The Number
SaaStr's AI VP of Marketing, called 10K, worked one hour and executed 125 actions. Cost: $13.42. That is less than California's $16.90 minimum wage and well below what any human marketing operator would cost.
Jason Lemkin posted the data this morning. In that 61 minutes, 10K read 2,463 lines of context and changed 652 lines of code. No human VP does that in an hour because humans context-switch, attend meetings, and wait on input.
The Monthly Cost
Monthly running cost for 10K: $94.51. A second agent, Qbee (AI VP of Customer Success), runs $159.55 per month. Combined: $254.06 for two AI VPs.
A human VP of Marketing runs around $225k base, $300k-plus fully loaded. That is roughly $145 per hour spread across 2,080 working hours. The AI version costs about 1% of that per hour.
Building vs Running
The $13.42 hour was a build session with frontier models (expensive). Running the agent day-to-day costs $30 to $60 per month because most tasks use small, cheap models. SaaStr's full agent stack, six production agents and 14 apps serving 1.9M requests, runs about $2,300 monthly.
Architecture drives the cost gap. Small models handle routine work (ranking posts, drafting tweets). Postgres cache handles most reads, not live API calls. Expensive jobs run once daily on a schedule, not in user-triggered loops.
What This Means for Sales Teams
The bottleneck shifts from budget to task design. Teams that can specify valuable work and point AI at it will scale output without adding headcount. The constraint is no longer "can we afford another hire" but "how well can we direct the agents we have."
Comp data for context: ANZ enterprise AE base averages $120k-$140k, OTE $180k-$220k. An AI agent running $100 monthly replaces specific high-volume tasks, not strategic judgment. The question for sales leaders: which parts of your process are repeatable enough to automate at $13 per hour?