The Numbers
SaaStr went from zero AI agents in early 2025 to 25+ in production by 2026. The result: a 20-person team (including 10 in sales) replaced by 3 humans plus AI. Revenue closed: $2.4M. Pipeline built: $4.8M. Deal volume and win rates: doubled.
Founder Jason Lemkin is not celebrating a productivity hack. He is worried about 2027.
The Operational Reality
Lemkin's take is blunt: AI agents are not better than great human reps. They are just easier to deploy than average ones.
The math: agents require daily training (60-90 minutes of human oversight), frequent rebuilds, and constant corrections. But compare that to recruiting, interviewing, negotiating comp, onboarding over 60-90 days, managing performance reviews, and hoping your hire does not leave in 18 months.
No benefits. No equity refreshes. No bad weeks. No counter-offers from competitors.
SaaStr's agent stack includes Artisan (outbound SDR), Qualified (inbound support, handling 97% of queries), and Agentforce (CRM reactivation). The agents send 60,000+ emails with 5-7% response rates, above the industry standard 2-4%. Chief AI Officer Amélie Lerutte spends 30%+ of her time managing them.
What This Means for Sales Roles
The question is not whether AI agents can close complex enterprise deals. Most cannot, yet. The question is: how many roles in your org are structured, repeatable tasks that an agent could handle with less operational friction than a human?
SDR roles are the obvious target. Outbound email, list building, initial qualification: all structured workflows. SaaStr's data suggests the replacement math already works if you are honest about what average SDR performance looks like.
AE and AM roles are safer for now, but the timeline is compressing. Lemkin's SaaStr AI event grew 132% YoY. The tooling is improving fast. The operational advantage of agents over average performers is real today.
The Uncomfortable Part
Lemkin frames it this way: a truly great human still beats any agent at judgment-intensive work. But most hires are not great. They are solid B-players who execute reliably.
A good agent beats that hire on consistency, cost, availability, and total management burden. Not on raw capability, but on operational ease.
That is the 2027 problem. When agents are easier to work with than humans, the default shifts. Hiring becomes the exception, not the rule.
Worth noting: SaaStr has no ANZ presence. This is a US market data point. But the agent tools are global, and the operational logic applies everywhere.