A founder I spoke with recently is launching a second product. New buyer, new competitors, a different ICP than the one the core business has sold into for years. The go-to-market team for it is two people: one account executive and one go-to-market engineer. No SDRs, no field marketing, nobody to staff a booth. And the first question on the table was not “how do we build pipeline,” it was “do we even need SDRs, or can we just point AI agents at this?” That is the right instinct attached to the wrong first question.
The question is everywhere this year for a good reason. The economics of cadence-based sales development have flipped. SaaStr now argues that most SDRs will be AI SDRs in 2026 and beyond, and the more cited version of the story is Jason Lemkin replacing roughly a ten-person sales team with twenty AI agents managed by barely more than one human. If you are a founder watching that, the temptation is obvious: skip the SDR hire entirely and buy the agents instead.
But there is a line in SaaStr’s own advice that matters more than the headline. An AI SDR, they note, is a force multiplier for a motion that already works, not a fix for one that does not. That single distinction is the whole problem for a new product launch, and it is the reason “AI SDR versus human SDR” is the wrong place to start.
When I say outbound engine, I mean four things, and none of them is a person. The first is a target list grounded in a real ICP for this product, not inherited from the company that built it. The second is a reason to reach out now: a trigger event, a technology install, a hiring pattern, something that makes the timing defensible. The third is a sequence and an offer worth replying to, written and tested rather than assumed. The fourth is the data plumbing underneath all of it: the CRM object model, the enrichment, the routing, the reporting, so the motion is measurable from the first send.
That is the engine. The SDR, human or AI, is the labor that runs it. The labor is not the engine. And here is the part founders skip: if there is no engine, automating the labor automates nothing. You have bought a faster way to send a guess.
An AI SDR is genuinely a force multiplier. If outbound already works with humans, agents make it cheaper, faster, and more consistent, and the case for them is strong. The issue is the word “already.” A motion that works is one where a defined list, timed against a real signal, run through a tested sequence, produces replies and books meetings. When that exists, multiplying it is smart.
When the product launched last month and nobody has run a single sequence, there is no motion to multiply. You are pointing a multiplier at zero. The agents will dutifully send a thousand emails built on an ICP nobody validated, against a list nobody defended, with an offer nobody tested, and they will do it at a scale that makes the mistake harder to see, not easier. The honest version of “should I buy an AI SDR for my new product” is “no, because you do not yet have the thing it is supposed to multiply.”
The other half of this is the go-to-market engineer, and I want to be fair to the role because it is real and it is useful. The execution layer of a modern outbound engine genuinely is engineering now: Clay, enrichment waterfalls, Apollo, Claude-scriptable tooling, the whole stack. Someone has to build and run that. That is the GTM engineer, and you want one.
What I would not do is mistake that person for the person who designs the engine. Many operators who brand themselves as GTM engineers are Clay-certified, have built some sequences, and have never sat inside a real sales cycle. That is an execution skill, not a strategy one. The person who decides the ICP, the signal, and the offer needs to have watched deals win and lose, because the engine is only as good as those decisions. Build versus design is the distinction, and a lean team needs both even when it is two people wearing four hats.
We recently scoped exactly this with a data-infrastructure company spinning up a second product line alongside its core business. The founder had already automated their own recruiting sourcing function, with no sourcers left on the payroll, and reasonably asked why sales development should be any different. Fair question, and the answer is instructive.
Sourcing got automated because it had a working motion first: a defined candidate profile, a channel that produced replies, a sequence that booked calls. They automated something that already worked. The new product had none of that. There was no ICP written down, no target account list, no signal to time outreach against, and no sequence anyone had tested. Automating the SDR function there would not have automated a motion. It would have automated a hypothesis, at volume, before anyone checked whether it was true.
- Write the ICP and the buying persona for the new product. New product, new buyer. Do not inherit the parent company’s profile because the logo on the deck is the same.
- Build one target account list you can defend, grounded in a real signal: a trigger event, a tech install, a funding or hiring pattern. A list without a reason to reach out now is just a database.
- Write and test one sequence and one offer by hand. Get ten real replies before you automate a single send. Cheap, falsifiable, and it tells you whether the engine has any compression in it at all.
- Stand up the data plumbing. The CRM object model, enrichment, routing, and reporting, so the motion is instrumented from day one rather than reconstructed later.
- Only now decide the labor. With the engine built and proven on a small scale, one operator plus automation can run what used to take a team, and the AI SDR finally becomes the force multiplier it is sold as instead of an expensive way to scale a guess.
You will probably end up with fewer SDRs than you would have hired two years ago. That part of the prediction is correct, and I am not arguing against it. But you get there by building the engine first and automating a motion that already works, not by buying agents and hoping a motion shows up behind them. If you are launching outbound for a new product and want a second set of hands on the engine before you decide the labor, that is precisely the work we do in programmatic outbound and GTM strategy, so tell us what you are launching.
- SaaStr. “Dear SaaStr: Should I Use an AI SDR Before I Have a Working Human SDR Motion?” 2026. saastr.com
- SaaStr. “Why Most SDRs Will Be AI SDRs In 2026+.” 2026. saastr.com
- Lenny’s Newsletter. “We Replaced Our Sales Team With 20 AI Agents, Here’s What Happened Next.” January 2026. lennysnewsletter.com