We Compared an AI Agent's Cost to the Headcount It Replaced

A client built an agent to handle first-pass triage of inbound support tickets, with the explicit internal pitch of 'this replaces roughly one full-time triage role.' Six months in, they asked me to help them actually quantify whether that was true, because the finance team wanted a real number, not a pitch.

What 'cost of the agent' actually includes

The naive comparison is API spend versus a salary, and API spend wins by a mile every time, which is why that comparison gets used in every pitch deck. The real cost includes the engineering time to build and maintain the pipeline, the ongoing eval and monitoring work, the human review layer still needed for edge cases the agent correctly escalates, and the incident response time on the occasions it gets something wrong in a way that needs cleanup.

API cost versus salary is the comparison that gets the project approved. Total cost of ownership versus total cost of the role is the comparison that tells you if it actually worked.

The real numbers

• Raw API and infrastructure cost: roughly 4% of the triage role's fully loaded salary cost.
• Engineering maintenance time (prompt updates, eval runs, monitoring): equivalent to about 15% of one engineer's time, ongoing.
• Human review of escalated or low-confidence cases: still required, roughly 30% of the original role's hours, not eliminated.
• Net effect: the role wasn't eliminated, it was compressed to roughly a third of its original time commitment, redeployed to higher-judgment work.

The honest headline isn't as clean as the pitch

'It replaced a person' made a better slide than 'it compressed a role to a third of its time and freed that person for higher-judgment work,' but the second one is what actually happened, and it's still a genuinely good outcome — just not the one the original pitch promised. If you're evaluating an agent project on these grounds, insist on the total cost of ownership number before launch, not six months after, because the gap between the pitch and the TCO is exactly where the project's actual value lives or dies.