EP 011 - AI is likely to steal Recruiter Jobs

00:00:00: Hello and welcome to this week's episode of the AI Boardroom.

00:00:05: Today with a special guest and to tell you more about it, Svetlana.

00:00:10: Go ahead.

00:00:11: Yeah.

00:00:11: Hi, guys.

00:00:12: Welcome to this episode.

00:00:14: Really excited to have another guest on our show.

00:00:17: Mike joins us to talk about some fun stuff in recruiting specifically.

00:00:22: I think we're really excited to actually bring you tangible examples for how AI

00:00:28: could be applied in specific parts of the industries across business verticals.

00:00:33: And Mike comes from a vast amount of background in recruiting, which he'll talk

00:00:38: about in a second.

00:00:39: And...

00:00:39: So we'll start at a high level, just talking about Mike, what he is, some

00:00:45: exciting things he's been doing in the field.

00:00:47: He's published a book, just a quick little preview that he'll talk about and some

00:00:51: tools.

00:00:52: But then we'll get his expertise opinion about where the industry is headed with

00:00:57: AI.

00:00:58: And yeah, there's going to be some fun things at the end.

00:01:01: So stick around.

00:01:02: And should we dive in?

00:01:03: Mike, do you want to jump in and give us an introduction on who you are and what

00:01:08: brought you into the recruiting in the first place.

00:01:11: Yeah, thank you so much for having me and thank you for that kind introduction.

00:01:15: My name is Mike Wolford.

00:01:16: I spent about 20 years in talent acquisition.

00:01:20: Like everybody, as a young child, I grew up dreaming of being a recruiter.

00:01:25: Like everybody else, I fell into the profession, but I really enjoyed it early

00:01:29: on.

00:01:30: I got to help people find a job and that wasn't necessarily the point of what we

00:01:34: did, but at the end of the day, my candidate had a better job, my client had

00:01:39: the person they needed, and I got to get paid.

00:01:41: And I was like, everybody wins here.

00:01:42: I feel really good about doing this for a living.

00:01:45: So I enjoyed the profession early on and, uh, I started to be a writer and

00:01:50: contributor probably about eight years into the profession.

00:01:53: I've spent the last 12 years writing for recruiting daily and source con, which are

00:01:58: two publications in my field.

00:02:01: Uh, and speaking on topics and recruiting, but not super intensely, like once or

00:02:06: twice a year, a couple of art, you know, maybe eight to 12 articles a month or a

00:02:10: year wasn't.

00:02:11: an intense focus of mine until I came into AI.

00:02:16: After I left Twitter, I got into analytics and artificial intelligence.

00:02:20: And I started to watch an expert, his name is Dr.

00:02:24: Alan Thompson, he's in Australia.

00:02:26: He was the head of Mensa International.

00:02:28: He is a PhD in researches area of expertise is human intelligence.

00:02:34: And about three years ago, he switched to artificial intelligence and I came across

00:02:38: his interviews with the AI.

00:02:40: And he was automating his interview with a AI avatar.

00:02:45: So it looked like he was actually having conversations with the AI and I thought it

00:02:49: was brilliant, but I also thought it was brilliant.

00:02:52: Like, wow, the AI understands and can respond like appropriately.

00:02:57: Um, and sometimes it was, you know, out of line and he would call it be like, that's

00:03:01: not true.

00:03:01: We're like, it was melancholic or overly negative.

00:03:05: And he would call that out.

00:03:06: And I thought that was really interesting too.

00:03:08: Like he wasn't trying to hype it.

00:03:09: He wasn't trying to downplay it.

00:03:11: He was just trying to show it for what it was.

00:03:14: And, but I, but I understood then that that was a aha moment for me.

00:03:17: Like it's conversational now.

00:03:19: Um, it may not be perfect, but it's conversational.

00:03:23: And by the time GPT four came out, I had interviewed it myself and done a similar

00:03:29: thing.

00:03:30: And it was competent in answering questions about the field that it had

00:03:36: never been trained on.

00:03:37: And so that's when I got the idea.

00:03:39: I was like, well, let's see how far I could push it.

00:03:41: And I tried to get the book and I've written two books before full disclosure.

00:03:46: I do have ADD.

00:03:47: It does take me a long time to get writing done.

00:03:50: And there's a lot of typos and it's a process.

00:03:53: My editors are champions.

00:03:54: They earn their pay.

00:03:55: But I wrote the book in just over two weeks from blank page to sending it to my

00:04:02: publisher and the editor.

00:04:05: came back like a day or two after submittal and was like, this is a waste of

00:04:09: my time.

00:04:10: Like the only thing is wrong is maybe I kept the prompts that I had to write the

00:04:14: book in the book itself.

00:04:15: So there's like typos and grammar mistakes in my prompting of what to tell the AI to

00:04:21: do.

00:04:21: But in the responses from the AI, there wasn't a single grammatical error.

00:04:25: There wasn't a typo, there wasn't a spelling mistake, nothing.

00:04:29: 74 ,000 words perfect, right?

00:04:31: Like that's impressive.

00:04:33: book about how to prompt the system, how to write?

00:04:37: So it was intentional that you've included the prompt, so it wasn't an oversight.

00:04:41: That was an aha moment for me too.

00:04:43: It was like, you know, what's really valuable is not the fact that I could do

00:04:47: it.

00:04:47: It's the fact that I could teach other people how to do it and then they could

00:04:50: take this and do their work with it.

00:04:53: That was the second, like that was the real value to me.

00:04:56: It wasn't just like, hey, I can get AI to write content.

00:04:59: a meta book, right?

00:05:00: Like, write about what you're writing, basically.

00:05:04: How do you use it to get all of these different?

00:05:07: And the thing is, I'm an old recruiter and I'm a sorcerer, and I have been writing

00:05:11: about spam.

00:05:11: Our profession is notorious for just spamming people with jobs, right?

00:05:16: Doesn't relevant.

00:05:17: And I'm like, please don't just use this to automate your spam.

00:05:20: Elevate, don't automate.

00:05:22: Because now, a lot of us come from different fields and professions where we

00:05:26: were very creative types.

00:05:28: And now you don't have to just sit there and write an email saying, oh, I love your

00:05:31: resume.

00:05:32: You could say,

00:05:32: Hey, write this like a country song and add some emojis.

00:05:35: When I'm talking to a software engineer, put some code in there that would make a

00:05:41: Python developer giggle, like an inside joke for them.

00:05:44: Use industry lingo to reach out to these people.

00:05:47: Like customization.

00:05:48: I remember when I was in college, my professor said something oddly profound.

00:05:52: It stuck with me.

00:05:52: He said, the 20th century was defined by mass production.

00:05:56: The 21st century will be defined by mass customization.

00:06:00: And like that is to me, what AI actually is, it's a tool that allows for personal

00:06:05: and mass customization.

00:06:06: That's part of why it has such an impact, but also for context.

00:06:11: And I know we're just getting started here, so I'm going to end with this, but

00:06:14: for context.

00:06:14: So we understand about the volume of what we're really talking about.

00:06:18: If you take everything in human history that was written in 2022, all the way back

00:06:22: to the very beginning of time in 2023.

00:06:26: the AI not only wrote more, it wrote 20 times more content than has ever been

00:06:32: created.

00:06:33: So this is the speed at which we are moving now.

00:06:37: And...

00:06:38: there's a lot in there.

00:06:40: Let me start first with something I was seeing this week.

00:06:47: I think it was released two weeks ago.

00:06:49: It was an AI coding interview with an avatar.

00:06:55: And first of all, the questions were hard.

00:06:58: I was like, I'm a senior developer.

00:07:01: I should be better at this.

00:07:04: But even any other senior developers I've seen on YouTube were like, okay, that's

00:07:09: really hard.

00:07:10: But a lot of questions also got like, they lacked context.

00:07:14: Like for example, like what would you do to assume the best possible database?

00:07:19: But there was no info what kind of data you actually have, like that kind of

00:07:25: stuff.

00:07:25: So how did you approach your solution?

00:07:29: Because you have like built basically,

00:07:32: effectively use systems for recruiting, right?

00:07:35: How did you approach the solution?

00:07:37: So this is to me somebody who's an engineer but not a practitioner building a

00:07:41: solution because it gave the programming but not the examples.

00:07:46: And an example would say what type of data set would qualify it.

00:07:50: And that's actually the missing key for most of the tools right now.

00:07:54: They're built by engineers who know how to set instructions to a machine but they've

00:07:59: never had the capability maybe or the necessity to also combine those

00:08:04: instructions with examples.

00:08:06: And it's the two in combination that actually end up giving you the

00:08:10: consistently good results.

00:08:12: When you have a system, it's really frustrating because I don't know if

00:08:16: anybody has read Pippi Longstockings or something like that before, but it's like

00:08:19: completely literal.

00:08:21: And so you do it and you're like, it did exactly what I said, but not exactly what

00:08:25: I wanted.

00:08:26: Do you know what I mean?

00:08:27: And it's, and that's the place where examples come in because that's where you

00:08:32: say, this is what I mean.

00:08:34: by it.

00:08:34: And when you have those two combined, that's the solution.

00:08:39: And I think what you're talking about just to bridge the gap for the audience is the

00:08:43: intent.

00:08:44: And so capturing intent behind user queries or any systems, honestly, I think

00:08:49: in AI, it's hard to do.

00:08:52: And then there's a lot of work previously done in natural language processing.

00:08:57: I have some experience building tools as well, where capturing intent is actually

00:09:03: really, really hard because you need contacts, you need to understand where the

00:09:06: person is navigating from.

00:09:08: you need common sense.

00:09:10: And you need common sense.

00:09:12: I think that's also, yeah.

00:09:14: No, no, yeah, that's an important part of the point, sorry.

00:09:18: Yeah.

00:09:19: No, I think it's an important point.

00:09:22: And I think it's understanding where they are in their journey, too.

00:09:26: You know, kind of navigating because if they're just kind of logging into a system

00:09:30: and they're trying to get a response about something, or there are three queries in,

00:09:35: or they're like deep into their workflow, they need, again, common sense to also

00:09:40: needs to be part of that.

00:09:42: But contextualization is really hard.

00:09:44: to implement.

00:09:45: And I think, again, as you mentioned, just kind of connecting with what Mike said,

00:09:49: examples are so crucial to provide context, but then also to provide common

00:09:54: sense as to like, okay, I know you're going to take my literal instructions and

00:09:58: follow them, but this is actually what my intent is and what my outcome needs to be

00:10:03: based on these examples.

00:10:04: And I love that you kind of are bringing those things into the conversation.

00:10:08: Thanks.

00:10:09: Yeah, that's a missing piece of the puzzle right now.

00:10:13: And this is the interesting thing.

00:10:15: They've kind of turned us loose and there's no instruction manual, right?

00:10:18: There's no like, here's what you can use it for and here's how you get it to do

00:10:22: like everything else we've, right, exactly.

00:10:25: one has.

00:10:27: We have approximations and we can get near to a consistent system that we are not

00:10:34: there, maybe will not get because we don't get what the AI is actually doing.

00:10:39: So.

00:10:39: And somebody explained it to me, like, what if dogs invented us to take care of

00:10:43: them?

00:10:44: Like, they would not understand what a job is.

00:10:48: They wouldn't understand what money is or how we get it to feed them.

00:10:51: But we, they just know that food shows up, right?

00:10:54: And that's kind of what AI is doing to us.

00:10:57: We really were like, to me, the other analogy I like is like, we're moving from

00:11:01: the digital stone age to the digital bronze age.

00:11:04: Mm -hmm.

00:11:05: our ancestors used fire.

00:11:07: They didn't totally understand it.

00:11:09: They didn't understand chemical reactions or energy or any of that, but they, they

00:11:12: used it.

00:11:13: And it's the same thing with us in AI.

00:11:15: Like we're, we're like digital fire.

00:11:17: Like AI is digital fire.

00:11:19: We don't understand it, but that doesn't mean we can't use it, but it's also

00:11:22: dangerous, right?

00:11:23: Fire is also dangerous.

00:11:24: We have a long history with fire for good and bad, right?

00:11:28: And AI is the same sort of thing, except this is the fire that can light itself.

00:11:33: Oh, that's, yeah.

00:11:34: press that writes the books.

00:11:36: It doesn't just print them.

00:11:37: think that's something for another discussion because I think we have still a

00:11:41: lot of things we can take from the AI recruiting space.

00:11:47: So I would love to reiterate my previous questions.

00:11:52: Where did you get started?

00:11:56: I think you're told about the professor you met.

00:12:01: But where did you actively in your day -to -day use started using AI?

00:12:06: to do active recruiting challenges.

00:12:10: How did you get there?

00:12:12: I started.

00:12:12: layer on the question that's like, cause a lot of the listeners would probably be

00:12:16: like, what problems exist in the recruiting space?

00:12:19: And then what problems were you trying to solve with AI?

00:12:22: I think those would actually also spark ideas as to, okay, I incur that in my day

00:12:27: to day where I've heard that happen.

00:12:29: Like that's an awesome use case.

00:12:30: So, and I'm really excited to hear about like the tool that you've developed

00:12:33: because you've showed it, you showcased it to us and we're like, oh,

00:12:37: not all the listeners would be able to see it, but if you can walk us through, again,

00:12:41: the problems that you were trying to solve and just kind of talk us through kind of

00:12:45: the reasoning and what you've come up with, because that, I think, people will

00:12:49: benefit from hearing it.

00:12:50: Thanks.

00:12:52: So for me, when I got into AI to start, I was afraid of it making decisions.

00:12:57: I didn't want to start there, but I wanted, I came across some research that

00:13:01: said, you know, 30 out of 40 hours of a recruiter's week is actually spent doing

00:13:05: administrative work.

00:13:06: And I was like, I believe that I was a recruiter.

00:13:08: I believe that.

00:13:09: And I started to look at how I could get AI to do that type of work for me.

00:13:14: Like, how could I get it to do the type of work that was eating up all my time with

00:13:18: administrative work?

00:13:20: So I could free up the time to do the value add work that I want, because if

00:13:23: you're a recruiter, one thing that you know is you get evaluated almost solely on

00:13:27: how many people you place.

00:13:29: And so all that administrative work is really just slowing you down between you

00:13:33: and your objective.

00:13:35: And that's your performance rating.

00:13:36: So it's really important if you can give back that time.

00:13:39: So I started looking about, you know, I've been trying to teach people to write

00:13:42: outbound emails for years to little effect.

00:13:47: Spam is still a problem.

00:13:49: But AI actually can be a solution here because one of the things that recruiters

00:13:55: are trying to do is they're trying to get enough people in enough time and they

00:13:58: don't have the mental time, the physical time to write 30 customized emails in a

00:14:03: row.

00:14:03: So in lieu of that, they send one email to a thousand people and hope it hits the

00:14:07: right people.

00:14:08: With AI, they can actually not only target better, but actually take the time to say,

00:14:13: hey, this person mentioned that they like.

00:14:16: Star Trek and their profile.

00:14:17: Let's mention that in the outreach email, right?

00:14:20: Like, and then that's all you have to do is give it the instruction and then the

00:14:24: creativity and the writing of the email, the AI takes over.

00:14:27: And instead of taking five minutes to write an email out and think about it and

00:14:31: how you want to do it, it's right there and it's done for you.

00:14:35: So to me, that was like, okay, this is a win, right?

00:14:37: Like I can actually get either more high quality out in the same time, or I can do

00:14:42: the same amount, get the same amount of results.

00:14:45: that I'm looking for in the end game, because I get a higher response rate from

00:14:48: custom outreach than I'm going to from spam.

00:14:51: And I could take that time difference and I can invest it in something else.

00:14:55: And to me, what AI is exciting about is from several perspectives.

00:15:01: One for me as a recruiter, I want that administrative work gone because the work

00:15:04: I like doing is being a counselor and an advisor and an advocate for my candidate.

00:15:10: And an advisor and a consultant to my business partner, my hiring partner,

00:15:14: that's the part of the job I like.

00:15:16: And so to me, that's the pitch of AI is the pitch is it can take away.

00:15:21: And what I was looking to do with it is to take away the low hanging fruit, the stuff

00:15:26: that's monotonous, the stuff that we have writing job descriptions, candidate

00:15:30: summaries, and, you know, trying to convince the hiring manager that this

00:15:33: person's a good fit for this job.

00:15:35: Right now I have the AI do it.

00:15:37: I take my interview notes, I take the resume, take the job description, and I

00:15:42: say, okay, based on this person's resume and my interview notes, create a table

00:15:46: with the must have requirements on one side, and then how the candidate's

00:15:50: experience lines up to it on the other.

00:15:52: And you're thinking, Mike, that's pretty obvious, that's not done today?

00:15:55: No, because as a recruiter, you'd have to do that 10 times a day.

00:15:58: And like, nobody has the time or the patience or the attention span to do 10 of

00:16:03: those.

00:16:03: But if you have an AI,

00:16:05: And it can do it, all you have to do is give it the instructions.

00:16:08: And then 10 seconds later, you get the output.

00:16:10: Anybody can do that.

00:16:12: And that's a better thing because another part of that process, most people aren't

00:16:16: privy to is the discussion between the recruiter and the hiring manager that goes

00:16:20: on.

00:16:20: The recruiter says, I like this candidate.

00:16:21: I want you to interview them.

00:16:22: The hiring manager says, no, they don't have the required skills.

00:16:25: Right.

00:16:25: And there's this back and forth.

00:16:27: Whereas it's a lot harder for the hiring manager to say no to an interview when the

00:16:32: recruiter says,

00:16:33: Here's all the requirements and here's how this candidate aligns to those

00:16:36: requirements.

00:16:37: This is exactly what you told me you needed and this person aligns.

00:16:41: Now it's much, much harder to say, no, I don't want to interview that person.

00:16:45: So it's better for the candidate, it's better for the recruiter, it's better for

00:16:48: everybody.

00:16:50: And I think, can I just summarize also, because I think we've talked previously

00:16:54: about the opportunity mapping.

00:16:56: I plan to do a little bit more maybe in the future episodes, deep dives maybe with

00:17:01: Edgar.

00:17:02: But you basically chose high benefit, low cost things that are repetitive in your

00:17:09: business.

00:17:10: If you tackle it, they're not as complex to actually implement, but they provide so

00:17:15: much value to the organization.

00:17:16: So the ROI is truly clear.

00:17:19: And I think what you're kind of even talking about just to bridge the gap to

00:17:24: the solution that you've built.

00:17:25: Cause if I'm not mistaken, you kind of did all of that kind of back and forth work

00:17:30: that currently happens between the conversations between the hiring manager,

00:17:35: the recruiter as to is this candidate even suited for this particular job?

00:17:41: And then you have basically a report that is customized or hyper -personalized

00:17:47: basically for the role.

00:17:48: and the candidate and their likelihood of being the right person for the job.

00:17:53: And so the decision is still a human decision, but I thought it was really

00:17:58: intriguing how you, and maybe you were going to go into that next, and I'm

00:18:02: stealing your thunder a little bit, but I thought it was really fascinating.

00:18:05: And I think I'm now connecting the dots and thank you for, yeah, I just want you

00:18:11: to kind of mention that it was really, really smart for you to pick these.

00:18:17: repetitive tasks and yeah, I think bridge the gap with a solution.

00:18:21: It sounds like you're able to roll out.

00:18:24: let me add to that.

00:18:26: I think one point I would love to point out is you chose an approach which is

00:18:33: deeply connected to language understanding.

00:18:37: And like a comparison is basically like language understanding, understand what

00:18:42: it's like.

00:18:42: Because you have a skillset on the one hand and you have a rhythm on the other

00:18:45: hand.

00:18:46: They are not like...

00:18:47: exactly aligned, like you won't find the same skills you need on the resume.

00:18:56: So I found it really good.

00:18:58: Like that's what I say, like if you have tasks that involve language understanding,

00:19:03: you're pretty much set to at least challenge it with AI.

00:19:08: And also the really practical approach, like really, really good work on that.

00:19:14: I really quite like it.

00:19:15: Yeah.

00:19:16: And that's where I, so I took the, what I learned from writing the book and what I

00:19:19: learned from building products commercially, like we built an AI analyst.

00:19:23: Um, and I put those together to create an agent for myself.

00:19:26: And so now that I'm speaking and doing presentations and I'm doing all this

00:19:30: writing, I now have my own AI agent who's programmed with my way of writing, given

00:19:36: examples of how I write and can now like, I wrote an article about three weeks ago

00:19:42: about spam.

00:19:44: Spam is dead, long live spam, right?

00:19:47: Where I was trying to get back to the community about like, hey, now you have

00:19:50: this new capability, please don't just make the problem worse.

00:19:53: And I believe please take like five seconds and then just gauge a little bit

00:19:58: of creativity.

00:19:59: And I think that's the other thing like I would ask across professions is don't just

00:20:03: take this as a chance to automate whatever repetitive task it is you have to do.

00:20:09: You're a creative human being.

00:20:11: You've been at...

00:20:12: whatever it is you're doing for a minute, you've probably thought there's a better

00:20:16: way to do this.

00:20:18: Now is your opportunity to try it, right?

00:20:21: At low risk to yourself.

00:20:23: And if it works, share it and be the hero, right?

00:20:27: Because this type of opportunity to disrupt so many different types of

00:20:31: business functions simultaneously is unlikely to come again.

00:20:35: So, you know, don't just use it and be like, oh, I can use it for A.

00:20:39: Be like, how can I make A even better?

00:20:41: better if I do it this way.

00:20:44: Because to your point, it understands.

00:20:46: If you give it examples and instructions, it'll get what you mean.

00:20:51: And I think I was just going to add, it takes skill even for me who uses AI on a

00:20:57: daily basis.

00:20:58: I use lots of these chatbot agents and other tools, but it takes skill to even

00:21:04: get started, but then also identify use cases for what AI could be suited for,

00:21:08: even these chat, chat GPT type of tools, because I even catch myself every now and

00:21:15: then and I'm like, oh, I need to like read through.

00:21:17: I will like, wait a minute.

00:21:19: Why am I doing this work now when I can upload this document and ask for it to

00:21:23: summarize?

00:21:23: I was just doing some research around a company and I pulled an annual report,

00:21:29: which I have an MBA, so I know how to read them and what information I'm looking for.

00:21:33: And I started flipping through the pages and it was 124.

00:21:37: And halfway through it, don't ask me why halfway, but I was like, why am I doing

00:21:42: this work?

00:21:44: At that point, I just downloaded the file, uploaded it into ChadGBT, but I feel

00:21:48: like...

00:21:48: It takes time for you to get accustomed to just trading tasks to chat GPT or AI.

00:21:58: So yeah, you may have free freed up time through these repetitive and automated

00:22:02: workflows.

00:22:03: But I think the best way to get started is to start delegating things that you would

00:22:08: commonly do on your own and ask the question, is this something that I could

00:22:13: delegate?

00:22:13: Again, there's privacy and other concerns that you have to keep in mind as well, but

00:22:17: like,

00:22:18: Is this something that I could just ask Chad GPT to do?

00:22:21: Yes or no?

00:22:22: I mean, like if you're not happy with the output, then just go do it on your own,

00:22:26: but start building that skill.

00:22:27: I feel like that takes a little bit of time for you to even identify the right

00:22:31: things to delegate to Chad GPT or other chatbots.

00:22:35: Yeah, that's part.

00:22:36: So I'm, I've been laid off of my job.

00:22:39: And so part of what I'm doing is I'm actually having a bootcamp where I'll

00:22:43: treat train people in my profession, how to build these tools for themselves,

00:22:47: because that's kind of the next skill is, yeah, you know how to use it.

00:22:51: And now you're getting good at prompting it to do your emails and write your

00:22:54: LinkedIn posts.

00:22:55: And that's cool.

00:22:56: But like to take it to the next level and really get the benefit of it.

00:23:00: You have to be able to understand the prompts and then be able to put them into

00:23:03: one of these GPTs.

00:23:05: And if you can do that, you create a really powerful tool for yourself.

00:23:08: Now in recruiting, what will happen is even more powerful because we have API

00:23:12: access to CRMs and ATS systems with candidate information.

00:23:17: And so in the not too distant future, you're going to have recruiters who have

00:23:21: an agent like mine, an AI recruiter agent.

00:23:24: connected to their applicant tracking system, connected to their client

00:23:28: relationship management system.

00:23:30: And they're gonna be able to pull the information directly out of that.

00:23:33: And that's when you're going to get the email that says, you know, remember when

00:23:37: we kept said we kept your name on file and if anything else came up, we'd call you,

00:23:43: surprise, something else came up, right?

00:23:45: Like that's always been a polite way to get rid of people, right?

00:23:48: Like, hey, we'll keep your name on file.

00:23:50: And if something else comes by, we'll call you, right?

00:23:52: someone slash something cared about what you were.

00:23:58: goes and is saved.

00:23:59: Companies do save it.

00:24:00: And the problem is today, like if you have a thousand candidates in there, the search

00:24:04: results will only bring you the top 100.

00:24:06: So the bottom non -none, it's not their fault.

00:24:08: They just can never access that.

00:24:10: Whereas if you API to a system like that, the AI can access all the records and it's

00:24:15: going to be able to go back.

00:24:16: It's going to find your old application.

00:24:18: It's going to find you on LinkedIn or wherever today.

00:24:21: update your information.

00:24:22: It's going to say, oh, is this much of a match?

00:24:23: Here's the information from last time you interviewed.

00:24:26: Put together an outreach email.

00:24:28: Hey, remember how we said we keep your information on file?

00:24:31: We did.

00:24:32: And we have a new position that might even be a better fit for you than the one you

00:24:36: interviewed for before.

00:24:37: And here's the details.

00:24:39: giving competitive advantage, right?

00:24:42: So, whilst others don't even bother looking in their CRM because it's like

00:24:47: maybe even just not clean and sorted and stuff like that.

00:24:51: And now you can even with unsorted unstructured data, you just try to use the

00:24:56: AI and it makes sense of itself.

00:24:58: So for me, it's exciting because I think one of like, it's one of the great

00:25:02: stresses in life is to look for a job, right?

00:25:04: Like it's up there with like dying and public speaking.

00:25:07: Like it's like a major life event.

00:25:09: It's a major trauma.

00:25:10: And I get it right now.

00:25:12: And it's the recruiters and sources who do the job.

00:25:15: They take a lot of pride in their work, but the fact of the matter is they're not

00:25:19: humanly capable of doing what's being asked of them.

00:25:22: There's no human recruiter that can conduct 500 interviews in a week.

00:25:26: It's just not possible and it's unfair to ask.

00:25:29: And so that's what they're up against.

00:25:31: They're really professionals who take a lot of pride in their work, who literally

00:25:35: have an impossible task placed in front of them.

00:25:38: And that's just the first impossible task that's placed in front of them, right?

00:25:41: Like imagine you walk into a room and somebody gives you a piece of paper and

00:25:46: they're like, you have to make a human being.

00:25:48: appear and do work, actual work based on me giving you this piece of paper.

00:25:53: Like go find a human, bye.

00:25:55: Good luck.

00:25:56: Like Legend of Zelda stuff, right?

00:25:58: Like that's a quest.

00:26:00: So like they take it personally when I say things like the AI will help because it

00:26:08: can talk to everybody.

00:26:10: Oh yeah.

00:26:11: And the common pushback, not only in recruiting, but everyone is it's not

00:26:14: creative.

00:26:15: It's not, it's not that smart.

00:26:16: It's not there yet.

00:26:18: Um, this is denials, not just a river in Egypt.

00:26:23: it is.

00:26:24: So I think that's kind of the deal because I just read a Medium article that was sent

00:26:30: to me in my email today talking about a new API which helps parsing documents for

00:26:37: better rack retrieval.

00:26:38: And I myself had just yesterday put in like four hours to improve our rack

00:26:43: pipeline.

00:26:45: So there is still...

00:26:48: a way to go.

00:26:49: So I'm also curious, honestly, to get your book and look at what you've done and how

00:26:55: you approach it because from just my experience, like processing documents and

00:27:01: like unstructured data is possible, but it's also not a small feat.

00:27:06: Right, right.

00:27:06: Yeah, this is the very, we're the very first iterations of this.

00:27:10: So like the very first phone screeners and video phone interview systems are coming

00:27:13: out today.

00:27:14: But the fact of the matter is a human being can't interview that many people.

00:27:18: It's just not possible.

00:27:19: And so it becomes unfair because a lot of people who would be really good fits for

00:27:24: the job never ever get seen.

00:27:27: You know, they never, it's never even seen.

00:27:30: And that's unfortunate.

00:27:31: And it's a bad experience for everybody.

00:27:33: It's inefficient for the company because they just lost it.

00:27:35: You know, they went out looking for that person.

00:27:37: That person actually came to them.

00:27:39: And because of their inefficiencies in sorting and other things and matching and

00:27:44: an inability to tell just from a resume, whether somebody's a match against a job

00:27:47: description, there's this impasse and it becomes this catch 22 problem.

00:27:51: And it's a huge problem and it costs billions of dollars a year in lost

00:27:56: productivity and frustration and all kinds of inefficiencies.

00:28:00: I mean, anybody from the logistics industry would have a heart attack looking

00:28:03: at the pipelines and recruiting like Six Sigma, oh, please forget it.

00:28:07: Like we're not even close.

00:28:09: We're not even a one Sigma.

00:28:10: So we have a long way to go in the profession, but the technology is here

00:28:15: today to automate recruiting.

00:28:17: You could take a phone screen.

00:28:18: You could take a video interview.

00:28:20: There's challenges on that though, because...

00:28:24: You know, when I talked to the testing companies, they say one of the dead

00:28:27: giveaways that somebody's using an AI is they score a hundred percent on the coding

00:28:30: test.

00:28:31: Is that good or bad?

00:28:33: Like, do you, you were testing to see what kind of coder you got, if they could

00:28:38: produce code that's a hundred percent passable on a test, isn't that what you're

00:28:41: really after?

00:28:42: Like, I don't know.

00:28:44: The industry hasn't decided, right.

00:28:46: The industry hasn't decided what the answer to that is yet.

00:28:49: Some companies are like, oh, that's the person we want.

00:28:52: They're innovative, they're forward thinking, they're using the tools, and

00:28:54: others are like, they're cheating.

00:28:56: just doing 100 % coding thing is what you want is how people, what you want is when

00:29:03: you hire a programmer as a problem solver.

00:29:05: So you don't want someone who writes 100 % code for that.

00:29:08: Okay, that's fine, but how does he behave if the code doesn't work 100 % and how

00:29:13: does he approach that?

00:29:15: And that's also something a lot of the AI coder questions which I mentioned earlier,

00:29:21: I had in there like,

00:29:22: And describe what your biggest problem was doing that and how did you solve it?

00:29:28: And so like, and that I get because like, no one needs a coding monkey.

00:29:33: So if you can write a hundred percent code, so good for you.

00:29:36: So, but what if the code breaks?

00:29:38: And that's also the thing with the local solutions oftentimes, like what if it

00:29:42: breaks?

00:29:43: Like, yeah.

00:29:44: is like next week I'll be or week after I'll be recording my own avatar.

00:29:49: And so, you know, what's prevent me from hooking it up to a software engineering

00:29:53: backend and sending my avatar out on interviews, right?

00:29:55: Like this is something.

00:29:58: So we have a lot to think about in the field when it comes to interviewing and

00:30:01: validating skills.

00:30:02: This is actually a big challenge in recruiting today.

00:30:05: And AI is actually not being used here, which is kind of ironic to be really

00:30:10: thoughtful on this particular challenge because.

00:30:13: it's going to become increasingly hard to validate skills over a phone interview or

00:30:17: a video interview when more and more people have access to an LLM that can take

00:30:22: the transcript and run a search and answer the question, right?

00:30:26: So we actually have, we're in a period right now where we don't really have a

00:30:30: good solution, but we need to rethink the interview process.

00:30:34: Can I ask maybe a last question before we part?

00:30:38: We're about time for a controversial one.

00:30:42: So I want you to make it a good one.

00:30:44: So do you think in your perspective, can a recruiter be fully automated or are there

00:30:52: still human tasks that could not be delegated to AI?

00:30:58: Right.

00:30:59: So the question to answer technically, the technical answer is yes.

00:31:04: Now here's the thing about my reading of the law.

00:31:07: You cannot reject a candidate with an AI.

00:31:10: GDPR is explicit, more and more candidate, but there's no law anywhere saying you

00:31:15: can't hire one with it.

00:31:18: So yes can come from an AI.

00:31:21: So I'm not saying that people want that.

00:31:23: Although if you're saying you're a technology company,

00:31:27: your AI recruiter extends you the offer that's us using, right?

00:31:31: So maybe that in the right market, that actually works.

00:31:35: Technically it is feasible to automate end to end recruiting today with AI systems

00:31:40: and to never have a recruiter talk to a human being and get them an offer.

00:31:44: That is technically possible.

00:31:47: Legally, it is challenging at the moment because you're going to have to justify

00:31:52: yes, no decisions.

00:31:54: However,

00:31:55: Most recruiters have very little authority over yes, no decision.

00:31:58: They're a gatekeeper.

00:31:59: They'll interview somebody and pass them on, but there's a ratio.

00:32:03: A good recruiter will have 80 % acceptance rates, which means 20 % of the people they

00:32:07: think should be interviewed don't get interviewed.

00:32:09: So they don't even have really decision -making ability in the process to begin

00:32:14: with to have it taken away from them.

00:32:17: This is something that is unnerving to them.

00:32:20: as a profession because then where's the value add?

00:32:23: And this is why I've been trying to get us to be data driven for so long is we have

00:32:27: to become the business advisors and say, well, unemployment is this rate and

00:32:31: salaries are this rate.

00:32:32: And here's what our competitors are paying, right?

00:32:35: And advise them because that's the information they won't have.

00:32:38: That's what they need to get from the recruiter.

00:32:40: They don't no longer going to need whether this person's technically qualified or

00:32:43: not.

00:32:43: The AI is better at qualifying candidates than recruiters are because

00:32:48: I was a tech recruiter.

00:32:49: I was never a software engineer.

00:32:51: The AI is going to be better at evaluating a software engineer than I am.

00:32:55: I think where the dangerous thing is, and Edgar probably has a lot more knowledge on

00:33:00: this from maybe a technical perspective, but these systems have been more recently

00:33:05: exploited more through kind of reverse prompting.

00:33:09: I don't know what the right terminology it is, but in the last couple of days, I also

00:33:14: read about OpenAI being exploited again and people calling them out like, oh,

00:33:19: Open AI is acting out funky again.

00:33:22: So I'm just kind of relating it back to here.

00:33:24: Could candidates also exploit these chatbots to negotiate themselves like an

00:33:29: unrealistic kind of salary or benefits with AI?

00:33:34: Because there's no human common sense, I think, Edgar, as you mentioned, to kind of

00:33:39: level set the expectations.

00:33:41: Yeah, decisioning will still, and that's where I think humans, the no call, I think

00:33:46: the human will still extend the offer.

00:33:48: I think that's still what most people want, but the decision to say no has to

00:33:52: still come from a human on any level on any of this for the moment.

00:33:56: So I don't think it's going to be the AI rejecting you at a stage or you could

00:34:01: reverse engineer it into giving you better benefits because it can't sign the

00:34:05: benefits package at the end of the day.

00:34:06: The offer letter will still be signed and approved by a human for now.

00:34:10: Five years from now, however, I wouldn't be surprised if that's like a like

00:34:16: Superman 3 .14 is them reverse hack, you know, like how we did the banks with the

00:34:21: pennies.

00:34:22: Like he'll do the GPTs and then he'll get the reverse.

00:34:24: Anyway, it'll be something like that.

00:34:25: Just the AI version of a super heist, but yeah.

00:34:30: all the systems, AI is not even like RabbitMQ with the large action model and

00:34:36: stuff like that.

00:34:38: That's not really what happens in the background.

00:34:41: It's text in, text out, and then the software has to work on it.

00:34:46: So if you don't know the underlying APIs and how they work, and you should do all

00:34:51: this stuff on the backend, please, developers out there, don't do it on the

00:34:55: front end.

00:34:57: And and and yeah having that said it's like you of course you can exploit it and

00:35:02: then the answer is broken and then the whole system breaks so that's Like that

00:35:08: won't get you far Yeah, so so if you use normal security standards as you should

00:35:16: and you also try to break it before you release it That should prevent a lot of

00:35:24: stuff I'm not saying that's not exploitable

00:35:26: Like Mike said, we shouldn't do the yes, no decision, like accepting an offer.

00:35:33: Or me, honestly, even sending the offer should already have some human

00:35:39: intervention in it.

00:35:40: Because as soon as there's money involved and then numbers, I think someone has to

00:35:46: make sense of it.

00:35:49: But it's also, I don't know, I think everything we do today as service,

00:35:55: as like mostly getting information, structure it and evaluating it will all be

00:36:01: replaced by AI.

00:36:03: And using a human for it will be just expensive path like for the rich and

00:36:07: wealthy.

00:36:08: Like today, if you go shopping to IKEA, like you barely get someone to explain you

00:36:13: anything.

00:36:14: Whilst if you go like an expensive store, you have like your personal assistant

00:36:20: explaining you everything about the sofa.

00:36:23: So...

00:36:24: I think that's kind of where this will go as well.

00:36:28: But cautious, like with every ISA system, be cautious.

00:36:32: Um, I actually ended on a high note here.

00:36:35: This is something because there are caution is wise.

00:36:38: Um, but I was just watching a Ted talk from Mustafa Suleman, who's the CEO of

00:36:43: Microsoft AI.

00:36:44: And this caught, I had to pause it, rewind it and catch this word for word.

00:36:49: Cause I thought this was amazing.

00:36:50: He said AI is to the mind.

00:36:52: What nuclear fission is to energy, limitless, abundant and world changing.

00:36:57: Yeah, it is.

00:37:00: I would love before I know we are like, like out of time, but I would love for for

00:37:06: you Mike to at least give us some some hints like we mentioned your book, the

00:37:11: book will be linked on the YouTube video.

00:37:13: I'm not sure how we can do it on the podcast, but I'll figure this out.

00:37:17: Is there anything?

00:37:18: title of your book.

00:37:19: Can you mention the title of your book?

00:37:21: it's the AI recruiter is the title of my book, revolutionizing hiring with advanced

00:37:27: GPT powered prompts.

00:37:28: And then I also have recruiting classes .com, which is where I train.

00:37:33: And then I think you also mentioned you have a GPT in the GPT store.

00:37:40: in the GPT store it is called the AI recruiter and it's a purple squirrel and

00:37:46: anybody who has access to the GPT store can use my AI recruiter for free.

00:37:52: I think you should give it a chance because again, Mike was very nice to show

00:37:57: us some kind of demo how it works.

00:37:59: I think it's really impressive and he invested quite some time to improve it and

00:38:04: get it to a quality where he's happy and he's actually used it in his previous

00:38:08: roles.

00:38:09: So I definitely think you should give it a try.

00:38:13: Yeah, go ahead.

00:38:15: because I find it so interesting, especially like for startups and young

00:38:18: companies, recruiting is hell.

00:38:21: What would you suggest like startups, young companies, small companies, they

00:38:26: want to hire?

00:38:26: Like I have one friend of mine, he just like hired three more people, like he's

00:38:31: hired like 10 people in the last three months from two starting like, so what...

00:38:38: How would you approach an AI recruiting from a young company's perspective?

00:38:45: Maybe you can give some hints.

00:38:48: Well, besides hiring me for consulting, teaching you how to do it, I, you know,

00:38:53: it's a matter of tools.

00:38:54: Sourcing is challenging.

00:38:56: LinkedIn is limited, especially for startups.

00:38:59: Um, there's seek out, um, they work with small companies.

00:39:02: They're like 160 bucks a month for small companies.

00:39:04: They have a license.

00:39:05: So you only need one.

00:39:07: Um, that'll give you access to, to the database.

00:39:10: Cause you're not going to come with one.

00:39:12: You're not going to have a database.

00:39:14: Um, I would start to use them.

00:39:15: They have some AI powered tools, uh, to help you with sourcing.

00:39:18: And then I would look for seek out.

00:39:24: And yeah, they're, they're really good on the front end.

00:39:26: And then they would be a CRM either something like higher easy or higher

00:39:31: suite.

00:39:31: They're AI powered.

00:39:32: This is nice for our profession is actually a lot of the tools are AI powered

00:39:37: and they're available.

00:39:39: you know, subscriptions like $150, $160 a month a seat, which for a startup is very

00:39:44: affordable and you typically only need one as a startup.

00:39:47: So you can get, you know, an AI recruiter, a sourcer, and a CRM for about $350, $400

00:39:54: a month.

00:39:54: If you're going to be doing recruiting that's, and you're using AI, that's

00:39:58: probably the way to start.

00:39:59: It's low investment.

00:40:01: It's not super intense.

00:40:02: It's all SaaS.

00:40:03: Mm -hmm.

00:40:03: like an implementation unless you wanna do API implementations and those are becoming

00:40:08: more and more simple, right?

00:40:10: And straightforward to do.

00:40:11: So those would be my suggestions to start with.

00:40:15: Thank you very much.

00:40:16: Mm -hmm.

00:40:17: I love that you got that one in before we part.

00:40:22: Again, thank you, Mike, for all the valuable lessons.

00:40:27: I haven't been this close to the recruiting space, so I definitely learned

00:40:31: a lot.

00:40:32: Thank you for taking the time and sharing your knowledge, expertise with our

00:40:37: audience.

00:40:37: So any parting thoughts from you, Edgar, Mike?

00:40:40: Yeah, also from my side, thank you very much.

00:40:43: It was really insightful.

00:40:45: I'm really looking forward to look at this because, like I said, it's a hard

00:40:51: challenge for every company out there and even harder, I think, because even getting

00:40:57: your stuff on StepStone is so expensive.

00:41:02: So having any alternative is really welcome.

00:41:06: And...

00:41:07: Yeah, thank you.

00:41:08: Thank you very much.

00:41:08: It was really insightful also for all you out there listening.

00:41:12: Please take this example.

00:41:14: If you build AI solutions, try to start where it's most effective and takes away

00:41:19: mundane administrative tasks, where it involves language understanding.

00:41:22: If you go in with that mindset, you will pretty soon get really great results and

00:41:27: see what AI can do for you.

00:41:30: And yeah, thank you very much.

00:41:31: Last words from Mike.

00:41:33: Oh, well thank you for having me.

00:41:36: I hope this was insightful.

00:41:37: For those of you who are yet to get started, I guess my last parting advice

00:41:41: would be ask it a question you never got a good answer to when you were a kid.

00:41:46: Like, why is the sky really blue?

00:41:48: Why is the grass really green?

00:41:49: Something you never really fully got your hand around.

00:41:51: And just start having a conversation with it.

00:41:53: Treat it like you would a text conversation with somebody who's, I don't

00:41:59: know, extremely well educated and could talk to you about any subject.

00:42:02: Awesome advice.

00:42:04: Thanks again.

00:42:06: All right.

00:42:07: Bye.

00:42:08: Bye bye.