Frank Sondors posted on August 15: stop hiring SDRs, start hiring GTM engineers. It's the kind of line built to travel, and it did, fast, across a corner of LinkedIn that has been arguing some version of this fight for months. Somewhere underneath the discourse it kicked off is an actual, useful question about who you should hire next, and neither side of the argument, taken alone, has the full answer.
The post that started this round
Frank Sondors runs Salesforge and Mailforge, which puts him inside the infrastructure layer of outbound every day, and his post landed hard because it names something a lot of revenue leaders have felt but haven't said out loud: the traditional SDR job description, source a list, write a sequence, dial for meetings, is increasingly a job a well-built automation stack does faster and cheaper than a person can. His pitch isn't subtle. Replace the headcount that was doing manual research and manual sequencing with a technical operator who builds the system that does both, at a scale no individual SDR could match by hand.
The market data underneath that pitch is real, even if the framing is deliberately provocative. Clay, the GTM-automation platform at the center of most of this discourse, closed 2025 at roughly $100M in annual revenue and reportedly reached a $5B valuation in 2026, growth numbers that don't happen because a handful of enterprises are experimenting on the margins. They happen because a large number of revenue teams have already voted with budget for the automation-first approach Sondors is describing, whether or not they've adopted his exact job title for it.
That last figure is the one worth sitting with the longest. The fastest-growing GTM organizations right now are not the ones that bought a turnkey AI SDR product and called the headcount problem solved. They are the ones investing in the automation infrastructure layer, and then staffing it with someone who can actually build and maintain it, which is a meaningfully different purchase decision than the "replace a person with a subscription" pitch a lot of AI SDR vendors are still selling.
The counter-argument, and it's a good one
Jason Bay runs Outbound Squad, which trains SDR and AE teams at companies like Gong, Zoom, and Rippling, and his position is not the reflexive "AI will never replace people" defense you'd expect from someone whose business is training the people in question. It's sharper than that. Bay's argument is that AI is already better than most reps at writing a cold email, and the job of the rep isn't to out-write the model, it's to know the buyer well enough that the AI has something true to say in the first place.
That's a materially different claim than "don't automate." It's closer to: automation handles the mechanical half of outbound well, and the human half of the job was never really the typing. It was knowing which account, which trigger, which specific pain to lead with, and that judgment doesn't automate away just because the sentence construction does. Bay's real-hiring-data point, that leading AI companies themselves are still hiring reps rather than eliminating the role outright, is worth taking seriously precisely because those companies have the least reason to keep a role around out of sentiment, and the most reason of anyone to have already automated it away if the technology genuinely supported doing so.
It's also worth noticing what the two aren't disagreeing about, because the areas of overlap are more instructive than the areas of conflict. Neither Sondors nor Bay is arguing that outbound should go back to the way it worked in 2019, before volume-based cold email arbitrage collapsed under the weight of stricter spam filtering and buyer fatigue. Both are describing a world where the old, undifferentiated SDR job, manually researching and manually sequencing at low volume, is already gone, regardless of which replacement each of them is pitching. The disagreement is entirely about what fills the gap, not about whether there's a gap to fill.
What a GTM engineer actually does all day
Before deciding who's right, it's worth being concrete about what the job title actually means, because "GTM engineer" gets used loosely enough to cover several different roles. Stripped to the mechanics, it's someone who builds and maintains the data and automation layer that used to be manual SDR work: enrichment waterfalls, intent-signal routing, sequencing logic, and the connective tissue between a CRM, an enrichment tool, and a sending platform.
Notice what's missing from that list: closing skill, buyer psychology, the judgment call about which specific pain point actually lands with a specific economic buyer. That's the gap Bay is pointing at, and it's a real one. A GTM engineer who has never worked a deal cycle can build a technically excellent automation stack that still sends the wrong message to the right list, because nobody on the build side understood the buyer well enough to know it was wrong. The system will execute a flawed strategy just as efficiently as a sound one, which is precisely why the person directing it still matters as much as the system itself.
Where each side is right
| COMPANY PROFILE | THE STRONGER ARGUMENT | WHY |
|---|---|---|
| Early-stage, high list volume, low ACV | Sondors, automate first | Volume and speed matter more than deal-by-deal judgment at this stage; a GTM engineer's system covers ten times the surface area a manual SDR team could. |
| Enterprise, long cycle, complex buying committee | Bay, human judgment stays load-bearing | Multi-stakeholder deals reward the exact contextual read a rep builds over a cycle, not a system tuned for volume. |
| Mid-market, standardized offer, repeatable playbook | Both, in sequence | Automate the top-of-funnel research and sequencing, keep a human on the qualification and close motion once a real signal appears. |
This is the honest resolution, and it's less satisfying than either original post because it isn't a single verdict: the two positions aren't actually opposed once you separate them by company stage and deal complexity. Sondors is describing what happens to the mechanical, high-volume, low-judgment half of outbound, and he's right that a system beats a person at that half now. Bay is describing what happens to the relational, judgment-heavy half of a complex deal, and he's right that no current system replaces a rep who genuinely understands the buyer there. Most real revenue orgs need both halves covered, which means the actual hiring answer for most teams isn't "replace SDRs with GTM engineers." It's "build the automation layer so the humans you keep are working the accounts where judgment actually changes the outcome."
“The system does the parts that don't need a person. The person does the parts that genuinely don't automate. Getting that split wrong, in either direction, is the actual mistake here, not picking the wrong job title to put on the req.”
The hire that's actually wrong
There's a third option neither Sondors nor Bay is arguing for, and it's the one still showing up most often in job postings and vendor pitch decks alike: hiring an "AI SDR" tool as a wholesale headcount replacement, with no technical operator behind it and no human judgment layered on top. It's the easiest option to greenlight, because it looks like a subscription line item instead of a hiring decision, and that's exactly why it keeps getting bought despite underperforming both of the actual positions in this debate. That configuration inherits the weaknesses of both positions and the strengths of neither. It has none of a GTM engineer's ability to build and maintain a real data pipeline, and none of a trained rep's ability to read a buying committee. It is, as the current data on AI outbound tooling's unproven ROI already shows, the configuration teams are least able to justify once someone actually asks for a number, and it's a large part of why clients fire cold email agencies that lean too hard on the tool and not hard enough on anyone directing it.
The pattern generalizes past this specific debate too. We've made a version of this same argument about signal orchestration tools replacing rather than augmenting a workflow: the tool that promises to eliminate a role wholesale usually underperforms the tool that's wired into a workflow a person still owns and directs. GTM engineering, done right, is squarely in the second category. It's infrastructure a skilled operator builds and directs, not a replacement for judgment, and the moment a team treats it as the latter, they've recreated the exact failure mode the "AI SDR" category is already earning a bad reputation for.
Do this next: before posting either job req, map your last twenty closed-won deals against how much of the cycle was mechanical (research, sequencing, follow-up cadence) versus judgment-heavy (objection handling, multi-threading, negotiating scope). Whichever half is larger tells you which hire actually moves your pipeline, and it's very unlikely to be a clean 100% on either side for most companies reading this. Write the split down before you post either job req, not after, because the posting itself has a way of narrowing your thinking toward whichever title sounded more current in the LinkedIn post that convinced you to open the role in the first place. Our cold email team builds exactly this kind of automation layer for clients who still keep reps on the accounts that need one, because the debate online is binary and the right team structure almost never is.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.