| GUEST PROFILE Mike Cohen, Founder, Wayne Technologies, also known as the Batman Recruiter. Connect: LinkedIn |
This episode was hosted by Amrit Acharya, Co-Founder and COO of Xobin.
Table of Contents
TL;DR – Key Takeaways!
- ChatGPT is not replacing recruiting. It is replacing the tactical, time-consuming parts of recruiting and freeing up time for the parts that actually require human judgement.
- Mike used ChatGPT plus Clay to build hyper-personalized outreach emails. His first-email response rate jumped from 7-9% to 19-36%, depending on the role.
- Most “AI in recruiting” tools are not actually AI. They are machine learning algorithms that learn from yes/no inputs. When evaluating tools, it’s important to understand the differences.
- AI bias is real, but it is not more real than the unconscious bias already in your own job descriptions. The right response is a rigorous review process for both.
- If Mike started his career today, then learning to code would be the one thing he would choose to do differently.
Mike Cohen answers his LinkedIn messages as Batman Recruiter; his URL is literally linkedin.com/in/batmanrecruiter. He has been an agency-side sourcing specialist since he was 21; he runs his own firm called Wayne Technologies, and he is one of the most candid voices on AI in recruiting you will find.
In episode #12 of Xobin Talks, Amrit Acharya, Co-Founder and COO of Xobin, sits down with Mike to discuss what ChatGPT has really changed in sourcing just seven months after its launch. The conversation covers practical use cases, actual results, clear limitations, and the big question many recruiters are quietly asking: “Will AI eventually replace recruitment jobs?”
“Wait – Why Is Your LinkedIn URL Batman Recruiter?”
Amrit can’t stop himself before any of the AI talks.
Amrit: “Your LinkedIn profile is linkedin.com/batmanrecruiter. What’s the story?”
Mike: “When I was working in New York City, I began exploring the idea of creating my own personal brand. Donald Trump launched his presidential campaign about the same time. Back then, his lawyer, Michael Cohen, was all over the news because of the way he treated people during interviews. Since he was already dominating search results, I quickly realized there was no chance I’d ever rank on the first page of Google with the same name.”
So he thought about Batman.
Mike: “I didn’t grow up reading comic books. However, when I picked up Batman: The Long Halloween at 19, everything changed. I instantly loved the contrast between Batman, Superman, and Wonder Woman, and the way their personalities balanced each other out really drew me in. Superman and Wonder Woman disguise themselves as ordinary humans so they can fit in. But Batman? Bruce Wayne is the disguise. The human is the mask, not the hero. And Batman was one of the world’s first great detectives. Detective Comics – DC Comics – literally stands for Detective Comics. And I thought: I’m kind of like a detective. Let’s see how this goes.”
Eight and a half years later, it stuck.
“ChatGPT Has Been Out for Seven Months. Has Anything Actually Changed?”
Amrit: “It’s June 2023 today, seven months after ChatGPT launched in November 2022. Has anything meaningfully shifted in sourcing and recruiting?”
Mike: “Such a loaded question. So GPT in general has been around for years as a large language model for developers to build off of. What I’m going to focus on is ChatGPT specifically, the site you can go to and interact with via a chat prompt. Has it changed anything? Yes. But I think the change is more of a paradigm shift than a tactical one.”
He describes two immediate tactical changes he has seen.
The first is the obvious one: drafting job descriptions, candidate messages, and outreach copy. Instead of writing from scratch, recruiters generate a first draft and edit from there.
Mike: “Even more significant than changing recruiting has been done. It’s helped people realize what technology can actually do. People are like, ‘Wait, I just typed a prompt, and it created a job description for a chemical engineer in Minnesota that actually sounds real?’ And honestly, yeah, it can.”
The second use case is where it gets genuinely interesting.
“Here Is the Use Case Nobody Is Talking About”
Mike: “I think ChatGPT is an amazing aggregator and summarizer of data. And I don’t hear people talking about this as much as I wish they would.”
He describes the workflow he built using ChatGPT alongside a tool called Clay, a spreadsheet platform with built-in data enrichments and APIs.
Mike: “In Clay, I use one column to track company news. Then, I keep separate columns for a candidate’s LinkedIn posts and LinkedIn summary. Along with that, I also store key details such as their name, current company, job title, location, and tenure at the company. That means I’m reviewing four different pieces of information for every single person. Even if you’re a fast reader—unlike me—it would still take close to two hours to read through data from 100 people. And that’s before you even start summarizing everything or writing a personalized outreach email for each candidate.”
ChatGPT reads it all and drafts a personalized email for each person, referencing their specific recent posts, their company news, and their background.
Mike: “It spits out a recruiting email that does not sound like me; it is not what I would normally write.” But I do not have to spend two hours going through and reading all these articles. It is already there for me, and all I have to do is spend maybe ten minutes rewording it into my own style.”
Amrit: “So you’re taking a typical three-hour workload down to ten minutes?”
Mike: “That’s right.”
“Okay, But Does It Actually Work? What Are the Numbers?”
Amrit pushes for proof, not just theory, but actual response rates.
Mike: “I have been doing this for every first email in my sequences for clients for the past three months. One of my clients is a security SaaS company in New York. Every first email for every role I have worked on for them has been this highly customized email, written by ChatGPT and then adjusted by me.”
He pulls up his data live on the call.
Mike: “Four different roles. Two executives and two non-executives. The overall response rate across the sequence is 51.14 percent. That is a cold outreach response rate.”
Amrit: “What was it before?”
Mike: “Our average score for the entire sequence stayed slightly below 30 percent. But let me get you the first-email numbers specifically. The government affairs role is an executive position and not an easy hire; the first email alone got a 36 percent reply rate. VP of Sales’s first email got 19 percent, and VP of Professional Services’s first email got 25.5 percent.”
For context, his previous first-email rates on roles without this approach were as follows: A full stack engineering role got 7 percent, and a product manager role got 9 percent.
Amrit: “You’ve more than tripled the response rate on the first email.”
Mike: “That’s right. And something that previously required doing very one-off things, you were not going to do that for 200 people a day; there was just no time. I can actually do it now.”
For reference, the industry benchmark for sourcing campaign reply rates is around 19.6% across the full sequence (Ashby Talent Trends Report, 2024). Mike’s first email alone is beating many teams’ full-sequence averages.
“But what is AI actually? Because Everyone’s Throwing That Word Around.”
Amrit raises the question his clients keep asking: with every recruiting tool claiming to use AI, how do you know what is real?
Mike: “People may have different opinions on this. Still, one thing becomes pretty obvious: most hiring platforms that claim to use AI are not actually built on genuine artificial intelligence. In reality, they are advanced machine learning systems that study yes-or-no responses over time and improve based on patterns they find. That does not make them actual AI. Still, the technology is impressive, and I am not taking away from that. But whenever I hear the word “AI,” I instantly feel like the term is being used more for hype than reality.”
He gives a more grounding framing.
Mike: “Google uses AI every time you run a search. You just don’t know it. LinkedIn uses its own AI. When you type ‘software engineer,’ it may return ‘software developers’ because it has learned those are associated. That is not a mistake. That is the algorithm working. The use case for AI is fluffy and overstated generally. When people say ‘we need AI,’ my question is: why? Tell me the business problem you are trying to solve, and then we can figure out if AI is the thing that can solve it. You do not adopt AI and then find a use case for it. That is backwards.”
Amrit: “I say this to my clients all the time. For certain problems, AI simply is not necessary because there is no real use case for it. However, since AI is the latest trend, everyone wants to be part of it and chase the next shiny thing.”
Mike: “It is human. We cannot control it.”
“Will AI Cost Recruiters Their Jobs? Be Honest.”
Amrit asks the question that most people are too afraid to bring up.
Mike: “Yes.”
A pause.
Mike: “Okay, scary answer. But hear me out. Yes, the same way that innovation causes job loss and job creation consistently, as you look back through the Agricultural Revolution and the Industrial Revolution. They all have the same effect: people lose jobs, and more jobs are created.”
But his real answer is more nuanced.
Mike: “Do you think a huge number of people could eventually lose their jobs? I’m not saying every single person, but maybe 95% to 99% of the workforce. Honestly, I don’t think so. The kind of work we do depends heavily on human judgment, real-world understanding, and years of experience. We spend years studying profiles, meeting people, and having conversations within the industry. Because of that, we develop instincts and insights that are hard to explain, let alone train into an AI system anytime soon.”
He does identify who is at risk.
Mike: “If all you are doing all day is messaging candidates, writing job descriptions, and scheduling interviews – yeah, maybe. Could a recruiting coordinator role, depending on their scope, be replaced? Yes, it could. So the question becomes: what are you doing to differentiate your skill set from something a computer could do? And if you do not know the answer to that question, you may want to spend some time figuring that out.”
“What Would You Tell Yourself If You Were Starting in 2023?”
Mike has been in recruiting since he was 21. Amrit asks: if he were walking in fresh today, what would he do differently?
Mike: “Learn how to code. Straight up. That is it.”
He does not dress it up.
Mike: “We spend hours doing tasks a certain way simply because that’s how we’ve always done them. However, many of these repetitive processes could easily be automated if you knew how to create a few simple scripts. I do not know how to code, so I have to use ChatGPT to do most of that now. If I had invested more time in that earlier, it would be very convenient.”
Amrit: “I come from a technical background, so I still rely on ChatGPT when I need help with coding. It saves time, and honestly, it makes the process a lot easier.”
Mike: “Oh, perfect. So it works both ways. From what I can tell, your requirements seem more advanced compared to mine.”
Amrit: “Slightly.”
“How Do You Handle the Bias Problem in AI?”
Amrit raises the issue that comes up in every serious AI-in-recruiting conversation: the tools carry bias. What does Mike actually do about it?
Mike: “People have made quite a distinct point about the bias in AI, and they are totally right. There is bias in AI. It is unconscious bias because it is written by humans, and humans have unconscious bias. But my response in return would be, “What are you doing about your own unconscious bias?” Because if you are really focused on the unconscious bias of this tool, I assume you are also focused on your own, in which case, what are you doing to alleviate that? Do the same thing.”
His Two Rules for Handling AI Bias
His practical approach has two rules.
Mike: “Number one: do not take the output from ChatGPT and run with it directly. Take that job description, read through it, have somebody else read through it, run it through a system like Textio or other tools that give you scores based on bias and readability. Do not take the output as gold.”
Mike: “Number two: keep in mind that no system is completely free from bias. ChatGPT is no different because it reflects unconscious bias in its own way. At the same time, it is not the ultimate source of bias either. Instead, it is simply another example of it. That’s why organizations need to address bias in a structured and consistent way. You cannot solve it just by saying, ‘I don’t like this, so I won’t use it.'”
He makes the comparison explicit.
Mike: “Let’s say you eliminate ChatGPT entirely and a human writes a job description alone. Okay. What are you doing about the unconscious bias in that? Because if the answer is “nothing,” then what really sets them apart? With ChatGPT, you are using somebody else’s unconscious bias. With a human alone, you are using your own. The process should be just as rigorous either way.”
“And One More Thing – Don’t Trust ChatGPT with Math.”
Before the conversation wraps, Mike has a warning.
Mike: “Quick random note: do not trust ChatGPT with math. It sounds weird because this thing can write JavaScript, explain moving averages, and help you understand basically anything conceptually. But ask it to add two numbers and square root them, and you have like a 50 percent chance it is going to give you the wrong answer.”
Amrit: “I asked it to divide six by three, and it gave me four.”
Mike: “Makes sense, right? It’s a large language model, not a calculator built for perfect math. It learns from huge amounts of public content across the internet, and honestly, a lot of people online aren’t great at math either. So naturally, the model picks up some of those mistakes and patterns along the way.”
“What Does Your Actual Tech Stack Look Like?”
Amrit ends by asking the practical question recruiters always want answered: what are the tools Mike actually uses every day?
Mike lists 16 platforms he has full enterprise access to, not just Chrome extensions:
Amazing Hiring, Chatter Works, Clay, Crunchbase Pro, Entello, Exact Buyer, HireEZ, Heartbeat, Horsefly, Human Predictions, Lead411, LinkedIn Recruiter, Loxo, Recruit’em, SeekOut, ZoomInfo Recruiter, and Perfect.
On the obvious question, do they all return the same candidates?
Mike: “When you looked up the same candidates on LinkedIn Recruiter and SeekOut, how much overlap did you expect to see in the results?”
Amrit: “Maybe around 60 percent?”
Mike: “To be honest, I haven’t seen the overlap surpass 40% in all the years I’ve been doing this. Most of the time, it stays below 10 percent. Plus, even when you add a third platform, the overlap barely increases.”
His reasoning: it is not which database you access; it is how you search through the data.
Mike: “HireEZ is very AI-driven; it has algorithms doing work in the background. SeekOut has the most deep-dive Boolean capabilities of any platform. LinkedIn Recruiter uses its own form of AI to associate what it thinks you want with what you are actually searching for. You wind up with completely different lists. I use a large swath of tooling intentionally to try and find the gems that other tools cannot find.”
His Favorites From the List
His single favorite? SeekOut.
Mike: “The reason is the deep-dive Boolean capabilities. I love Boolean. I love being in control of my search and knowing exactly what is going to come back to me. There is no other tool that does that as well.”
For email automation he uses SourceWhale, for contact info Swordfish, and for candidate screening either Honeit or SmartRank.
Mike: “Honeit helps recruiters produce richer and more genuine prospect submissions by capturing phone screen behavior. On the other hand, SmartRank works well when you want to give candidates a quick three-minute self-screening option instead of asking them to join a 30-minute call. This way, candidates can explore the opportunity at their own pace and decide whether the role feels like the right fit before investing more of their time.”
🎧 Watch the Full Episode
Xobin Talks – Episode 12 | Mike Cohen (Batman Recruiter), Founder, Wayne Technologies | Hosted by Amrit Acharya, Co-Founder and COO, Xobin
▶ Play Episode #12 of Xobin Talks
About Mike Cohen
Mike Cohen founded Wayne Technologies and stepped into agency recruiting at just 21. Over the years, he has built deep expertise in sourcing and spent the past six years running his own sourcing practice, helping companies from different industries find the right talent. Along the way, he has maintained enterprise-level access to 16 sourcing platforms and regularly shares insights on recruiting technology, AI, and hiring workflows.
Because of his practical and hands-on approach to AI in talent acquisition, many professionals see him as one of the most grounded voices in the space. Online, many people also know him as Batman Recruiter.
Connect with Mike on LinkedIn | Company: Wayne Technologies
Looking for more insights from Xobin Talks? Explore all the episodes.
Frequently Asked Questions
What is the best way to use ChatGPT in recruiting?
Instead of using it only to draft emails, use it as a smart data aggregator and email personalization tool. You can feed in candidate LinkedIn posts, company news updates, and profile summaries through platforms like Clay. Then, let it create highly personalized outreach emails for every prospect. Because of this approach, Mike improved his first-email response rates from just 7–9% to an impressive 19–36%.
Is most “AI in recruiting” actually AI?
No. Most recruiting tools using the AI label are actually machine learning algorithms that learn from yes/no inputs over time. That is powerful and useful, but it is not artificial intelligence. Knowing the difference helps you evaluate tools honestly rather than getting caught up in marketing language.
Will AI replace recruiters?
Not most of them. The intuitive judgment built from years of experience, reading people, and understanding organizational context is not something current AI can replicate. Recruiting coordination roles with narrow, automatable scope face higher risk. The right question is, “What in your current role could a computer do, and what would be left?”
How do you handle AI bias in job descriptions?
Never use AI output directly. Read it, have a second person read it, and run it through a bias-scoring tool like Textio. More importantly, apply the same rigor to human-written job descriptions. The unconscious bias problem exists either way; the review process should be consistent regardless of whether a human or an AI wrote the first draft.
How much candidate overlap is there across sourcing tools?
Less than you think. Mike has never seen crossover above 40% between any two platforms, and typically it sits below 10%. The reason is not the underlying data; it is the search methodology. Each tool accesses profiles differently, so the same search returns meaningfully different candidates.
What is Mike’s favorite sourcing tool and why?
SeekOut, for its deep Boolean search capabilities. When you need to find very specific combinations of skills, experience, and background, no other platform lets you construct and control a search to the same depth. He also rates Clay highly for data enrichment and email personalization workflows.
Should recruiters learn to code?
Yes, if they are early in their career. Knowing how to write basic scripts automates the repetitive, time-consuming parts of a sourcing workflow – column organization, deduplication, and data formatting. Mike did not learn to code early and now relies on ChatGPT to write the scripts he needs. He wishes he had built that skill earlier.
Can ChatGPT do math?
No. Or at least, not reliably. It is a large language model trained on internet text which includes a lot of people getting math wrong. It can explain complex mathematical concepts beautifully and even write out the steps of a calculation correctly, then give you the wrong answer at the end. Do not use it for anything where the numbers have to be right.
