Field Guide · hiring
What Does It Cost to Hire an AI Agent?
A working answer for the budget conversation. Compare against the part of the role that scales — not the salary as a whole — and you stop arguing about the wrong number.
The cost question is almost never about the cost. The buyer who types it into a search bar already has a number in mind for what an AI agent should cost — usually around what they pay for a SaaS seat, or around what a chatbot costs, or around what they spent on the last AI tool that disappointed them. The shape of the answer they want is a single dollar figure they can drop into a budget cell. The honest answer is more useful and slightly more annoying. It is a comparison frame: against the part of the role the agent does, not against the role as a whole. This piece walks through that frame in the order a CFO would want to read it.
Why it matters
A dollar figure on its own gets you to the wrong decision twice. If the figure is small, you compare it against ChatGPT and decide it is overpriced. If the figure is large, you compare it against an entry-level hire and decide you would rather have the human. Both comparisons are wrong because both compare the agent against the wrong reference point. The agent does not replace the model in your private chat window, and it does not replace the entry-level hire who handles the work that requires a person in the room. It replaces the part of a senior or mid-market role that scales — the briefs, the monitors, the structured first drafts, the rollups. Compare against that, and the math gets simple.
Start with what the agent is doing. A Fidelic Professional-tier agent runs the daily output of a mid-market role: the brief by Monday, the monitor that fired overnight, the structured first draft that was supposed to be in the doc by lunch. A senior-tier agent runs the harder version of the same shape: the analysis, the early-warning signal, the question worth thinking about that should have surfaced from the data already.
The salary is the anchor for the budget conversation, not the price.
A mid-market role in NYC — a content marketer, a customer service lead, a finance analyst, a sales development rep — runs roughly eight to twelve thousand dollars a month fully loaded. That number includes the base, the benefits, the equipment, the share of the manager who manages them. A senior version of the same role runs twenty to thirty thousand a month fully loaded — director-level work, ten years of context, the judgment somebody pays for.
Those numbers are real. They are the right anchor for the budget conversation. They are also not what an agent costs.
The agent runs at a small fraction of those numbers.
A Fidelic Professional-tier agent costs a small fraction of the mid-market salary it draws from. A Fidelic Expert-tier agent costs a small fraction of what a senior in the same role costs. The exact figures are on the pricing page; the shape of the math is the same regardless: the agent is not priced against the salary, it is priced against the part of the salary that scales.
The part that scales is most of the daily output. The drafts, the monitors, the briefings, the recurring analysis, the structured first cuts. That work, if a human did it, would take fifteen to twenty hours of a forty-hour week. The agent does it in the background, on schedule, in Slack where the team can read it.
What the agent does not replace.
The other twenty to twenty-five hours of the human role is not the part the agent does. Judgment in unfamiliar territory. Accountability your customers can shake hands with. Taste built from ten years of doing the work. The relationship a senior holds with a counterpart at another company. The decision-making that requires being in the room. Those parts do not scale. The agent does not pretend to do them.
The economically correct read of an agent is not “cheaper version of a hire.” It is “the part of the role that scales, separated from the part that doesn't, priced separately, deployed differently.” Most teams keep both. The agent does the part that should already be in the inbox by Monday morning. The senior person, a real human, does the part that doesn't scale.
The CFO question.
The right CFO question is not what does an agent cost; it is what would I pay to have the bottom forty percent of a senior role done before the senior arrives in the morning. Phrased that way, the math gets simple. Phrased the other way, it gets confused.
The edge
The frame buyers usually arrive with — agent versus human, dollar for dollar — is the one Sutherland would call the focusing illusion. It makes one variable carry too much weight. Reframed against the part of the role that scales, the agent is no longer a substitute for the person on the team; it is an addition to the team's daily output, doing the recurring work the team should already have done by the time they sit down. The senior person stays. The mid-market role stays, where the work needs a human in the room. The agent does the third thing — the part that should already exist by Monday morning.
Honest take
This frame falls apart when the team's actual problem is not capacity but expertise. If the work that's missing requires a senior person to be in the room — a regulator meeting, a strategic call with a counterpart, a customer who needs to look another human in the eye — the agent is not the answer. Hire the human. The agent is built for the work that piles up between meetings, not for the meetings themselves. Buyers who confuse these two end up disappointed in the agent for not doing a job it was never priced to do. The honest version of the cost conversation includes naming the work the agent is not for.
Cost is a comparison, not a number. Compare against the salary, you get a confused argument. Compare against the part of the salary that scales, you get a clean budget line.