Jason Lemkin published a piece on SaaStr last week saying AI agents are hitting 120% of human SDR output. The number is moving around LinkedIn. Most people are reading it as "AI is winning."
That reading misses the part of the story that matters.
The 120% is real. SaaStr now runs 21 agents and three humans across their GTM. They have crossed a million in AI-driven closed revenue. Qualified set 682 meetings in the run-up to their Annual event. Lemkin is honest about the wins and equally honest, in a different post one week earlier, that the same agents have run rogue A/B tests that gave away free conference tickets nobody authorized, promoted events that had already happened, and broken on vendor hot-fixes. Both are true.
What everyone is missing is who the AI is beating.
The SDR market in 2026 is full of people who took the job because the other job they wanted did not call back. Trained for two weeks on a sequence the founder copied from a LinkedIn carousel. Cold-called between TikTok videos. Quit in month four. The actual average outbound BDR in this market dials 22 times a day, gets 1.4 connects, books 0.2 meetings, watches 14% of those no-show, and is gone before the second quarter. That is the human number Lemkin's 120% is measured against.
So yes, an AI agent at 120% of that is winning. It is winning a race where the other runner never showed up.
The harder claim, the one nobody is fact-checking, is that AI replaces a senior closer with twelve years of objection-handling muscle on a six-figure deal with a skeptical CFO. That claim is unproven. It is also where the actual revenue lives. Nobody is showing the AI-vs-human numbers on the deals that are hard to close. They are showing the numbers on the deals nobody bothered to close.
Two things are true at the same time.
AI is eating the bad SDRs. It should. Anyone running outbound with a team of average kids who phone-screened well and showed up in month one will be cheaper, faster, and less wrong with an agent doing that job. Pay for the agent. Stop pretending the previous setup was working.
AI is not eating the actual job. The actual job is humans on the phone with conviction, context, and the seniority to handle a real objection. That job is not commoditizable. That job is also not what most founders are buying when they buy "BDR-as-a-service" or "AI SDR." They are buying the cheap version of a job that was never about cost in the first place.
If the math is not working on your outbound right now, do not buy AI to fix it. Do not buy commission-only humans to fix it. Both are the same move: pretending the job is cheaper than it is. Build the small, expensive, dedicated version of the team first. Get the messaging proven by humans who can defend a real objection. Then layer AI to scale the parts that are proven.
The 120% number is a permission slip to skip that step.
Do not take it.
One piece worth your time this week
"Schmoozing Is Dead, Agents Are Hitting 120% of Humans, and Growth Is the Only Thing That Matters." Jason Lemkin, SaaStr. May 17. Read it here.
Lemkin's full ten points are worth reading even if you disagree with half of them. The strongest one, buried in the middle, is that nobody is paying for table-stakes product anymore. If your product is not differentiated, no rep, human or agent, will fix that.
The weakest one is the schmoozing claim. Schmoozing is not dead. Schmoozing is dead in tech sales because nobody under 30 in tech learned to schmooze. In manufacturing, professional services, the trades, B2B insurance, and most of the deal sizes that actually keep mid-market businesses alive, the relationship is still the moat. The reps closing eight-figure books in those categories are not on Zoom. They are on planes. The framing that schmoozing is dead is what happens when the SaaS world spends too much time talking to itself.
Forward this to a founder who is about to buy their first AI SDR.
Talk soon,
Justin