompanies you throughout your life.
The Enterprise AI represents the company, protects its knowledge, maintains its operating state, governs access, and coordinates the activities occurring inside the enterprise.
When you work for that enterprise, those two intelligences cooperate.
Your AI asks.
The Enterprise AI decides what you are permitted to know or do.
Your AI performs the work.
The enterprise records the result.
And when you leave the company, something extraordinarily clean happens.
Your AI leaves with you.
The company's knowledge stays with the company.
That is far more natural than today's model.
And once you see the architecture this way, much of what we currently call “enterprise AI software” begins to look temporary.
The future enterprise is not a collection of employees sitting in front of AI applications.
It is an intelligent entity interacting with other intelligent entities.
Including you.