How ChatGPT Will Turn Recruiters Into Software Engineers With Robin Choy of HireSweet

Listen & Subscribe on your favorite platform
Apple | Spotify | Google | Amazon

Visit us at RecruitingDaily for all of your recruiting, sourcing, and HR content.
Follow on Twitter @RecruitingDaily
Attend one of our #HRTX Events

Transcript

William Tincup 0:32

This is William Tinkup, and you're listening to the Recruiting Daily Podcast. Today we have Robin on from HireSuite, and our topic is how ChatGPT will turn recruiters into software engineers. And I've never talked about this. So I'm really, really, really excited to get Robin's take. Robin, would you do us a favor and introduce yourself and

Robin Choy 0:54

HireSuite? Sure. And, uh, and I'm very happy to be, uh, on the Recruiting Daily podcast, uh, at last, cause I've been listening to it as well. So thanks for Well, it's you, it's

William Tincup 1:03

you and my mother, Robin. That's, that's the, the list of ship is It's broad and wide. It's you and my mom. That's it. That's what we

1:11

got.

Robin Choy 1:12

Did you leave a review on Apple podcast? I

William Tincup 1:15

should get her to actually, that's a really good idea. That is a really good idea. I'm going to do that. Cause you know, I'll go over over the weekend and she'll ask me questions from the week. It's craziest thing in the world, Robin, like she was like, I don't understand financial wellness now. What is that exactly? I'm like, she's 86 years old. Why do you care? Anyhow, go ahead. Tell me, tell me, tell me a little bit about harshly.

Robin Choy 1:42

Yeah, so Harry's suite, we, we created a company in 2016. So, uh, seven years, we'll be, um, uh, we'll be blowing our seventh candle this year. Um, and we've been through a lot of stuff by votes, lots of things, but, um, the, the core of what we do remains the same. It's. Building software for recruiting teams. Um, and we started with lead generation, then it evolved into a recruiting CRM. And today we have two products. The first one is a town marketplace. Uh, so privileged tech talents operating only in France. Uh, that's where I'm from. So you can probably say from the exit. Where we're in France. So I'm originally from Bordeaux and, uh, and, uh, the company is based in Paris.

William Tincup 2:25

Oh, very cool. My, my son went over, was in Normandy for two weeks this, this summer and loved it.

Robin Choy 2:32

Of course. Yeah. It's pretty cool. There's the, um, the landing beaches. Yeah. It's pretty nice. So, so yeah, company based in France, first product type marketplace. And the second one is a recruiting CRM. And I myself am based in the Bay area in San Francisco Bay area, uh, in the city. And the second product is what we're now bringing to the international market. And it's a recruiting CRM to help recruiters be more efficient, meaning. Making more placement for agency recruiters or individual recruiters, or just making more hires for in house recruiters. So it's a productivity software that helps go after candidates, nurture talent pools, automate outreach, um, automate newsletters to candidates. Get data, um, yeah, recruiting CRM

William Tincup 3:18

again. I love it. I love it. I love it. I love it. And I want to talk to you a little bit more about that, but we'll do a different podcast on just HireSuite. But, so, tell me a little bit about your experience with ChatGPT. Uh, obviously you're, you understand technology a little bit more, a little bit better than the average bearer, but, like, what do you, what do you think of it right now and in this phase that we're kind of interacting with it? Well, the

Robin Choy 3:42

cool thing is because I'm the, um, the founder of this company, I'm basically paid to stay on top of trends so I can try and I can look at everything. I can, I can spend my time talking to people about it. And it's basically part of my job. So I don't know if I understand it better, but at least I spent more time. Um, and since Harriswood inception, we're always sad about a third of the team working in data sense team. Um, especially on AI models, but not only AI models because there's lots of stuff and so we've been pretty Embedded into AI since 2016 looking at all the innovations look all the technologies And new technologies the research and see what we could Implementing the product. So obviously when we heard about, uh, GPD 3 and then chat GPD, we were all ears and got pretty, uh, deep into it. And that's how we got started with this. And, um, and how I got started with chat GPD personally. So I'm not an engineer, uh, from trade, more of a business guy, but, uh, then been investing, uh, program a bit. I've been investing a bit into chat GPD, but mostly what's interesting to me is the. Not so much the R and D behind it and how it's built, but rather what you can do with it and the opportunities when you, when you apply to very practical use cases. Uh, so that's how I got started into it and started understanding what could be done, what could not be done. Um, it's easy to like the easiest use cases are drafting, drafting outreach messages. That's also what we do in our products. So we, uh, pretty, pretty quickly implemented this in our products. Um, also drafting job descriptions, job ads. These are like the most basic use cases. Then people will want to use Chagibili to do matching, like match a candidate to a job. Um, very. Like suspicious and cautious with this because, um, again, we've been building our own matching systems in since 2016 and we know what it takes. We also know what the regulation takes. Right. Uh, you need to be very transparent and stuff. And then, so, yeah. Okay.

William Tincup 5:47

No, it's, what's interesting about the matching side is We've gone through different abolitions of matching, right? So it started off with some, some, some basic things. Uh, now it's gotten to skills and, uh, and, and, and even further into what potentiality, you know, what else do you want to do other than the things, so we know what you can do. We know what we want you to do. Uh, and do you have the skills to do that? The breadth and depth and kind of the measurement and skills testing and all that sort of goes along with it. Okay. Those things match. Here's what we need. Here's what you have got it. But then it's like the things that aren't there that aren't explicit in either the job description or in the candidates desire, like that stuff's really fascinating to me. It's like, okay, how do we unpack the fine and to find out a true match? Of, of, okay, what's not in the job description that should be, so that's, that's a, that's a kind of a hiring manager recruiter kind of to figure out like, okay, what's not been said that should be said, but also for the candidates so that they actually do get matched to the right jobs. Understanding the things that maybe they don't say, you know, like, you know, like they don't, they don't say, Hey, you know what I want to do, I want to get a master's degree at the same time. I want to actually learn about AI. University of Texas has this MS and AI, and I want to give a master's degree while I'm doing this. Like, if we know that, you know, if that's something that we can pull out or find out, et cetera, then we can also use that to help make a better match.

Robin Choy 7:23

Absolutely. And then matching it's, it's also all about the, uh, preconception and biases of both the candidates and the companies. And if you're, uh, I remember, so when we started hire suites, um, thought that my job was to bring candidates that were, um, what the hiring managers, uh, described to me. So we're looking for that person. Okay. Let me go and find that person. But then I, I realized that the job of recruiters and the added value for recruiters. He's saying, actually, you believe that you need that, but let me show you people that, that are not exactly these, but that will still fulfill your need. And then there are some sales element to it. Like, okay, that person, uh, she's a junior. She doesn't have exactly the right skills, but look at that. And these, she could be a right fit because she's been actually passionate about the space, uh, since she's a kid. Um, and this is not something that you can automate and not with logs. So there is a very deep element to matching and then but we're slightly moving away from the main topic for the podcast But there is another example that was very interesting with our higher suite with our experience at higher suite Is that matching is all about comparison as well. So if if I show you the best candidate Possible like that person is the absolute 100 percent match for your job But you only see that person as a hiring manager you want to see more people and maybe you'll even reject that person because your your Um, your expectations are not right and you're like, okay, that was easy. I got that person Let's see 10 more of these people and for now, let's just put that person on the back burner I don't want to talk to them right now. Let's see 10 more people and then that person is gone and you're um We're with 10 people that was just less relevant in the experiment that we led at the less right.

William Tincup 9:16

Well, I'm sorry to interrupt Robin did less relevant. It learns you're what was less relevant to you in that particular job. So if done well, if I understand the way you're the way you're going, you'll say you have 20. You pick 10, the 10 that you, we would, we would historically call silver medalist. The, it, it now learns, okay, what are the traits? What, what are the silver medalists? What makes those silver medalists as opposed to the people that are gold medalists, and then the more you do that, the more it is actually trained within that, within that job. No. Is that, am I charging

9:53

that,

Robin Choy 9:53

that that's, so you, there, there are ways to build models that do this and this is how you will build a matching system. That's so how we do it, right. But what I was, what I was trying to say is that, so we had that experiment with aerospace. Okay. We, uh, before when we used the lead generation model, so we had much more data. We'll say, okay William, you are hiring someone. Um, let me show you a list of the. Top 100 people that are most relevant for you. Okay, I'll show you a list. And you, you'll reach out to, and you'll say, okay, these people are good. These people are not good. And you're on average, reach out what we measure. You reach out to seven to 65 percent of those people. So the top 65. If we run that experiment to several people, we know, know who are the top 65, but what happens is if next time you come, and instead of just showing you a hundred people, I just show you the top 65 people, you'll still reach out to. 65 percent of them. Cause somehow when you're recruiting, you'll, you'll, you'll have this need of just saying, okay, that person is fine. That person is not fine. And you still need to eliminate some people. So now if I come to you and I show you the top 65 people, you'll only reach out to 50 people. Because you have this, uh, unconscious need to compare people. Um, so that's also why it's hard to build a matching system. Do you see what I mean by that? Oh, 100%. 100%. Measure this, and then next time, if I actually show you the same 100 people, but then I add 30 people that are not relevant. You'll reach out to the entire hundred people and keep the 30 people. So there is no, like, you cannot have an absolute matching score because it's also about the biases of the person that will receive them. The, um, so that makes it very difficult. That makes it, that makes it. Required to be very explainable, which which LGBT is not because it's very, um, more of a black box. You don't exactly know what's happening with it.

William Tincup 11:55

What do you think about, uh, the potentiality of building, uh, one to build their own large language model? So, and I know I'll use be Marie as an example. So be Marie. Um, I had him on a podcast at one point, and they've Uh, two or three years, they've been kind of in the background, kind of skunkworks, trying to build their own large language model just for their customers. So just for hiring managers and recruiters and people that are using BMRE, candidates that are applying through BMRE, et cetera. So I mean, uh, first of all, it sounds fascinating when I first, when I first heard it, I'm like, Oh, that's cool. So we can actually learn more about me. You know, it's one of the fails of chat GPT in general is that it's not specific enough. It's not, you know, it's not nuanced enough or, or, uh, it, it gives you, it's, it doesn't go into the crevices where you'd like for it too. So I can understand why people would do it, but that that's also a heavy lift. To build your own large language model, uh, yeah, that's not something you do overnight. A, B, it's still got to learn. It's still got to do all those things and still not going to get to the point to your point where it's a hundred percent.

Robin Choy 13:12

Exactly. Um, exactly. Cause there are big diminishing returns. Like I don't think an LLM will ever be capable of building a matching system. And it can be good at tons of stuff like generating jobs and again, generating message and then generating code as well, which is, uh, something that's very, uh, uh, underappreciated, but the very powerful as well. But I don't think it will be good at matching any. Because it's just not the right technology. And if you want to look into matching, there are other existing technologies. You can look at TextKernel for instance, they build their own matching systems. I wouldn't I wouldn't go into building my own LLM because that's a lot of work. And then you also need the adjacent learning. So the fact that ChatGPD was trained on, I don't know, Wikipedia pages better at writing stuff. So if you only train an LLM on job description, for instance, it might actually not be as good as a more generalized LLM. So the right approach will probably be fine tuning an existing LLM, just saying, okay, we'll use existing technology, and then we'll adapt it to our use cases.

William Tincup 14:26

Right, right, right, right. So one of the things I wanted to kind of track with you is the connection between recruiting and software engineers, right? So. They, they're learning as you have, because you're doing it for the company, you're learning, you know, all these things for the company, um, but they're learning in real time and they're learning how to actually use this. You, you said some of the easier cases, uh, messaging, job description, job ads, things like that, like, and you know, what's fascinating about that is they're learning prompting. Right. They're learning how to actually get it to where it's a good, you know, let's say it's a matches or it's an outreach and let's say it's a cold outreach, you know, this is someone I don't know. They don't know me and we've never had an interaction with each other. We, this would probably used to be, would be via LinkedIn in, in mail. Right. Okay. But how can I make it? snazzy enough? How can I make it sarcastic enough or how can I make it, you know, like they're learning all of that, the technical prompting right now, which, which was, again, is great. And then they're doing it while they're doing their job, which is also great. But where do you see that first of all, what do you see in that, but where do you see that going? What, what skills do they need to develop so that they can get the most out of ChatGPT?

Robin Choy 15:48

Well, there's, I think there's two approaches to Chargipedi. The first one is just being better and faster. Um, I could write that outreach message, but maybe if I write it myself, I get a 25 percent reply rate. And if I use Chargipedi and iterate enough, I'll get a 35 percent reply rate. So that's just doing what I used to do, but faster and better. And then, uh, and, and I think that's part of the training. You should look into it. You should look, uh, you can go to, there's a pretty good website. That's a nonprofit as well. It's called learnprompting. org. Helps you be a better prompter, a prompt engineer. So that's, that's skills that are necessary. Like you have to understand how to use Chattopaddy to be faster and better. But then there's an entire field that is not really touched on for now. Um. Which is using Chagipedi to do stuff that you didn't do before. And actually, so I did a whole webinar on this, on how you can use Chagipedi to be, to become basically a software engineer. So you used to be just a recruiter, and now you can actually turn into a software engineer, write your own programs, write your own automation, your own scripts, using Chagipedi. And I think that's interesting because once you save all this time, um, once you save all this time because you're now much more efficient, you're much more efficient. No, much faster. Uh, you get more replies to your outreach. Well, what do you do with all this time? And I think it's interesting to look into, you know, just doing more of what you already do, but you're doing new, uh, new stuff and very exciting opportunity for me is like, what will recruiter, uh, what will a modern recruiter look like in 10 years? And, um, that will probably be a person that's way more. Um, autonomous that can write their own code, their own scripts again, that could be using Chagibili, but that can also be using Chagibili to write the script or the code or the program and then learn from it so that in the end, you can basically write it yourself, but still use Chagibili, but at least you, you'll just, um, use Chagibili as an Accelerator of sorts.

William Tincup 18:00

So if, for the audience's sake, we're not just talking about technical talent. So recruiters that deal with the recruiting software engineers. So not to be confused, we're talking about no recruiters in the futures that are, that are recruiting call center employees, et cetera. It's it's so, so that the audience doesn't get confused, uh, that we're not just talking about tech recruiting. We're actually talking about. The technical aspects of recruiting, becoming more sophisticated and, and, and again, getting your time, getting some of your time back, uh, so you can go deeper into like, I get this question all the time. I'm sure you do as well. It's like, well, will this all be automated? I'm like, no, no, because at one point a human by and large wants to talk to another human. And again, not. Not at the beginning. They want a lot of this stuff that the, you know, the calls to be automated. They want this skills testing. They want all the assessments. They want everything to be sussed out. Then they want to talk to somebody and go tell me a little about the job. Maybe some of that can be automated. Great. The hiring manager wants to get involved. Okay, maybe some of that can be automated. At the end, very end, before there's a job offer made, I don't see in the next five years Candidates being totally comfortable or, or companies being totally comfortable with, without having some human to human contact. Now, will it happen much further past that? Well, yeah, we'll have flying cars at one point. So, yeah, of course it will happen, but not, I don't believe it's in the near term. Now, I might be wrong about that. What's, do you have a take on that?

Robin Choy 19:45

Um, you know, I'm on the, on the, on the subreddit, it's, uh, I was trying to find it, but something like recruiting or something like this, a person had an interview process with an AI and he was like, okay, what, what the fuck, what's happening there? He was supposed to have an actual call with an AI and the company gave it a first name and a last name, uh, the AI and the candidate was just like, okay, no way I'm going to join a company. And he dropped. And so. I do believe that we should look at AI in areas where it can help us be more efficient. Things that we already do that sometimes we don't do well. I don't know, like rejection messages, for instance. Very few companies send rejection messages. Maybe you can use AI to send more valuable to the candidate rejection messages. But then in the end, it remains a competitive process. So if you try to automate everything and your competitor just has a human at exactly the right step, the right points to improve the conversion, they'll steal all the top talent from you. And then you end up being with a very efficient process, but just not hiring best people because your competitors decided to invest slightly more, have the human exactly at the right place, and then just close all those top people. And because it's a competitive process as well, sometimes you just need to invest 10 percent more and you, you'll still get all the top people, right? If you are just 10 percent better, you'll still win all the deals. So in the end, one company will just hire everybody and the other will just struggle hiring. So I don't believe that AI will replace humans. And, uh, and I don't think That candidates want that, because candidates want information on a company, they want transparency, they want also respect, some kind of respect, be able to ask their questions. So some AI, sure, as long as it provides more transparency, more respect, but you absolutely have to keep the human, otherwise your competitors will And, uh, and they'll steal all the talent for you. So that's how you think about it. Yeah.

William Tincup 21:54

Yeah. I'm, I'm tracking the same thing. So advice you'd give recruiters right now, as they listen to this podcast and my mom, uh, as they listen to this podcast, they're probably thinking to themselves, okay, how do I get started? How do I, how do I, how do I kind of track where I should be? What, like, what should I be tracking towards so that I'm, you know, learning some of it's, you know, bumping into stuff and just kind of making mistakes, et cetera. But what's your guidance right now is sort of to, Hey, listen, stop doing some of the things that you used to do and slowly start to do these things so that you can get yourself maybe by the end of the year to this place.

Robin Choy 22:33

So my advice would be, look at your work, look at everything that you do and think. So look at everything that you do that has no added value. There are a lot of areas where a machine could do it. And you have to, you have to automate yourself, otherwise other people will do it, or your competitors will do it, or your company will do it at some point. So look at where it doesn't add value. and think of how you could automate it. And then look at all the stuff that you don't do and that you could do and think of a strategy to get there. And maybe it doesn't even require chatgpd. Maybe it's like spending more time in the intake meeting, the the kickoff interview with the hiring manager. Maybe you're just not spending enough time now and maybe You could send a form to the hiring manager just prior to the meeting to get more information. So that could be stuff like this. Doesn't have to be, don't think technology first, because now what I see a lot of people do is think technology first and like, okay, I have to be, what can I, where can I apply to my work? But do the opposite. Look at your work. What's not adding value? What else could you be doing? And then look at the other ways that you can solve this. And then also remember that now the sky's the limit. You can use Chattopaddy to be a software engineer yourself. So if something's missing to your workflow, don't wait for a company to build it. Or don't wait for a team of software engineers to build it or a vendor to build it. You can build it yourself. Log into Chattopaddy, ask it to write the code. And there are lots of ways to do it. Run basic code. So you can run basic code in Google Spreadsheets, for instance. Um, it's called Apps Script. You can also use Google's, uh, there is something called Google Notebooks. It's called Google Colab. It allows you to run Python scripts in the cloud. Um, and again, even if you don't know anything about programming, this is the perfect, perfect time to learn and to use, and you don't even know, you don't even have to know how to write the actual code snippets. You can have Chagibili. Just write it. You just need to define your needs. So start with the need. Uh, that would be my very long advice. No,

William Tincup 24:43

it's great advice. It's great advice. It's freeing too. As you were, as you were talking about that, I'm like, you know, it's liberating that you can go and learn this stuff. Exactly. It's a mindset. You got to look at it, not as. But it's something really, really exciting. Like my job is now going to be more exciting because I'll have more control. I just need to learn some new

Robin Choy 25:05

things. Exactly. Imagine that you're, you're a software engineer and a copywriter and a, a designer at the same time. So you have all those superpowers. What can you do differently in your job now? Like you are, you literally have the ability to, to, to create paintings. that are the quality of the best painters in the history of the world, uh, to create illustrations, to create videos. You have all these superpowers. Um, You won't be replaced by AI, you'll just be empowered, but you have to embrace the fact that it's not all yours, but also all your competitors have these superpowers as well. So if you don't use them, you'll, you'll end up being crushed.

William Tincup 25:48

That's right. That's right. Drops mic, walks off stage. Robin, thank you so much. It's been, I mean, a fantastic topic and I love your take. So thank you so much for coming on the show.

Robin Choy 25:59

Thanks, William. And I hope your mother will enjoy it as well.

William Tincup 26:03

Well, I will find out, I'll give you feedback. Alrighty. And thanks for everyone listening until next time.