SaaStr Now Runs 20+ AI Agents (More Than Humans): Here's the Data

How SaaStr Uses AI Agents for Sales Development

This is SaaStr showing their receipts on AI agents for GTM - not vendor marketing, but actual operator data from running 20+ agents across their business. The numbers challenge a lot of conventional wisdom about AI SDRs.

32x output increase is real, but the insight is deeper. Human SDRs sent 75-300 emails per month. AI agents sent 60,000 hyperpersonalized emails in 6 months. That's the headline, but the real insight: humans were only doing 3% of what was possible. Why? Not laziness - they were rationally managing their time. Every salesperson force-ranks leads in their head and puts effort into the big deals closing this quarter.

"Do the work humans don't want to do" is the killer use case. The 70% open rate on Agent Force sounds impossibly high until you understand what it's doing: following up with leads that human SDRs ghosted because "$10K wasn't worth their time." These were people who actively reached out wanting to sponsor 5-6 figure deals. Humans wouldn't respond. AI agents did. 15% of London event ticket revenue came from AI agents doing work humans refused to do for 6 years.

The real bar: "as good as average human with 24/7 consistency." Lemkin pushes back hard on critics who say AI emails aren't impressive. The customization ranges from "3 to 6 on a scale of 1 to 10" - not the artisanal email someone spent a month crafting. But average human SDR emails are often worse and wildly inconsistent. "Would I have written this? Maybe 50/50" - that's good enough when it runs while you sleep.

"Your problem becomes keeping up with your agents." Once you deploy multiple agents, you enter "a new world where you have to learn to keep up with your agents." Amelia gets constant Slack notifications - AI agent did this, AI agent booked that. People arrive at events saying "Thanks Amelia" and she has to clarify "Oh, Amelia AI helped you, not real Amelia."

Start with work that isn't happening, not mission-critical. Don't unleash AI on your entire database. Train sub-agents on batches of 800-1000 leads with specific personas. Start with ghosted leads or low-priority segments your team says they'll work but doesn't. Don't expect magic - take what humans figured out, document it, train the agent on what works.

11 Insights From Lemkin on AI SDR Performance

• 20+ agents, more agents than humans - SaaStr now operates with more AI agents than full-time employees
• 60,000 personalized emails in 6 months - 32x output vs human SDRs (75-300/month)
• 6% response rate - Double the 2-4% average for human SDRs
• 70% open rate on ghosted leads - Agent Force following up on leads humans wouldn't touch
• 15% of ticket revenue from AI - Doing work humans refused for 6 years
• "Do work humans don't want to do" - The real killer use case for AI SDRs
• Customization is "3-6 out of 10" - Not artisanal; just consistent and 24/7
• Don't unleash on full database - Sub-agents with 800-1000 lead batches, specific personas
• Keeping up with agents is the new problem - Constant Slack notifications, meetings booked while sleeping
• People don't mind talking to AI - They want help; great AI beats mediocre human
• Start low-stakes - Ghosted leads, low-ASP products, not mission-critical deals

What This Means for B2B Sales Teams

SaaStr runs more AI agents than humans. The killer use case isn't replacing good salespeople - it's doing work humans refuse to do. 70% open rates on ghosted leads, 15% of ticket revenue from AI following up on prospects humans ignored for 6 years. The new problem: keeping up with your agents.