.png)
.svg)
Curious how TeamOhana can help you? Talk to us.
In this article
Never miss new content
When Karina Bernacki, VP of People at CoreWeave, showed me their Ask HR digital worker, what stood out most about the tech was how they used their org chart.
Ask HR is a stack of agents. One takes intake in Slack. One asks follow-up questions. One routes the request. One learns from repeat questions and flags gaps. One drafts documentation updates. Together, Karina said, they do the work of an IC2 role. Which means they fill a seat.
That's the shift I think most workforce plans haven't caught up to. We still plan headcount, and headcount only counts people. Meanwhile, a growing share of the work that used to become a req is going to agents. Salaries get budgeted a year ahead; AI spend shows up on the bill after the work is done. Two workforces, one plan, and only half of it written down.
The seat is the unit that holds both
A position is a seat: a defined scope of work with a budget, an owner and a history. It exists whether or not anyone is in it. Historically, the only thing that could fill it was a person. That's no longer true.
Thinking in seats changes how you approach an open position. Instead of automatically asking who to hire, you can first ask whether the work needs to be done by a person at all.
Today, at most companies, a departure produces a req, and a req assumes the answer is a hire. Planning by the seat makes room for the other answers. Does the work need the same person it had before? A revised role? A process change? An agent?
A person is the most expensive answer, so it should come last. Not because people matter less, but because hiring is a long-term financial commitment that should be a deliberate decision, not the default.
Why jobs and reqs can't carry this
Many HR systems were built around two other objects. A job is a template, and a req is a request to hire. Neither one is the seat.
In job management, a role only exists while someone holds it, so an empty seat has nowhere to live and neither does the work an agent is doing. In req management, the system starts from the answer, hire, and tracks the hiring activity well. But it can't tell you what the seat costs, who held it before, or whether a hire was the right answer at all.
What is position management? Position management tracks the seat rather than the person or the req. Every funded position keeps its own ID, budget, owner and history, whether it's filled by an employee, a contractor or an AI agent, or sitting open.
Position management starts from the seat. When the seat opens, it keeps its budget and history, and the decision about what fills it happens with all of that in view.
Copilots and digital workers belong in different places
Not every AI tool fills a seat. Karina draws the line clearly. A copilot makes one person faster at their existing job, so its cost belongs on that person's seat, like any other tool. A digital worker owns a scope of work end to end, the way a role would. It gets a seat of its own, with a budget you can compare against a hire.
That comparison is the point. Once agent work maps to a seat, you can put a person, a contractor and an agent side by side for the same scope and the same budget, and make a call you can explain to your CFO and your board. (Ben Murray at The SaaS CFO frames the same test for finance teams: is this spend filling a seat a human would otherwise fill?)
What happens if nobody plans it
Karina's warning is the one I'd put on every planning kickoff. If HR isn't in the room, she said, "the workforce gets shaped by accident." Finance sees the AI bill. Engineering knows what the tools deliver. And the decision about how the company does its work gets made by whoever is closest to the invoice.
The cost doesn't show up right away. It shows up two years later, as broken spans of control, career paths with missing rungs, and no bench of early-career people, because the work that used to train them went to agents and nobody planned for that.
Where to start
Start with the seats you already have. List every funded seat, filled and open. Give every AI key an owner and sort the spend into enablement, copilots and digital workers. Put each digital worker in a seat of its own, with its budget. Then, at the next departure, ask the seat question before anyone opens a req.
At TeamOhana, Position Management gives every seat its own ID, budget and history, and Token Spend Management brings AI spend into the same plan, so Finance, HR and IT make hire-versus-deploy decisions with the same numbers. See how it works.
See what a governance layer built for workforce decisions looks like in practice. Book a demo.
Take control of your headcount spend
Planning AI Agents in the Workforce Plan FAQs
Simplifying TeamOhana: your questions, answered.
A position is a seat: a defined scope of work with its own budget, owner and history. It exists whether or not anyone is in it, so it outlives any one employee who holds it. Planning by position lets a company decide what should fill a seat before anyone opens a req.
When an AI agent, or a stack of agents, owns a scope of work end to end, treat it as a digital worker that fills a seat of its own, with its own budget. That puts agent spend in the same plan as salaries, so it can be compared directly with the cost of a hire. AI spend that only makes one person faster belongs on that person's seat as a tool cost.
A copilot helps one person do their existing job faster, so its cost belongs on that person's seat like any other tool. A digital worker owns a scope of work end to end, the way a role would, such as a set of agents that handles HR help requests from intake to resolution. Because it does the work of a role, a digital worker gets its own seat and budget in the workforce plan.
Start from the seat, not the req. When a position opens, look at its scope, budget and history, then compare the options side by side: the same role, a revised role, a process change, a contractor or an agent. Because hiring is a long-term financial commitment, it should be a deliberate choice rather than the default answer to every departure.
In job management, a role only exists while a person holds it, so an empty seat, or work done by an agent, has nowhere to live. Req management starts from the assumption that the answer is a hire and tracks hiring activity, not the seat itself. Neither shows what a seat costs, who held it before, or whether a hire was the right answer. Position management starts from the seat, so it can hold people and agents in one plan.
The workforce gets shaped by accident, by whoever is closest to the AI invoice, rather than by a plan. The effects tend to show up a year or two later as broken spans of control, career paths with missing rungs and too few early-career employees, because the work that used to train them went to agents.
List every funded seat, filled and open. Give every AI key an owner and sort AI spend into enablement, copilots and digital workers. Put each digital worker in a seat of its own with its budget, and at the next departure, ask what should fill the seat before anyone opens a req.
