Replit's ARM playbook: Running SaaStr with 3 humans and 21 AI agents

The Gist

• SaaStr AI 2026 event run by 3 humans orchestrating 21+ specialized AI agents
• Banned historical talks, focused exclusively on practical agent implementation
• Predicts agent swarms will be commonplace within 18-24 months
• Shift from 'democratization of software' to 'post-software' agentic workflows

• Replit's team demonstrated how they run SaaStr with 3 humans and 21+ AI agents, showcasing radical productivity gains and operational efficiency.
• The energy in the AI space is reminiscent of 2015, signaling a transformative moment in tech where agentic workflows are becoming mainstream.
• Product fluency in sales teams is now a critical leading indicator, surpassing traditional metrics like pipeline.
• AI agents like 10K (VP of Marketing) and QBee (VP of Customer Success) are achieving 70% fewer human hours in operations and generating $1M+ in revenue.
• The democratization of API consumption, not just software development, is the underdiscussed shift enabling non-technical operators to build powerful tools.
• Labeling agent outputs (yes/no feedback) over time compounds into a moat, as agents learn to produce higher-quality results with minimal prompts.

• Measure and prioritize product fluency in sales teams as rigorously as pipeline metrics.
• Implement a 30-day labeling phase for AI agents to refine outputs and build a competitive moat.
• Build N=1 apps on top of existing tools to address specific gaps rather than replicating entire SaaS products.
• Leverage AI agents for hyper-personalized, real-time engagement on social platforms to outperform human-led efforts.

• 143% (SaaStr AI Annual attendance growth year-over-year.)
• 70% fewer human hours (Reduction in customer success hours achieved by AI agent QBee.)
• $1M+ (Revenue generated from AI-qualified leads.)
• 72% open rates (Performance of Agentforce win-back campaigns.)
• 14,230 lines of code (Size of the autonomous VP of Marketing (10K) agent.)