The Stack
SaaStr runs on 3 humans and 20+ AI agents. Jason Lemkin walked through the back end: commit counts, API costs, live output. The numbers are specific.
10K (AI VP of Marketing): $257 a month. Replaced an entire BI workflow. Runs daily forecasting, tracks every campaign, pushes the top 3 marketing ideas each morning. That is $3,084 a year, or under 3% of one loaded marketing analyst.
Amelia AI (inbound agent on Qualified): 402,000 chat interactions across 2.25M sessions. Booked 614 qualified meetings at ~$85K average deal size. That is roughly $52M of theoretical pipeline. A strong BDR books 10 to 15 qualified meetings a month. 614 in one event cycle is 3 to 5 BDR-years of output, compressed.
Artisan (outbound agent): $500K recovered from B-leads. Not A-leads. B-leads. The ones with real signal but never worth a human's time, so they rot in the database. Point sales agents at the leads humans will not work.
The Costs
Total stack cost sits in the low thousands per month. 10K at $257. Amelia's cost was not disclosed but runs on Qualified's platform. Artisan pricing was not specified. Compare that to loaded headcount: one BDR in Sydney costs $80K to $100K all-in. One marketing analyst costs $90K to $120K.
The ROI math is clean when the work exceeds human capacity. 402,000 conversations cannot be staffed. 614 meetings in one cycle cannot be booked by one person.
Where Not to Point Them
Lemkin's clearest take: do not put AI on A-leads. Your million-dollar inbound gets a reply from your laziest rep in 60 seconds. The gold is in the B-leads that humans never touch.
One more data point: QB (AI VP of Customer Success) manages 150 accounts at under 90 days old, with zero Salesforce integration. No human CSM wants 150 accounts. They want 5, plus a renewal commission and an incentive to hide problems.
The Discount Problem
Amelia gives the right discount inside hard rules and never panics. A rep who smells a deal slipping goes from 20% off to 25%, then 30%, then 34%. The data says it does not move the close rate. On a $10K ticket, sliding from 20% to 34% off is $1,400 of pure margin per deal for zero measurable lift. Across 100 deals that is $140K gone. Guardrail the discount and you stop paying for the panic.
What This Means
This is not a vendor pitch. SaaStr is using itself as a proof point. The stack is live, the numbers are real, and the takeaway is about deployment strategy, not tool selection. Point agents at volume work humans cannot staff, at B-leads humans will not work, and at workflow guardrails that stop margin bleed. Do not point them at your core enterprise accounts.
If you are running a lean sales org and wondering where AI fits, this is the clearest public breakdown of what actually happens when you deploy 20+ agents with 3 humans. The output is measurable. The costs are disclosed. The strategy is replicable.