AI SDR
/eɪ aɪ ɛs diː ɑːr/
What is an AI SDR?
An AI SDR (Sales Development Representative) is an AI-powered system that autonomously handles sales prospecting activities traditionally done by human SDRs:
- Outbound emails: Personalized, multivariate-tested sequences
- LinkedIn outreach: Connection requests and messages
- Lead qualification: Scoring and routing based on engagement
- Follow-up: Automated sequences with human handoff
The key difference from email automation: AI SDRs generate personalized content, test variants autonomously, and adapt based on results.
The Q2 2024 Inflection Point
A critical reality: AI SDR products didn't work until Q2 2024.
Jason Lemkin documents this clearly:
- Gamma (founded 2020): 5 years to $1M, then 1 year to $80M
- Replit (founded ~2014): 10 years to $1M, exploded in 2024
- Qualified (founded ~2019): "Finally worked after 5 years"
The cause: LLMs weren't good enough. Claude 4 and GPT-4 improvements in early 2024 made the difference.
"If you had a bad experience with AI SDR tools before March 2024, write it off. Different LLMs, different world." — Jason Lemkin
How AI SDRs Work
1. Training (Like a Human SDR)
You can't just "turn it on":
- Figure out what works with real customers
- Tell the AI what worked (scripts, proof points, CTAs)
- Connect to data sources (CRM, enrichment)
- Review outputs and provide feedback daily
- Iterate for ~1 month until it's trained
"If you can't sell it yourself, the AI can't sell it for you." — Jason Lemkin
2. Multivariate Testing at Scale
AI SDRs test combinations humans can't:
- 10+ variants of pain points
- 10+ variants of solutions
- Multiple CTAs
- Multiple proof points
Finding from Artisan: Two lowercase words in subject lines outperform everything else. Only discoverable through massive testing.
3. Autonomous Optimization
The system improves without manual intervention:
- SaaStr saw response rates improve from 3.7% to 4.5% over weeks
- Optimization happens across thousands of variants simultaneously
- Better than any human could manage
Real Metrics
From SaaStr's public data using Artisan:
| Metric | Value |
|---|---|
| Messages sent | 21,000 (5 months) |
| Overall response rate | 7.5% |
| Positive response rate | 4.5% |
| Trend | Improving over time |
Key finding: Warm outbound performs 2-3x better than cold. Most companies have "hundreds of thousands" of unnurtured CRM contacts.
Who Owns AI SDRs?
Deployment varies by company size:
| Company Stage | Owner |
|---|---|
| Early-stage | Founder or Head of Sales |
| Growth | RevOps or Sales Ops |
| Enterprise | Sales Ops (can serve thousands of AEs) |
Warning: "We'll just buy it and hand it to our SDRs" doesn't work. Requires intentional deployment and training.
Common Mistakes
- Buying and turning on without training: Doesn't work
- Not connecting to your data: AI needs context from CRM
- Expecting magic with no inbound: You need some closed customers to train on
- Not reading the outputs: Review every email for the first month
- Trying before Q2 2024: Those products were fundamentally different
The Future
AI SDRs represent the first wave of AI agents replacing knowledge work at scale. The pattern:
- LLMs get good enough (Q2 2024)
- Products suddenly work
- Companies with AI-native GTM outpace traditional sales teams
- Human SDRs focus on higher-value activities
Related Reading
- AI Agents - The broader category
- Jason Lemkin - Leading voice on AI GTM
- Jasper Carmichael-Jack - Artisan founder
Mentioned In
Jasper Carmichael-Jack at 00:02:41
"21,000 messages sent, 7.5% response rate. We test 10+ variants simultaneously - humans can't do multivariate testing at this scale."
Jason Lemkin at 00:14:49
"AI SDR products didn't work before Claude 4 in Q2 2024. Gamma, Replit, Qualified - none worked until the LLMs got better."
Philip Lor at 00:44:00
"Expansion SDRs love having this AI assistant. It makes their job better. AI needs daily oversight but gains are nonlinear."