The Numbers
SaaStr has been running AI SDRs for over a year. Not as a test. As part of their actual go-to-market motion. Jason Lemkin, founder and CEO, shared the deployment data across three platforms.
Artisan (outbound): 3,221 emails per month from one platform. Previous human SDR output sat at 75 to 285 emails per rep per month. That is an 11 to 40x volume increase. Response rates hit 5.5% on cold audiences, 11 to 12% on warm leads like recent event attendees. Result: 11 to 13x more responses from the same lead pools.
Agentforce (Salesforce-native): Deployed to roughly 3,000 ghosted CRM contacts. 72% open rate. Highest response rate of all platforms tested. Already closing deals from contacts that got zero follow-up for months.
Qualified (inbound): 668,591 sessions. 91 meetings booked. $1.01M in closed revenue in 90 days. $2.5M in pipeline. In one month, 71% of closed deals came from AI-qualified inbound. Historic average: 29 to 34%.
The Actual Use Case That Works
The insight: AI SDRs perform best on leads your human team is ignoring. Not your best motion. The motion you are not running at all.
Human SDRs force-rank every day. They chase big deals, warm inbound, accounts that could close this quarter. The rest gets nothing. Return event attendees, old website visitors, ghosted CRM contacts. Zero follow-up.
SaaStr deployed AI on exactly those leads. Generated 15% of London event revenue from a segment that was producing $0. Not 15% lift. 15% of total revenue.
What Actually Takes Time
Lemkin's team went through 47 iterations to stop the AI being too aggressive on pricing. Took 30 days of daily tuning. First 1,000 emails required manual review. They still spot-check for 20 to 30 minutes every day.
Message quality matters. Most AI outbound reads like AI outbound. What works: human-written frameworks with AI handling timing, sequencing, personalization. Your best AE writes 5 to 7 templates based on what has actually closed deals. AI executes.
ICP work has to come first. AI does not fix fuzzy targeting. It amplifies it. Sharp ICP produces more of your best customers. Vague ICP produces a spam cannon that burns your domain reputation.
ANZ Context
SaaStr is US-focused with no disclosed ANZ operations, but the data matters for local sales leaders exploring AI to scale without adding headcount. The use case translates: every ANZ CRM has the same ghosted contacts, the same leads human reps deprioritize, the same speed-to-lead gaps at 2am on a Sunday.
The comp math matters too. If human SDRs cost $80k to $100k base in Melbourne or Sydney, and AI covers high-volume, lower-ACV segments or reactivation motions, the ROI case shifts fast. Worth noting: this is not about replacing your best performers. It is about covering the motions you cannot afford to staff.
What to Actually Do
Pick one platform and one bounded use case. Inbound follow-up, CRM reactivation, or high-volume SMB segments under $5k ACV. Get that working before expanding.
Define the handoff before you launch. When AI books a meeting, who picks it up? Most deployments fail here, not at the AI layer.
And track real pipeline, not activity metrics. Emails sent and open rates are interesting. Closed revenue from previously ignored leads is the number that matters.