Why AI SDRs Failed, and What It Actually Takes

Yeah, today I'll be talking about why the first wave of AI SDRs failed and what it actually takes to automate outbound go-to-market well. So dive right into it. As many of you might know, the AI SDR cycle went through a big hype cycle and a dramatic downfall between 2023 and around 2025. So just as a quick summary of what AI SDR is, essentially the promise was, you know, we would be able to fully automate the SDR role. We would find leads automatically using AI. AI is really smart. It's going to be able to qualify and target those leads, reach out to them automatically, handle the replies, and you take the meetings, right? And the whole narrative kind of around this that the companies, extremely well-funded companies were pushing is that, you know, we're going to replace the humans. You know, there were billboards all over San Francisco, stop hiring humans, hire our AI BDR instead. That was kind of the whole narrative is now we're going to have these digital workers, right? And, you know, over the next year, year and a half or so, customers started using these products and everyone kind of slowly realized that they just didn't really work. There was a dramatic overpromise of AI. And, you know, people started really hating these things. Customers hated them because they didn't work. The email recipients hated them because they would get horribly written and irrelevant emails in their mailboxes. And employees hated them too, because, you know, the whole narrative was that they were going to get replaced. But the one thing was true is that the demand was real, you know, businesses have always and continue to want to do more with less, especially with so much new technology at our fingertips. So, yeah, just to kind of round this out, the CEO of 11x, one of the best-funded AI SDR companies, actually went on a podcast and said this himself. I don't think AI SDRs work. I think they don't work in their current form. And, you know, he realized that over the course of several years. And I agree with this. And he said that the narrative around replacement Doesn't really work. It's the wrong question, and I agree with that. So if replacement isn't the right way to think about AI tooling, you know, what is the right way? And we'll get into that. So just some quick background on myself. As Gerard mentioned, I'm the co-founder and CTO of Aveena. We're a go-to-market automation platform. And previously, I co-founded Bowtie. It was an AI receptionist that was acquired by Mindbody back in 2019. And as part of that business, we scaled our AI systems into thousands of SMBs and deployed custom AI systems into those businesses that essentially manned the front desk for them. And by the way, we grew that entire business basically on the back of cold outbound. a mix of outbound emails and cold calling. The other thing I'll mention is that I'm a CTO, so my background is engineering, and I think I, I might be one of the only engineers they let slip through onto the speaker circuit here, but I promise I'm only gonna show one very simple math equation during this talk. Okay, so jumping into it. 20 years ago, sales was an art. You know, reps had their Rolodexes, they had their entire target market in their head, the pitch, who to call, when to call. It was art and a lot of blood, sweat, and tears. So now sales is still an art. Anyone who's doing sales knows that there's a lot of judgment involved, but you Increasingly, it's becoming an engineering problem. And now with all the technology available to us, a big challenge is how do we engineer these systems to actually give us more efficiency so we can focus on the art? And with AISDRs, I actually think they got both of these wrong. They got the art part wrong and the engineering part wrong, and we'll, we'll get into that. And if you, If you don't believe me for whatever reason, this is just a chart showing the number of go-to-market engineer or RevOps job postings in the last 5 years. So we can see that clearly there's a rise in the technical aspects of running efficient sales organization, especially, and I'll focus on in this talk on the outbound side of things. So If you think about, you know, many of you have worked with amazing sales teams and individuals, and, you know, if you think about what your best SDR does, I'm gonna break that down into 7 things. So the first one is targeting. You know, the first thing is, are we reaching the right companies? Are we reaching the companies that can actually physically use our products and services and will physically get value out of them? And then also, how do we find those companies? How do we actually get those leads and the contact data to reach out to them? Number 2 is timing. You know, what is actually happening at that business on their side? You know, what is their life like right now? And is their life actually one where they're feeling the pain of whatever solution whatever your solution solves. So they might, you know, technically be able to use your product, but are they actually in a position where they would gain value from it right now? Number 3 is the persona. So who's the champion? Who's the decision maker? Who's the economic buyer? Who's the actual user of the product? What is the strategy to reach out to them? What— in what order? And yeah, so number 4, research. Finding the actual facts about the business and understanding your product so you're equipped to communicate the value of your product effectively based on their specific needs. And then number 5, the actual pitch. So are you able to— now that you've done all this targeting, you've timed it well, you know how to reach out, who to reach out to, you've researched that business well, Can you actually craft a message that communicates all of that effectively, that shows them in that message that you've done all that work and do it in a compelling and empathetic way? You know, that in a lot of the, you know, this one definitely has a lot of art involved. Number 6, the actual sending and delivery of the message. So, you know, 10 years ago, there weren't as— there's more of the Wild West. So you could kind of just send a million emails a month and not really get in trouble. Today, it's different. You know, Gmail and Microsoft have both implemented very strict domain reputation rules. You basically don't want to send more than 20 emails a day per mailbox. There's a whole infrastructure system that you want to set up. And this is also the whole reason why we, again, we can't just email every single business in our target market every day. So, you know, every email and LinkedIn DM that you send actually counts because, you know, there's a limit to how many you can send. Okay, number 7, reply handling. You know, once you get that interested response, how are you actually turning that into a meeting? very important as well. And there's a whole system of wanting to nurture them, wanting to get them to the next yes. Okay, so now that we've kind of established what the SDR, what a human SDR that's really good at their job does, you know, what happened with AI SDR? So this is kind of what they promised. They said, you know, we're again, we're replacing the humans, hire us, this digital worker instead of a human. So, you know, they should be doing all of this, right? But in reality, here's what happened. They, they only really did one piece well, which was the piece that was easy to automate, which is the sending, the infrastructure. This is very mechanical. And they, they did do all the other things too. Like they would find businesses in your target market. They would try to craft a plausible pitch for those recipients, but they didn't do any of these things well. They did them very poorly, in fact. And so, so they got the art wrong because, yeah, I would say they, they got both the art and the engineering wrong because the art of it was the human, right? The judgment and the human input going into each of these steps. That's why these systems didn't work well when the AI tried to do it, because they just didn't have that judgment. And they got the engineering of this wrong too, because there were a lot of— there were a lot of pieces. You know, these are very complicated. Each of these steps has a very complex set of things that need to happen for it to be done well. And so they were oversimplified in how they actually handled those things and really only did the sending well. Okay, so to kind of showcase, you know, what happens when you don't do all of these things well, this is the only math in this talk, so So bear with me here, but it shouldn't be too bad. So outbound, I would argue, is a sum. It's a simple equation, and it's— sorry, not a sum. It's a product of all of these steps, not a sum. And what does that mean? If you multiply all of these steps together, if any of them are zero, then Then the whole equation becomes zero. So if you have, if you have all of these things implemented and executed well, then you get meetings. You've targeted the right company, hit them at the right time, reached the right person, researched them well, crafted the right pitch for them, actually delivered the message and not landed in spam, and handled the reply properly. That's going to get meetings. And I think that is pretty intuitive. What might be a little bit less intuitive is if you have a zero on any of these, you're very unlikely to get meetings. So you can do everything well but one and still not get meetings. So what does that look like? If we just take targeting, for example, if you're hitting companies that physically can't use your product, you know, this one is pretty obvious. If I'm selling manufacturing AI software and I send an email to a D2C e-commerce brand, you know, it's just, I'm not gonna make a sale there. So it's pretty obvious, but it's harder than it looks because to do the good targeting, you have to have good data and that takes some engineering and human judgment too. So the timing piece, even if you target the right company well, if you reach them at the wrong time, that's also— it's not gonna convert either. So people have very little attention to spare. They, you know, have certain pain points that they're feeling. If whatever your solution solves is not top of mind for them, They're gonna prioritize other things, and, and your email will just, you know, fall down to the bottom of their inbox. And so what about the pitch itself? You could again perfectly target the right company, have the right timing, reach the right person, do all your research. But then if the actual message that you deliver them doesn't communicate any of that well, Then you've done all that work for, for nothing. So, so the message really needs to map on, you know, what is, you know, all the work that you've done here needs to be communicated. You know, why should they care specifically, and why are you helping them and showing that you actually have something that could benefit them? Okay, so just as a quick example here, I wanted to To do a little autopsy of 2 emails. So the email on the left, I would argue this one is illustrative, but I would argue this would go immediately to spam and we'll break that down. The email on the right, this was actually an email that I sent for our own cold outbound. So for selling Aveena, our platform. So we use our own platform to run outbound. This is a real email that went out to a startup that sold commission software for home services companies like HVAC, plumbing, and electrical. So do a quick read of the one on the left. Hi Jordan, I came across Glow Ritual and saw you make a clean dermatologist-tested skincare line for Gen Z shoppers. And, and our AI platform helps businesses generate more leads. Would you be open to a quick 15-minute call? The one on the right, hey James, we figured out how to find every HVAC, plumbing, electrical, and cleaning business running Housecall Pro with 3 to 30 techs and refresh that list on a daily basis. So you can see the one on the left is, first of all, it's targeting a B2C brand, so they can't even use our product. Second of all, it's It's saying it's complimenting them on a LinkedIn post that has nothing to do with the value we're providing at all. And, and third, it's a generic pitch. And by the way, this recipient actually turned into a customer. They responded within 5 minutes, and that became a sale. So, so to break this one down, you know, we targeted it to the right companies, this commission tracking startup for home services businesses, and they just hired their first sales rep. So that was part of the signal that went into generating this outreach, and that indicated that they're growing, scaling, and looking potentially for ways to be more efficient. And then third, it was customized to them. So this gave them pretty much exactly what they wanted, right? A way to find all of their exact target customers, a way to reach out to them. And it spoke directly to them. Okay, I'm gonna try to speed up a little bit because I see I'm a little short on time. But this is kind of the conclusion here, right? Buyers don't reject automation, they reject irrelevance. And if you take away anything from this talk today, this is what I want you to take away. It's the relevance of the outreach. Does it actually solve Someone's pain right now that matters most. And this is— so what you're seeing here is responses to some of the outbound that we sent. And, you know, when people receive a relevant email, they're actually delighted to receive it because it's, you know, going to help them out.

Okay.

So what is the actual ideal combination of technology And humans. This is a view of kind of the AI-enabled SDR, right? It's not an AI SDR, it's an AI-enabled SDR. And now each of these 7 steps, we've got human judgment plus AI for execution. And just run through these. So targeting, you've got the human who makes those decisions on who to target. The AI is the, the system that Goes and finds those companies, taking away all of that grunt work. Timing— the human strategizes around what indicators might indicate that a prospect is feeling pain for their product or service. The AI can help them brainstorm those ideas, and it can also actively monitor for those companies and monitor for those timing signals.

Persona.

The human strategizes around who to reach out to. The AI goes and finds that contact info and populates the lead list. Research, the human decides what's important to know to craft the right pitch. The AI goes and physically gets all that information. The pitch itself, the human crafts sequences that are gonna speak to that customer and the AI For every individual recipient, it's gonna select the right pitch and tailor the message for that person. So the human defines the parameters around it, but the AI is, is going to execute that for thousands of recipients much more efficiently than the, the person could. Sending, you know, again, that's, that's automation. That's all good. And then reply handling. The, the person is, is going to be working that lead and getting the meeting. The AI can help them triage those replies and book and do the physical booking. Okay. And a wise man once told me, technology doesn't do the thinking for you. It's an execution tool. So that's, that's what, what I would like to leave with you today. And I'll go ahead and move to the demo. Okay, so increase the size of my screen here a little bit. So as I mentioned, Avena is really designed to help you with a lot of those steps. And just grabbing a sentence here. So, okay, so Avena is really designed to help you with many of those 7 steps and be the AI execution partner for the human. And first, we'll start with targeting and timing. So those are the first 2 steps, and I'll show you how Avina can help you execute on both of those. So as an example, what, what you do in Avina is essentially ask for exactly the kinds of companies that you want to sell to, and our AI will go and find those companies. And you can describe ICP criteria as specific as, you know, what you're seeing on the screen here, and even timing signals as well that indicate that business is ready for a conversation. So for us, the example that I want to walk through is actually that same one that I showed you on the slide there. We're going to look for Businesses that integrate with Housecall Pro. So that's the home services, home services industry, and have less than $100 million in revenue. It's a qualification criteria and have an open SDR or equivalent role. So this is a timing signal, right? They have an indicator that they need AI selling software right now. I'm just going to submit that search. Okay, so Aveena cleans that up a little bit, cleans up the search criteria, and then what we've created here is essentially a dynamic audience. So we've defined this search criteria. Aveena's AI agents are going to go find those businesses for us, and we're able to preview the results here so we can iterate on this and make sure that our search is bringing good results. We can add or remove criteria, and you can see the results coming through here. So again, this is targeting and timing, right? But I've brought my own judgment into this. I know what kinds of companies need my product or services, and this— the AI is now executing that search for me and saving me loss of time and effort to actually find the right perfect leads. So this all happened in real time. It's finding these companies that matched all this criteria. And how this works is now I basically tell it, this is the next step, right? The personas. Step 3, who do I want— actually want to reach out to in these businesses? So for our account, we reach out to the executive, go-to-market leadership, RevOps and sales roles, sometimes the head of marketing and the head of sales leadership as well. So this is all configured in the settings. For each of these personas, we've set up around 10+ different titles. That the system will automatically look for. And then we essentially tell this how many leads we want per day, and this becomes a dynamic audience that keeps refreshing every day. And so anytime a new business posts job postings and become qualified into these criteria, the system will find them. Okay, and just to show you what those leads actually look like, I'll switch over to an audience that We've already been running for a while. But so you can see the leads here. So it's, it's actually gone and enriched the contact info of all of these folks matching the personas that we decided. So targeting, timing, and personas. Now we've got a clean list of businesses that match all of our criteria and are primed to be reached out to. Okay, what's the next step? It's research. So moving on to the research and pitch side of things. So this is again the exact sequence that you saw on the slide earlier that we've been sending and closing deals off of. So within Aveena, you create these sequences. Here, I'll just back up. Really quick to the sequence creator. Essentially, you come in here, you simply create the sequence, you pick the signals audience that you want to trigger the sequence. So we just created this one, Housecall Pro users with SDR openings. We pick the the actual sequence sending platform. So we integrate with Instantly, Lemlist, SmartLead, HubSpot, Apollo, and Outreach, and, and just create the sequence there. And so that gets you to this screen where you can define all the sequence steps. And what you're seeing here is the definitions and rules for how to actually get the AI in the system to send the perfect email or LinkedIn message to these folks. So, and notice again that we have a reference template here. So this is the basis for, for our instructions for the AI to write this email. As you can see, we've written out actually most of it ourselves. It's not a fully AI-generated email. The AI is not actually writing every word. And that's a little secret for you is that, you know, we've tried AI-generated messages. They actually sound very AI-generated. So we've— so again, we're bringing the art part of it into the AI system. And you can use AI, you know, we use AI, we have a ton of skills to actually craft these for us more quickly. But we're bringing the knowledge of how to speak to our customers into the AI system for it to execute, and we're allowing it to have some freedom to customize the, the email to the actual recipient. And we can add a bunch of rules. So this is the research part, right? In the actual guidelines here, we can give the AI rules on how to research what information to pull about the recipient that it can use to fill in the, the email details. And then we have instructions on how to fill in those actual details. And this can go as deep as we have a corpus of knowledge for the AI to use as it writes those emails. So this is an example of a document that's attached to every email that just tells it how to pitch Aveena to different titles. So if it happens to be emailing a founder, it's going to tailor the pitch for that. If it happens to be emailing ahead of SCR, it's going to tailor the pitch for that. So there's a very, you know, a lot of power and input that you can provide into the system here. And once we turn this on, essentially every lead that comes through from that signal audience goes through the system. And these emails are generated and fired out. So that's essentially steps 1 through 6. And yeah, so just jumping back to the presentation here, I want to just conclude with a little preview here of What we're excited about, where we're actually going next, which is, you know, we have the system where the humans are bringing the art and the judgment into the AI execution tool. And we want to make that even more efficient, right? So we're giving the humans— when we do this for our customers, we have a white glove tier where we handle every single aspect of this for them. And, you know, we have humans on our end that are bringing that judgment and art to it, but we also give our humans a very powerful toolset to do all those things. And part of that toolset is this essentially Claude Coworkspace, if any of you have used that. But it's a collection of skills that essentially creates documents for the target market, the brand voice, the how-to-pitch docs, all of that. And this allows us to work extremely efficiently and actually tackle a lot of those human judgment art pieces, but, you know, AI-assisted. And eventually we're going to take away even more work and hopefully allow the humans to focus even more on the high-level art piece. And what we're really excited about is essentially bringing in forward-deployed GTM engineers. So it's a little jargony, but it is really where we believe things are going, which is you have the human judgment again, but someone who's incredibly AI-enabled, solving problems for every individual business. So yeah, that's the end of the talk, and hope you took something away from it.

Thank you so much. I mean, this is— it's exciting. I love how you're taking the human approach and adding it in. You know, I noticed there's so many new AI filters on email, right? So the emails don't even get through. And so it's really smart to actually come in and create some of those templates and showcase that it's not AI because you're right, you know, sometimes the AI generator really does sound like it is AI.

Yeah, for sure. And, you know, we think that the future is, You know, email filters that are going to intelligently decide which emails to let through, right? And it's going to be the emails that are actually relevant that get let through. And so, so having, yeah, being in the, being in the top 1% of, this is actually something that is very well researched and you should care about, will make the difference.

Yeah, what I like is that you can set filters for your search criteria. For example, I was able to tune into, let's say, technology companies that offer sales technology that recently have received funding, and it immediately shows those companies, and it saves so much time because you can really dial into your ICP and, um, all the, you know, the variables that, um, come at play. Uh, what level do you want to, uh, uh, connect with? Or maybe you can, uh, even filter the 3, uh, best decision makers that you want to reach out to. So you build in a lot of flexibility and, um, And the other thing that I like is once you let this run, um, it continues to search afterwards. So even if you have completed a campaign, uh, you can go back and check that filter, and then it finds companies that just have gotten funded like yesterday.

Exactly. Yeah, the idea is that you create these criteria and it's always on, you know, and that's the power of AI that hopefully will save us from having to do a lot of the repetitive stuff that, you know, reps were doing every day and, you know, just looking for new businesses and, yeah, just doing a lot of that manual kind of list building work.

Well, thank you for taking the time to share the demo and also share your perspective. And what's refreshing in all of this is that, uh, you put human judgment at the top, um, so that, um, people can execute faster with AI, but, uh, they can guide the process, and, uh, failure is not an option.

That's right. Yeah. And trust me, we've tried to do a lot of this stuff fully powered by AI as well. And the results just aren't as good as when we inject a little bit of art into it.

Well, I suggest that anybody who would like to learn more go to avena.io. And there are a few guides that you can find on their website. About agentic prospecting, and I recommend that you educate yourself and test it out. And they also have an interesting blog about how to use intent data for outbound. So there's lots to learn. And I want to thank Mike for being the pioneer in that field and showing us what is possible and the possibilities for the future. wide open because you strike me as an entrepreneur that never stops innovating. So thank you for that.

Appreciate the kind words. Yeah, thanks for having me. Thanks, Gerard and Stephanie, for hosting.