What Does an Unfilled Seat Cost vs. an Agent Seat?

An unfilled seat costs far more than the salary you're not paying: it's months of undone work, recruiting and ramp time, and — the expensive part — the founder absorbing the overflow. An agent seat costs design time up front and usage costs ongoing, and it starts producing in days. The honest comparison isn't "agent vs. employee." It's "agent vs. the empty box on your chart that's been open for a quarter."

A note before the math: every number below is illustrative — "say you..." arithmetic to make the structure visible, not a claim about your business or anyone else's results. Plug in your own figures; the shape of the conclusion survives.

What does an unfilled seat actually cost?

Founders track the cost of employees obsessively and the cost of vacancies not at all, because a vacancy never shows up on a P&L line. It shows up in four hidden places:

• The undone work. Say the open seat is an ops coordinator you've budgeted at $90k. The seat exists because roughly $90k worth of coordination per year needs doing. Three months unfilled ≈ $22k of work that either didn't happen or happened badly.
• The overflow to expensive people. The work doesn't vanish — it flows uphill. If you bill your own time at, say, $400/hour and the vacancy eats five hours of your week, that's $2,000/week of Architect time spent doing coordinator work: over $25k per quarter, before counting what you didn't do with those hours.
• The search itself. Sourcing, screening, interviewing — hours from you and your team, plus recruiter fees if you use one (commonly quoted at a meaningful percentage of first-year salary).
• The ramp. Even after the hire, months pass before the seat produces at full capacity — while costing full price from day one.

Stack those and a "$90k seat" open for a quarter plausibly costs more in that quarter than the salary would have. The empty box is the most expensive box on the chart.

What does an agent seat cost?

An agent seat — a function running a FAST loop on the agentic org chart — has a different cost shape entirely:

• Design time. The Architect (you, or a pod lead) defines the loop: input, skills, tools, output, and the written standard for "good." This is real work and the part most people skip. Budget focused hours, not months.
• The training period. While the loop matures, a human reviews output and gives feedback each cycle. This costs review attention — the same attention the vacancy was already eating, but shrinking weekly instead of compounding.
• Usage. Model calls, tool subscriptions, infrastructure. Costs scale with volume of work actually done — usage-shaped, not salary-shaped. No employer taxes, benefits, equipment, or severance attached to the seat.

Two structural differences matter more than any dollar figure. Time-to-productive: days to a working draft loop versus months to a ramped hire. Failure cost: a loop that doesn't work out cost you design hours; a mis-hire costs a salary, a severance, a restart of the search, and morale.

The comparison, side by side

When is the hire still the right call?

Whenever the seat's real product is judgment, trust, relationships, or hands in the physical world. An agent seat filling a taste-shaped hole is a worse failure than a vacancy, because it produces confident output nobody with standards approved. The dividing line — which seats convert and which don't — is drawn carefully in which roles should AI agents fill first?

The practical move for most open reqs: split the seat. Give the repeatable digital slice to an agent loop now, and re-scope the human req around what's left. Sometimes the leftover justifies the hire; often it folds into a seat you already have. If you want the agent side built and run as a service rather than a project, that's MAKO's job.

The real question the empty box is asking

An org chart with a chronic vacancy is telling you something: the business needs the output, but the market price of a human producing it hasn't cleared your willingness to pay. The agentic chart resolves that honestly — the output gets produced by a loop, the human budget goes to seats that need humans, and you stop paying the compounding tax of the empty box. What that does to team structure and reporting lines is the subject of span of control when your reports are agents.

FAQ

Is an agent seat always cheaper than a hire?

No. For work that needs taste, trust, relationships, or physical presence, a good hire is worth every dollar and an agent seat is a false economy. The agent seat wins specifically on repeatable, digital, reviewable work — which is a large share of most open roles, but never all of it.

What is the biggest hidden cost of an unfilled seat?

The founder's attention. Unfilled work doesn't disappear — it flows uphill to the most expensive person in the building. The visible cost is the undone work; the invisible cost is what the founder didn't design, sell, or decide while covering the gap.

What does an agent seat actually cost to run?

Three components: the Architect's time to design the loop and write the standard, a training period of human review while the loop matures, and ongoing usage costs for models and tools. Exact figures depend on your stack and volume — the structural point is that the ongoing cost is usage-shaped, not salary-shaped, and starts in days rather than months.

Should I stop hiring entirely?

No — change the default. Make every open req justify why the work isn't a loop before it becomes a salary. You'll still hire for judgment, relationships, and outcome ownership; you'll stop hiring to buy throughput on repeatable digital work.