AI SDRs work — but not as the ad promises. In 2026, the volume per rep jumped from ~1,150 to ~7,400 touches/month, while response rate dropped from 4.7% to 2.9%. The cost per qualified opportunity fell from US$487 to US$224 — but only in hybrid teams (AI + human). Volume isn't the metric; dollar efficiency is.
Do they generate pipeline?
Yes. 41% of B2B enterprise teams have at least one AI SDR in production (compared to 12% a year earlier). But the cost savings come from hybrid pods, not AI alone.
The problem: the channel physics
When everyone fires 6x more with AI, the recipient recognizes the pattern and filters. Rates drop 60%+ in 18 months. Domains in volume suffer a reputation drop of ~38 points within 90 days — and there is no public countermeasure against pattern detection. Warming the domain slows down, it does not reverse.
What makes it work
Hybrid model: AI handles the mechanical part (research, enrichment, screening), humans handle the conversation. Relevance over volume. Protect the domain (and WhatsApp number) as an asset. Measure opportunity by dollar, not touches per day.
The effective AI SDR is a system, not a volume button. This is the type of structure I build.