Episode 61

Guy Attia on AI and the Future of Hiring & Leveraging Technology for Smarter Recruitment

TIM: Guy, welcome to the objective hiring show. Thank you so much for joining us.

GUY ATTIA: Hey Tim, it's nice to be here. Thank you for having me.

TIM: It's absolutely our pleasure, Guy, and I was thinking about where we could start our conversation today, and where I think it would make sense is to have a discussion about AI and hiring because AI is taking over the world. Is that the right way to put it? I'm not sure. They used to say that software was eating the world. So AI is doing something to the world. I don't know what, but it's certainly having a big impact, and yeah, it'd be interesting to chat about it in a hiring context, so I'd love to know: Have you started to experiment with any AI in the hiring process? Have you seen candidates on their side start to use AI? I'd love to get your initial thoughts there.

GUY ATTIA: So first of all, I will just say that as an experienced AI and ML developer, I also graduated with my master's degree in AI and ML, so I truly believe this revolution of AI is real, and there are a lot of benefits and a lot of practices that we can take and will be taken in the future across the globe. Specifically, in one bit, currently we are not using AI in our hiring processes, mainly because our recruitment pace is quite low. we are a slow We are a fast-growing startup, but still we are around 100 people, so we are not recruiting on a daily basis, and also the pace of or the amount of CVs that we are getting is quite low in reference to relatively bigger companies, so we don't have the painful process enough. It's not painful enough for us to establish an automatic process of new technologies. I do believe that AI is of great value in these processes, like screening resumes and matching candidates to job descriptions, and these tools can save us a lot of time and a lot of effort and ensure consistency. However, they must be carefully calibrated to avoid bias, and we know we already saw some examples against all of us in the AI world of embarrassing bias and comparison between people and other animals or stuff like that, so it needs to be taken into account and be sure to have a fair consideration of the diversity of the candidates.

TIM: And what about on the candidate side? Have you started to see them dabble, for example, in, I don't know, creating CVs or maybe trying to get some interview assistance with ChatGPT? What have you seen so far?

GUY ATTIA: So I already see some sparks of using ChatGPT to write the resumes. I can see similarities between the resumes of different people if it's mainly in the context part, like ChatGPT can't invent experience; it can't invent terms, but it will design your experience in the right approach to pass the automatic tests. and also there is I already heard about some new hacks that the candidates are using; they are trying to hack the system because they know that AI bots are going to screen them, and, for example, they are writing additional text in white color that people can't see but bots can. So for example, if I'm a candidate for an AI solution or an AI job, sorry and I know that the chatbot will pass me if I have six years of experience, and I have four, so I, in white words in white text, will write six years, and then the chatbot will move me to the next level, the next step, so I already heard about stuff like that.

TIM: It's funny how circular things are because that was like Google search engine optimization, a 1999 tactic, and now it's coming to, I don't know what you call this AI ATS optimization in 2024, so some beautiful circularity with these tactics that takes place.

GUY ATTIA: So I think what is even funnier is that the AI bots are going to talk between themselves because the people are going to use them in order to write the CVs, and they are disqualifying or not the CVs, so they are going to optimize the same thing at the same time. At the end, it might happen that the CVs will be the same for people that use only AI bots.

TIM: This kind of intermediate period where, for most companies at the moment, it's still a human reading a CV or humans reading the CVs to make the decisions, so candidates probably shouldn't be optimizing for an AI reading their CV now, but yes, that will presumably very quickly switch and change. I feel like it's inevitable that companies would have to implement this, especially in some markets. Like, I've spoken to so many people recently in the United States and Europe, and their application volumes are outrageous. Some companies in America would get a thousand in a day easily. I guess the Israeli market is probably slightly different in that regard at the moment because maybe it's just a contained market, and maybe people from overseas aren't applying for jobs there as much as they would in Europe, for example. I'd like to get your thoughts on that. But basically the companies are going to have to start doing this AI CV screening now. One issue I think is, though, no matter how good this screening is, the data set you're using to do the screening is still a CV, which is now maybe even less accurate than it was before because it's been written by Chachabuty. Like what state are we going to get into? Like how accurate can it be when the data set itself is pretty cracked? What do you think?

GUY ATTIA: Yeah, and mainly if CVs are still relevant or not to today's world, I will start by talking for a second about the Israeli market and the amount of applications, so I will give our company in one bit as an example, although we are in the Israeli market, and most of the developers are from Israel. Still, we are working globally, and we have employees also in Russia and in Europe and in the US, and we are open to candidates, but I totally agree that the amount and the traffic of candidates that apply from their own, like they push it from their side, is quite low. Although we are starting to publish it and market it more in Europe as well, so maybe it's a matter of traffic and focus of us, like we are focusing on the Israeli market in terms of developers. And maybe if we focus across the globe, we will get more CVs as well. On the CV side and the relevancy of a resume in today's world, I don't think that CV is dead. I think it does need a significant evolution. On one side, CVs are optimized by CGPTs and other Gen AI tools such that the text is better worded and sounds amazing, but like I said before, from the other side, the actual experience and skills won't change by this optimization. and it will still be possible to distinguish between candidates. In addition, I think that resumes have a significant impact on the later stages of the hiring process besides the screening phase, which is crucial, and it's the main painful part for most of the companies to screen thousands of applications per day. But still, the CVs will be crucial in the next stage. It's usually used to review projects, skills, and mutual interest between the interviewer and the candidate, so if you want to have a resume, you will need to tell me that in the beginning, and I will need to think about it in real time instead of before the meeting. and so I don't think it's that, but I think that precisely because of the use of AI, I do believe that CVs will become more narrow and will contain only a closed list of skills per job type. This is my dream. I really like classification and the categorizing stuff; it makes life easier, and I think with LinkedIn these days, it's even easier to see that. Let's take a role, for example, let's take a data analyst. So instead of people describing in words what skills they have, there will be a closed list that they can check: Okay, I know SQL; I know this database; I know how to develop dashboards, and then it will be even easier for the AI solution to distinguish the real needs.

TIM: And I guess also part of the use case or benefit of large language models is they're quite good at taking unstructured data and making structured data out of it, so it might still be a candidate supplying with a CV and some other worthy things, but then the large language model can then convert that into the structured data to then do some kind of ranking. Is that how you would see it working?

GUY ATTIA: Actually, I didn't think about it, but it sounds like a cool and great solution. We can use AI in order to do that as well, even if it is not related to Gen AI. The AI solutions are like clustering, and we can find similarities and non-similarities between candidates, so maybe, you know what, maybe I will implement it here later. Maybe we can create an AI and ML model that will cluster all similar candidates into groups, and then I will look for the outliers. I will ask him to give me the one that is not like the others, and I will explore his cv

TIM: That's an interesting thought, so let's unpack that a bit more. Because you're thinking that everything's going to be condensed and everything's going to look more and more similar, you almost want the people who are a little bit outside the box, outliers on some dimension. Would there be I'm just thinking about it now: would there be, like, a good and a bad outlier in a way, like, depending on which side of the distribution it's on? What would you be looking for in those outliers? do you

GUY ATTIA: CVs—that's a good question. It's there is no bad publicity, so maybe there is no bad outlier, like in this case. I would like to have someone that is not using the most common way and is thinking out of the box, maybe to distinguish himself. Still, of course, you need to pass all the minimum requirements and skill set and whatever we defined in the job description. But if everyone is writing the same thing using the same AI bot, I can't distinguish between them, and maybe the one that we write is an additional paragraph that will distinguish him. This is the guy I want; maybe I don't know, but it will help me to identify these specific persons out of thousands, for example.

TIM: Thinking of it froma slightly different perspective now, but the same thing, so from the candidate's perspective at the moment, again, especially in markets like Europe and North America, they are going to LinkedIn, they're seeing a data analyst, a data scientist role, they're saying, Wow, a thousand applicants in a day. Are you kidding me? And they're thinking to themselves, Oh my God, I thought I had to apply to 50 jobs to get one interview; now I probably have to apply to 500 to get one interview. So then they're applying en masse, which is then causing this vicious cycle problem where there's just more and more applications, which is then causing their application to be lost in the sea of noise, if you can call it that, for these candidates. If they want to stand out as a candidate as opposed to you finding them standing out, how could they stand out? Do you think in that application process you have any thoughts?

GUY ATTIA: That's a big challenge, and honestly I faced the same challenge a few years ago, and what I try to do is create a specific CV for each kind of job, and I like to manually handcraft the CV for different positions that I submitted to. I don't really think there is a solution unless people will, you know, from the job perspective, from the company perspective, there can be an automation that will write how many CVs there are already in line in the queue and then the set of expectations. will get lower, and then if I were a candidate, I would see, Okay, there are 40,000 CVs that are waiting for review before mine. I won't even comply; I won't even apply, and then again it's a cycle, so the number of CVs will go down, and I will apply for the most common, or maybe we can tell them in percentage what is the chance to get to see my CV will get examined in the next few days. It will give me hope, I know.

TIM: Yeah, you've touched on something I think is really important there that I feel like AI could drastically improve, and that is just the transparency of the process. I feel like the way hiring is done, maybe because it's so manual, it's so tedious. I feel like we're not really in a position to measure things as part of the process and then share those metrics. So, yeah, something as simple as like average time from applying to reviewing has my application been reviewed yet? How long did someone spend reviewing it? Who was reviewing it? What did they like about it? All of that is so manual at the moment; it's never really going to go anywhere, but if an AI was doing these first few steps and that information could be shared with anyone, that could be shared internally, that could be shared in an HR tool, that could be pinged to the candidate. There's just so much upside, I think, from moving to some clever automation in these early stages for sure.

GUY ATTIA: So I think even if not related to AI solutions with today's technology, there can be a lot of automation of data processing and data analysis to show these kinds of things, and even if I do it manually, I can still log how much time it was in process, right? or waiting for me, and it can be reviewed by the platform that I'm using, if it's LinkedIn or whatever I choose.

TIM: You're right, and maybe that then makes me think, why hasn't that already been done? So let's just take that as an example at the moment: in an applicant tracking system, a human is sitting there opening up a CV, looking at a CV, clicking reject. Like, obviously, you can track all of that, but isn't it at the moment, like, if I think of the ATS as I've used, they're not in some cases not tracking that at all, at least not visible to you as a user, and they're certainly not sharing it with the candidates? So I wonder if the companies aren't demanding that maybe they don't want that level of transparency; maybe they're worried about that. I don't know; it's an interesting idea to think about.

GUY ATTIA: I personally don't know why I didn't do it, but I think that for companies with thousands of applications per day, it's even a good perspective for the HR departments to know what the traffic is, to know what the processing time is, and to know how good the process is and how good the feedback to the applicants is. I guess the HR department cares about how much time this person is waiting, and it's their client. By the way, the candidate is a company client, as per se, so you want to give good service or not. If you ever give it good service, you can get perfect publicity also because of that. I think it's just not important enough for the candidates, sorry for the technical persons in the companies, and I don't think the HR department has enough power and influence to open a project like that because it will involve technical and data guys to develop this streaming process or this analysis process, and it will require justification, which is not easy. No, there is a specific budget for different developments, and maybe they don't see the value in that.

TIM: Yes, I would also argue, though in theory, the applicant tracking systems, the HR tech that companies buy, should have This functionality is the fact that they don't have it. Even though we both know it's technically pretty easy, it must mean that people aren't demanding it. Now this leads me into another thing I want to talk to you about, which is maybe the fact that talent and HR teams and the leadership in those teams very rarely would be from technical backgrounds. Very rarely would they come up through machine learning or data science or software engineering or hard sciences; typically not. They'd come from other areas of the business, including sometimes HR specialists. The fact that they maybe don't have that kind of scientific analytical experimentation lens means that they aren't thinking about these problems in the way that a data person would or a software engineer would. What do you think?

GUY ATTIA: I think it does pose a challenge, especially when evaluating highly technical roles. Collaborative hiring, where technical leads and talent teams work closely, can bridge this gap. For example, upskilling talent teams with basic data literacy can also improve hiring outcomes in OneBit; for example, for specific roles, we cooperate between us, the data department, and other departments, the R&D and the more developers, tech leads. Thank you for assessing candidate skills when needed. For example, I reviewed the CV of a data engineer. I also asked the VP of R&D to share his feedback with me because it's collaboration between data and development skills, so also for other candidates it's relevant. In my opinion, there is a huge jump during the last few years in the development skills of data developers since there are bigger expectations from them to work with common best practices like Git and other code development, so I do believe that this gap will get smaller and smaller in the future.

TIM: The other angle I was thinking about recently was when I'd spoken to marketing analytics people who had been involved in hiring at scale in their company. It was very interesting how they took a very marketing funnel lens to hiring, and to me, hiring in the hiring funnel is no different from any other lead funnel. like you get a bunch of people at the top, they eventually filtered down through some conversion gates until you hire one, or you sell something, and it was interesting to see how perfectly they'd applied their marketing hats to the hiring hat. Another example is a company that works with us, Agoda, a travel company based in Southeast Asia. They have, within their marketing hiring, applied all their marketing principles to hiring, and so they have all these metrics they capture internally, including things around, like, interviewer performance, how many interviews have you done this week, and how effective a screener are you as an interviewer. So, for example, if you consistently interview people and then in the next round they all get rejected, you're not really an effective gate, and so they have hacked together all these metrics themselves that they built on their own back because they've got engineers, they've got the marketing brains, and they can combine those together. but I wonder again whether if there were more of these other skill sets in talent teams, they would then be demanding this as basic functionality from the ATS providers because they would have their marketing hat, their analyst hat, their science hat, and so maybe there's a lack of that diversity of hats in talent What do you reckon?

GUY ATTIA: I think that, yes, the lack of talent sets in the processes, like what you described in the Goddard, seems like a really healthy process and like a personal review of their processes, and I guess it came from and like a retrospective of not very good hiring for the last few years, and then they said, Okay, how can we get it better? So we will add another feedback or another analysis on this kind of stuff. Maybe what's needed here is a collaboration between some analysis developers and analytics developers with the department that is in charge of the hiring; usually it's the HR, and maybe there they can think together on what the KPIs are. What are the KPIs for the hiring process? How do you know if it's good? How do you know if you need three or five interviews? How do you know if the people that are joining or that you already selected after the first interview are good enough against the other candidates? How do you measure it? So first of all, we discussed it before; we need to gather this data. As a data person, I can tell we want to gather as much data as we want, and then we'll decide how to analyze it. So first we need to automate this kind of stuff, and hopefully all these AR platforms—sorry, HR platforms—are going to do it automatically for us. and later we will just need to justify and define the KPIs in order to analyze it, and it will close the loop as I can see it.

TIM: Yeah, I hope so, and this is why I'm so excited about the next few years, because I feel like, in, let's say, the recruitment hiring space, we're at a very nascent stage in the way we think about it and the kinds of questions that we would ask if we were looking at marketing or sales—the kinds of basic BI questions we would have an answer to immediately. We don't really have in hiring most of the time, maybe in certain situations or bigger companies that do it at a high scale, but if I asked the typical hiring manager or business what was the conversion rate per step, what's the average number of days it takes from a person to apply to get hired? What's the biggest bottleneck in the process? Who's done the most interviews? Who's the most effective interviewer? pretty basic questions that I don't think anyone would be able to answer because we just don't have the data or the systems or whatever to answer those, and so I think maybe with AI, it could automate a bunch of annoying, shitty things Maybe we'll have enough time to now think about these other problems. Maybe that's going to be the breakthrough, hopefully.

GUY ATTIA: That's an amazing approach and a great way to think forward. I don't see a reason why not to do it, though it also impacts the current managers or team leads that do all these interviews and are frustrated by talking with more and more people, and of course they want to do it like they want to find the perfect candidate on the first chart. and you also can get the feedback from them, and yeah, you can optimize these processes by a lot.

TIM: Yeah, I think you can, and what we've been describing I would say is almost a bit like data-driven hiring or at least data-informed hiring, and if I think back to the last five or six years, I've spoken to probably a thousand data leaders in that time, and what struck me was something really interesting that I'd love to get your thoughts on. and that is that when it came to actually making the hiring decisions, I would say probably 80 percent of these data leaders would use a very intuitive-based approach, a very gut-feel-based approach, rather than any data, and I find that kind of curious because these are data leaders who spend all their day running product analytics, sales analytics, machine learning, whatever. And then when it comes to hiring, they just change all of that and use intuition. Have you seen that? What's your thoughts on that, and why might this happen, do you think?

GUY ATTIA: That's totally a funny topic, and it's true. I will start by saying it's true, and most of the people, by the way, not only data but especially it's funny in the data field, still use gut feeling. I think that people often rely on gut feeling because hiring is fundamentally about trust and culture fit, which can feel subjective. When a manager is interviewing some candidates, he doesn't only think about his set of skills; he also examines if the candidate fits the environment and the culture of his team, of the company, or of the stage of the company, like between startups and based companies, and specifically in the data field, most of the jobs, besides maybe data engineering, combine both technical skills, not both, also technical skills, domain knowledge, and analytical thinking, which is so hard to evaluate and quantify. So therefore, usually there is more than one answer for an analytical question, which makes it harder to quantify the accuracy. I believe that it must be combined. Some of the tasks should be totally quantified and technically technical, while the others should represent the candidate's ability to deliver its solution in a good way. okay, for example I can present a specific analytical solution in two different dashboards, which you would say the first one is perfect, and I would say, Oh, I don't understand anything here, and the other one looks much better to me, so at the end it also depends on the hiring and the team lead or the current analytical thinking of the company. no, if it fits the strategy

TIM: Is there something to be said then for minimizing how much of the criteria are inherently subjective and focusing more on the objective bits so that you have a more consistent, accurate way of hiring, or if we did that, would we just miss out on things that are important, that are important but we just can't measure that well? GUY ATTIA: I think it's super important and you can't ignore that like part of it it's also the creativity of the candidate to think outside of the box and to show you some ideas or present you some ideas that you even didn't think of because when you let's compare it to a developer technical test okay So there is probably an optimized solution for this task, and there are like minor changes or medium-level solutions, but you can very easily put it on a scale, but whether an analyst or a data scientist will share with you their thoughts that might be like out of range, you didn't even think about it when you designed the question, right? You don't have the full set or full list of solutions that might come up in the interview, so usually, it's doing the interview; you can already feel if the candidate is going straight forward to the most binary and default solution or is also thinking and challenging you or challenging the current solution that you have in mind, which is, like, personally, this is a bigger trigger for me, like if he challenged me and he's saying something that I didn't think of because I was already familiar with the challenge, I had a lot of solutions; this is a bigger trigger for me, like to move on with him to the next step. TIM: So it's almost like an X factor that might not be in your criteria if you're writing them down, but it's something out of left field that shows they're a little bit higher than the other candidates in some ways.

GUY ATTIA: Yeah, also it's very subjective because if I, if the few team leads that are part of this hiring process, maybe I didn't think about this solution, but the other guy already thought about it, so it's not an X factor for him, but it's an X factor for me, so it wouldn't pass.

TIM: Is there something to be said for even if some of the criteria, or a lot of the criteria, are inherently subjective? So let's take, I don't know, a data analyst as an example. Maybe you need them to have some solid statistics knowledge, I don't know, some visualization skills, some SQL, some Python, but then good communication skills. problem-solving creativity So you've got like a blend of more objective and less objective criteria, but even with that, is it not possible for those subjective things that if you thought enough about them, let's say communication, if you really thought like when we say good communication, what do we mean? Oh, okay, that means they listened carefully to what I was saying. They gave a succinct answer. They gave an answer that I could understand. It was very clear, like you could probably break down what it is to have good communication by your definitions and then score people against that, so even though it's still someone's opinion over how well they did, at least it's like in a kind of box that you've reduced the subjectivity a lot. What do you think of that approach? Do you think that would work, or is that too limiting?

GUY ATTIA: So I will stick with my answer from before. I really like, as I said, to classify stuff and know the open world of the open idea into classes, and I think it will help here as well if I can just, as an interviewer, put a checkbox and my and Mark V if he thinks he's a good listener or if he thinks he understood the solution and understood the problem. So it will help me to quantify the end and, or at least know what set of skills he does have, like not technical ones, but still it will have a level of objectiveness inside it because I can interpret a person's listening mode or if he understood, I think he understood the solution or the problem while the other teammates didn't. It will stay like that. Maybe, you know what, maybe in the few years forward with AI and the chatbots that will interview by themselves, maybe they will be able to quantify it better.

TIM: Yeah, maybe I was talking to someone about this recently, and one thing we were thinking about was so there's probably, so there's definitely the moment things that we would understand and pick up on immediately as humans that an AI won't. Obviously, there's lots of those. What's interesting maybe is to think about the opposite. Are there going to be extra data points that we're not cognizant of that an AI could start to be measuring, like flickers in eyes and the movement of the head? Maybe that indicates something we figure out after years, and it's fascinating to see where this goes. Can you be an expert in this field? Can you imagine an AI interviewer in the next few years picking up on data points that even we're not aware of?

GUY ATTIA: Yeah, I think, like for both sides, first of all, I do imagine, and I think, I guess there are already experiments with AI bots that interview people. I didn't hear about, but I guess there are now. I think it's a problematic approach because at the end this person will join the company and come to work with people and not with AI bots, and also the experiment of talking with someone and trying to solve the challenge together with him. It's very much so important for a good candidate to be able to move forward, which won't be able to be accomplished or achieved using AI bots. I do think regarding your point that analysis of the AI or the video eye movement or head movement stuff like that can be an additional parameter and metric for me after the interview to raise an alert. Did you notice that this candidate didn't look into your eyes or something like that? Okay, I will say, Ah, yeah, I did notice, or I didn't, but it did affect my decision because you usually are getting the decision you decide if to move forward after the meeting so you can have a summary of the AI decision-taking and take it into account. Ah, another thing—sorry, I see another problematic cycle: if AI bots will interview only like without a human being, I can, from the candidate's side, put an AI bot for myself, and he will interview him. I can just say no to the system again, which is another problem.

TIM: Yeah, that's going to be fascinating. Oh wow, how's this going to play out? One other angle I was just thinking that one of the most fundamental issues we have in hiring is that in the interview stage, people aren't really themselves; the candidate isn't really giving you themselves. The interview is not giving you the full picture of themselves or the company; there's this little charade dance mask thing going on, which I feel like limits how accurate any interview could ever be. Now what I'm also wondering, thinking about it now, is if candidates start to interview with an AI interviewer, maybe their behavior will change again. Maybe it will be more realistic or less realistic. I don't know, but that's interesting. I find myself at the moment, if I have a conversation with Chachi PT, I find it very easy to receive its feedback. You see what I mean? Like, I ask it for feedback on what I'm doing well and what I'm not doing well, and I find it fine. I can accept all the criticism in the world from them. I find it harder to receive criticism from a human, and so I already realize that I interact with AI differently than with humans, so I'll be fascinated to see how that plays out in interviewing. What do you think will happen?

GUY ATTIA: It's an interesting question because personally I think that there is a lot of value when you see the candidates getting stressed or not knowing the answer to your question and how we react to it. It will reflect you on specific use cases doing the day-to-day job, and I see it as valuable. I know that AI can summarize it and try to do it. but I don't know if we can put more stress on it when it is or reduce the stress when needed. It needs to be more human, and maybe in the future it will, but also another approach that I noticed and I experienced in the past, which I really liked in the later stage of the process, instead of doing the last interview or one before the interview in a meeting or in one hour session, I was called to have five hours of daily jobs in the company to feel the environment and see the actual tasks that the team is facing. and also for them to like to see how I react to new stuff and how I react to the business and what the environment culture is and how good a fit I am in terms of personal connection with other teammates. I really like it. Maybe we can include this kind of step in a more narrow way. I'm thinking about it personally.

TIM: Yeah, I think anytime you have a step in hiring, the closer it is to the real job, the more accurate it is, so that's why I think an interview is not that predictive, and then you give a take-home project and then a presentation that's a bit closer, and then what you're saying is the next step, which is actually work effectively for five hours. I know McKinsey used to have the 40-hour interview idea from years ago. I don't think they do it anymore where they get you in for a whole week, which is the next end of the spectrum, I guess that's going to be Too difficult, yeah.

GUY ATTIA: I think maybe you know that maybe we can use AI and the automation to do a simulation of a day-to-day job, and then the candidate can do it on his own time. It's not affecting anything, and maybe with virtual reality you can feel the environment in the office. I have many ideas in this field, but it's far from accomplished.

TIM: Yeah, there's definitely something in that. What I think is that AI will probably start to get used from the top of the funnel down, so maybe like CV screening first and then a first basic interview, and then it might be like an interview assistant with a human and then eventually maybe the full thing. Because it's probably, I reckon, going to be more accurate at doing those basic CV versus job ad comparisons at scale than a human is, but it's going to be much worse at the moment at doing the actual Do I want to work with this person or not? high-touch interview, and so if we can just automate at least 80 percent of the bullshit, then it could be more like a value add as a sort of I'm just trying to think now, like if an AI was sitting in on an interview with someone, it could do, obviously, the transcription, the summarization, the getting ready, and the feedback notes for the candidates like that's already current with, like, current technology. It could also do things like evaluating the interviewer. Hey, did you realize you were making the candidates stressed at this point, or this question you asked, which seemed a little bit biased? You might want to think about this like friendly feedback for interviewers that, again, should already be possible with the current quality of the technology. So maybe it's just a matter of time for the applications to be built.

GUY ATTIA: I think it's also like a combination of what we discussed before; it can be a spectator of the actual human interview, like the later stage, and for sure it should help us screen, and even take it further, it should analyze and quantify the technical skills. the technical sets you can examine the test and let you know which one you can score it automatically It's not something difficult, but definitely it should take part of the hiring process.

TIM: and what about then from the candidate side of things so I feel like candidates so far have been quicker to adopt AI as we've discussed to write CVs or optimize CVs also probably to get help with taking tests either an online test or a take-home test Have you started to change or rethink your hiring process at all based on access to these kinds of tools?

GUY ATTIA: Yeah, definitely. In the past, I put more focus on the SQL query that I challenged the candidates to write. Now I don't really care about it because also in the day-to-day job, he won't be able to use whatever tool he needs in order to solve the problem. I don't care about it. I care more about the way he thinks and the way he uses the tools that he has. Right in the past, by the way, you could do it with Google Query and find the proper solution, and I say, Okay, it's fair you have all the tools that are open for you, and I even mark it down in the beginning of the test, right? I'm saying I tell them to use whatever is available for you; just use it right, and I will examine how you use it. If it takes you too much time to check with ChatGPT or search on Google or find a proper solution or debug the proposed solution, the hallucination challenge with ChatGPT can get very frustrating sometimes. and more than that Did you understand the proposed solution, or did you just copy it? So what I'm focusing on these days is giving them the challenge and then challenging them back on their solution. Why do you think this is the best one? Isn't there an optimized way to do it? Why did it do that? In addition to that, I also focus again mainly on the analytical data roles on the business side. How much value do you think it will give? what other things you could do in order to justify the business KPIs or whatever

TIM: I'm wondering if you've noticed a tendency with candidates who've used ChatGPT or any large language model to at least do the coding bit. Do you ever get the sense that they are now almost palming off the responsibility? It's Oh, this is what Chattopadhyay gave me as opposed to them needing to own the fact that yes, they've used a tool to generate it, but ultimately it's still their responsibility to write it. Have you noticed any maybe more junior candidates start to slip into that mindset?

GUY ATTIA: Not really. I think they are shy. They are too shy to say that yet it's convenient to use it, but to blame it is another level, and at the end you are responsible; you are the answer writer, so I didn't find it like no.

TIM: And in your view, is there a difference between using something like ChatGPT on a skill that you don't know at all versus using it on a skill you actually do know to make yourself more efficient? So, for example, a candidate who'd never written SQL using it for SQL versus one who's an expert in SQL.

GUY ATTIA: So it's again the same; it's quite the same answer from before, because if I use it for something that I don't know, I won't understand the solution, and if I don't understand the solution, I can't explain it, and I will catch it over there, but in any way I do expect them to use the tools that are optimizing If he tries to solve it without Google or any LLM model, it will take him 15 minutes, and he could solve it in one minute by querying Google. I will ask him, Why didn't you use the tool that you have? Because you're wasting time; you're wasting money, so it's also part of the things that I'm checking. Did you use the correct tool in order to do it? Maybe if it's something that's technology that you're not familiar with at all, maybe you will need to start by reading the documentation of this tool before going to ChatGPT or the solution itself.

TIM: to ask you about is If you had the proverbial magic wand, it could be AI; it could be any other kind of magic you like. If you could, with the click of a finger, fix the hiring process, how would you do it? What would you do?

GUY ATTIA: Oh, that's a nice thing to think about. I'd probably create a system that seamlessly integrates skill-based assessments. Automated candidate matching and real-time feedback while eliminating bias, which is most important for me because I am very sensitive to this topic, will also offer continuous learning opportunities for hiring managers and candidates. Both sides will benefit from it. The manager will learn what you need to do better in order to hire or to interview the next person, and the candidate will learn where he felt wrong or where he answered wrong, and then he will improve his skills, and we'll have better chances in the future. It will ensure alignment with evolving industry needs because it's getting out there for bigger companies and smaller companies to hire people and know exactly the set of skills and the seniority level of a new recruiter.

TIM: That sounds like a great use of your wish or your magic wand, and I feel like we can't be that far off that I don't think that's fantasy. I feel like that's 2025-2026 products. I feel like we're already pretty close, so I hope you get your wish, and your magic wand wish is granted. Guy, it's been an interesting and engaging conversation today. We've covered a lot of ground. Thank you so much for joining us and sharing all of your insights with our audience.

GUY ATTIA: Thanks a lot, Tim. It was interesting, and I had a lot of fun. Thanks for having me.