AI Employees: The Missing Piece
Executive Summary
As of today, what companies don't have yet is AI employees. We have AI tools, AI assistants, and on the frontier, people running "chief of staff" agents that coordinate other agents on a loop. But nobody has yet been given an AI that does a job, indefinitely, the way an employee does. When that arrives, it will be a massive change in how companies use AI, in the employment picture, and in many other social trends.
An AI employee changes the shape of the company itself. The pyramidal org chart exists to manage communication flows, to push a mandate from the top down so everyone moves in the same direction. With AI employees, that is no longer necessary, and the organization gets a lot flatter. Push the idea further and a single AI entity could replace an entire department, because we don't really need to split work anymore when one AI can spin up sessions on demand.
It also changes how companies pay for AI. Today they buy subscriptions and tokens, out of the technology budget. If they are effectively giving tasks to employees, they should be hiring, and at some point AI spend moves out of technology and into staffing and resourcing. That is when AI stops being a tool companies buy and becomes part of the workforce they manage.
Where We Are Today
There are lots of people on the forefront of AI adoption using something they call a chief of staff. Really, that means a single AI agent coordinating other AI agents, running on a continuous loop or a schedule of some sort.
It's impressive, and it points in the right direction. But it is still something an individual rigs up for themselves. It's a personal power tool, not something the company has hired.
What Makes an Employee an Employee
An employee is someone who is given a job, and they continue to do that job indefinitely. They learn on the job. They continue to get direction. They have a manager. And potentially they have direct reports.
AI Will Manage Humans
I think AI absolutely will manage humans. Humans are still going to be needed for a lot of the work in the near term, so the AIs at headquarters will be directing human work. That's not a question of if.
The Org Chart Is a Communication System
Think about a typical company. It has an org chart and reporting lines, and that chart is generally a pyramid. The reason it exists is to manage communication flows. It's how direction is communicated from the top down, so that people have a mandate and are aligned in what they do.
This is literally the only reason that middle management exists in many companies.
Even though it doesn't work all the time, it is intended to make sure everyone, from the top to the bottom, is moving in the same direction and has been given clear instruction on what they should be doing. It's very messy and very human.
With AI employees, that is no longer necessary. Every AI employee can receive the same mandate and the same context directly, with nothing lost in the relay.
The Organization Gets a Lot Flatter
If the pyramid exists to carry communication, and communication no longer needs carrying, the pyramid shrinks. The org chart definitely gets a lot flatter. Much of what middle layers did was relaying information, and that job goes away.
From AI Employees to AI Departments
The next question is whether a single AI entity can replace an entire organization of humans, an entire department.
I think yes. We split work across people because one human can't hold everything. AI doesn't have that limit. We don't really need to split work anymore, because a single AI can handle tasks at effectively infinite capacity by spinning up sessions on demand.
So the "employee" stops being a unit of headcount and becomes a unit of responsibility. "AI employees" might even be the wrong frame. The real shift may be AI departments.
How Companies Will Buy AI
Today, companies purchase AI based on subscriptions and tokens. That's a bit like paying a worker by the keystroke.
In the future, if they are effectively giving tasks to employees, they really should be hiring, whether that's based on headcount or on capabilities. It's a little unclear which. A salary-like fee for a role or department is one option. Paying for outcomes, per sale closed or per ticket resolved, is another.
The budget point is the really important one. Historically, AI spend has been part of the technology budget. At some point, that will change into staffing and resourcing. Once it does, AI is competing directly with hiring a person.