AI Won't Replace Your Best SDRs. It Will Make Average Ones Dangerous.
Rémi
Rémi
April 23, 2026
10 min read
There's a conversation that keeps happening in sales circles right now.
Someone posts "AI SDRs will replace human reps by 2026." A thousand people share it. A thousand more argue about it. And meanwhile, quota is still due on Friday.
We ran a live debate with two of the most experienced outbound practitioners we know. Erwan Gauthier, VP Growth at lemlist. Elric Legloire, who has spent years building and optimizing SDR teams across the US market. They disagreed on almost everything, which made it interesting.
What came out of it wasn't a manifesto for or against AI. It was something more useful: a real map of where AI helps SDRs, where it hurts them, and the one mistake that's silently making reps worse.

The problem with the "AI SDR" category
Before we get into the how, let's name the thing that's been poisoning this conversation.
The label "AI SDR" did real damage to the profession.
"Labeling a software category after a human role completely discredited an entire profession. The moment you do that, people start doing the math: one replaces the other." - Erwan Gauthier, VP Growth at lemlist
That math is wrong. But it's everywhere.
The reality: there are now more than 190,000 SDRs globally, up from 170,000 at the start of 2025. The role isn't dying. It's changing. And the reps who understand that change will pull so far ahead of those who don't that it'll look like a talent gap when it's actually a judgment gap.

What AI actually does well in outbound
Let's be specific. Because "use AI in your outbound" is useless advice.
Account research. This is where AI earns its place without question. What used to take 25 to 30 minutes per account, cross-referencing job postings, press releases, LinkedIn activity, and product reviews, can now be done in under 5 minutes. And the output is better: structured, sourced, with dates so you know what's still current.
Quote Icon
"For account research, I haven't found anything better than AI. Once you've properly trained it on where to look for information, you can get really strong results." - Elric Legloire
That matters because relevant research is the foundation of relevant outreach. Not personalization in the "I noticed you went to Stanford" sense. Relevance in the "I saw your company just expanded into the enterprise segment and you're probably running into this exact problem" sense.
Background preparation for cold calls. If a rep is running 80 to 150 cold calls in a day, they can now have a pre-call brief on every account before they dial. Most teams don't do this today. The ones that do are booking more meetings, not because they're saying something clever, but because they're saying something relevant.
Executing creative ideas at speed. This is the one people underestimate. At lemlist, we've been building prospect-specific sandbox environments: a version of lemlist pre-loaded with campaigns built for the prospect's exact use case, sent as part of the outreach. The concept is creative. The execution used to take hours. Now it takes 15 seconds.
Quote Icon
"Today, AI executes your creativity. If you have an idea, you can act on it. Whereas before, you'd have an idea and ask yourself three times whether it was worth spending an hour on research to make it happen." - Erwan Gauthier, VP Growth at lemlist
That's the real unlock. AI compresses the distance between a good idea and live outreach. The ideas still have to come from a human. The execution can be assisted.

Where AI is making SDRs worse
Here's what nobody in the vendor space wants to say.
The reps who are leaning hardest on AI to generate their emails, their LinkedIn messages, and their call scripts are not getting better. They're getting worse. Faster.
Quote Icon
"The more you rely on AI to generate your emails, your LinkedIn messages, your cold call scripts, the worse you become. You start basing your judgment on what the AI says, you think that's what you should be doing, and you just keep iterating in the wrong direction." - Erwan Gauthier, VP Growth at lemlist
This is the atrophy problem. A muscle you don't use stops working.
Writing a cold email forces you to answer a real question: why should this specific person care about this right now? When AI writes the email, you skip that question. You get an output. You send it. You get a 1% reply rate and you assume the sequence needs work.
What actually needs work is your diagnosis of the account.
Quote Icon
"A senior engineer today is someone who can review AI-generated code and think critically about it. The parallel is exactly the same for a salesperson. If you haven't seen thousands of situations, you won't be able to evaluate what the AI gives you." - Elric Legloire
Senior engineers use Claude. They don't let Claude think for them. The best SDRs will use AI the same way.

The RevOps multiplier nobody is talking about
Here's the second-order effect that most sales leaders are missing.
The ROI question shouldn't be "should we hire more SDRs or invest in AI?" It should be "what's the maximum output our current team is capable of, and are we anywhere near it?"
Most teams are not. Not even close.
The average SDR quota in the US is between 10 and 20 meetings per month. Nobody seriously questions whether that's the ceiling. A well-instrumented RevOps function, with the right AI stack underneath it, can identify every point of friction in the process: the time windows where calls aren't being made, the accounts being worked in the wrong order, the sequences that are running past their useful life.
Quote Icon
"A great RevOps person will anticipate those small friction points in the orchestration. And by solving them one by one, they'll get to a point where they realize: we don't need 5 or 10 more SDRs on this region." - Erwan Gauthier, VP Growth at lemlist
Not because SDRs aren't valuable. Because when the system is optimized, the same headcount produces dramatically more.
The companies that figured this out stopped asking "how many SDRs do we need to hit our number" and started asking "what does our system need to look like so the SDRs we have can hit a bigger number."
One RevOps hire with the right tooling can unlock the productivity equivalent of adding 10 reps. That's not a replacement argument. That's a leverage argument.

The creative advantage AI can never take
There's a dimension of outbound that is genuinely hard to systematize, and it's the one that Erwan kept coming back to throughout the debate.
Call it sales creativity. Not the "put {{first_name}} in the subject line" version. The real version.
Quote Icon
"Outbound creativity starts with the story you want to tell. You bring someone into the scenario you've built for them. It's not just about the ratio - it's about when you reach out, what you say, whether you lead with A then B or B then A." - Erwan Gauthier, VP Growth at lemlist
That's orchestration as a craft. Who do you reach first in the account? What do you say to them versus the economic buyer? What story do you tell the champion so they carry it up the chain?
AI doesn't have an opinion on any of that. It will write you an email for any answer you give it. But the answer has to come from you.
The SDRs who develop this skill, who learn to think in scenarios and sequences rather than in "personalized touches," are building a durable edge. One that compounds over time. One that AI makes faster to execute but cannot manufacture out of nothing.

What this means for how you use AI today
If you're an SDR:
Use AI to research faster and more thoroughly. Use it to execute ideas you already have. Use it as a critic of your outreach, not as the author. If you can't tell whether an AI-generated email is good or bad, that's the gap to close first, because improving your judgment about output is how you improve your output.
Stop using AI to generate your thinking. That's the part that makes you good.
If you're a sales leader:
The question isn't whether to buy an AI tool. The question is whether you've mapped where your team's time actually goes, and where AI can remove friction without removing judgment. Research is the obvious starting point. Preparation for calls is the underrated one.
And before you open a new SDR req, ask seriously: is the constraint headcount, or is it the system those heads are working inside?

The hard truth about the "best SDRs don't need AI" take
Erwan's original hot take was that the best SDRs don't need AI. After an hour of debate, the real version of that idea became clearer.
The best SDRs don't need AI to do their thinking. They use it to go faster. And because they have strong enough judgment to evaluate what comes out of it, they get better over time instead of more dependent.
The dangerous version is the SDR who outsources their thinking to AI, optimizes their process around AI output, and slowly loses the ability to diagnose what's actually wrong when things stop working.
AI amplifies. If your fundamentals are strong, it makes you faster and more creative. If your fundamentals are weak, it makes you confidently mediocre at scale.
That's not an argument against AI in sales. It's an argument for developing the judgment to use it well.

So no, AI won't replace your best SDRs. But it's quietly running a sorting test on every sales org right now, and most leaders don't realize the test has already started.
The two things this whole debate kept circling back to, individual judgment and system design, aren't actually separate questions. A rep with sharp judgment plugged into a sloppy system still wastes half their day on friction nobody fixed. A perfectly tuned RevOps stack handed to a rep who can't tell a good email from a generic one still produces generic outreach, just faster. AI doesn't fix either problem on its own. It just makes whichever one you have more visible, faster than it used to.
Which means the actual question for 2026 isn't "how do we adopt AI." It's "which of our two bottlenecks do we fix first." Most teams will be tempted to buy a tool and call that progress. The ones who pull ahead will be the ones who treat this as what it actually is: a forcing function to finally fix the system and train the judgment they'd been putting off for years.

Want to see what it looks like when AI executes your outbound creativity at scale? lemlist was built exactly for that. Start your free trial.
Hi there, I’m Rémi, Senior Sales at lemlist. Like you, I go from sales meeting to sales meeting - and somewhere in between, I tried to share the no-fluff content pieces I wish I’d read when I first started
LinkedIn

A calendar full of opportunities starts here.