Episode 79

Noa Lubin on The Future of AI in Recruitment and Education in 2025

TIM: We are live on the Objective Hiring show, our second show for 2025. And this evening I'm joined by Noa. Noa, thank you so much for joining us.

NOA: you for having me.

TIM: It is absolutely our pleasure to have you for 2025. I'm pumped for this year. I think so much is happening with AI and all sorts of other things. I think it's going to be a year of great change. I'm personally, personally feeling quite optimistic. And I'm really excited to chat to you. to begin with about the book that you've written. It's not every day we get to speak to a published author. And it's a really interesting topic and theme, very topical. And I think, especially for our audience members who've got children, I think it's going to be really interesting to hear about your book and what prompted you to write it.

NOA: Okay. I didn't expect you to start from there. So yeah, so, I wrote a children's book about AI. I have a five year old daughter and a two year old son, and my five year old daughter is really learning from books. She was always wondering what mommy does for work, and aside from the fact that I sit in front of my laptop I couldn't really explain it to her. as a a side job, I also lecture and about AI and I love doing it. then all the stars aligned and I was like, this is what I need to do. I need to write a book to explain children. not only how to use AI, but mainly what are the concepts behind AI so that they can really understand and be mature about using this technology. And hopefully maybe some of them can develop the future AI solutions once they become adults. And it was really successful. The, it was a headstart project. A kickstart project and 1200 books were sold like over two weeks or so. And most importantly, now my daughter understand what I do for a living and the concepts of AI and it's amazing to see she's only five. And, you know, we have the explain like I'm five library and machine learning that gives you the explanations of the model. So I feel like I I revealed for her how concept. that are very basic in our domain work. So that was really fun.

TIM: That's amazing. And how, how did you take something that is potentially so complex and make it easy enough for, for kids to understand used like analogies or, or yeah, how did you go about simplifying it?

NOA: Yeah. So I did use analogies and sometimes simplifications, but I must say I didn't cut them slack at all. And it's amazing to see how they capture the concepts. It is much more processed and explainable. but yeah, a lot of analogies, a lot of. Items from their day to day simplifications, like I said, but the children understand the concept of a language model of machine learning, learning for, from examples, regression, logistic regression, things that, you know, decision trees, things that I thought would be impossible for, for her age to understand. It was a huge challenge for me. I, I would say it was very challenging. I think it was the hardest teaching I've ever done.

TIM: Yeah, well, I think it would be because you have to distill something down to just like the simplest form. It's almost, almost like a tweet in a way, like that's the, the perfect communication is something that's very simple, straight to the point, but still leaves no ambiguity. It's, it's quite a, quite a difficult challenge I would have thought. NOA: Yeah, I did use a lot of infographics, so, you know, visualizing a lot of concepts also helped there.

TIM: and what's the book called?

NOA: it's called AI for All. Currently it's in Hebrew. I have it right here. I'll show you

TIM: Oh, nice.

NOA: booklet and it's in Hebrew. It's very, very colorful.

TIM: And it will be available soon in English. I understand.

NOA: yeah, so now I'm translating it to English and it will be available soon on Amazon.

TIM: Excellent. Well, I've got at least three little nephews who are going to receive that as a late Christmas present whenever that's available. So I'm excited to see how they will enjoy it.

NOA: Thank you. I'm excited to hear they did.

TIM: and what about then the world that we're going into? Like what, what, what can you imagine your, your kid, you've mentioned your daughter's five. Like, what, what is she going to be doing in 10 years? What are her skillsets going to be? How, how is life going to be so different for her than it has been for us?

NOA: Wow, this generation is so much smarter than us. No offense, Tim, they're

TIM: None taken. used to, you know, technology before they can even speak. I think, you know, using AI is a given. They will use AI. For everything for coding for doing their homework for now my five year old. She's communicating with the phone via speaking.

NOA: She's using speech to text without even knowing when we're entering the building. She puts her face and the camera recognizes her and opens the door. So, you know, a eyes in her day to day and she she's feeling it. She's excited about it, and she's using it frequently. I can't wait to see what the future holds for this generation. they know how to use, work those tools. They know how to make advantage of them, save up time. They're not afraid of them. So it's gonna be super fun to watch.

TIM: Yeah, and who knows how it's going to play out. But yeah, being an AI native is going to be to your advantage, I think. Because you would just take it as almost like an extension of you, I guess. It's, it's like where we're becoming almost like cyborgs, I guess, as we have these technologies and they're just a part of us, in a way.

NOA: Yeah, I, I agree with your vision.

TIM: Well, hopefully we go down the good path and not the terminator path.

NOA: that's where I'm a bit concerned. I mean, we as humans, we can take it to do good and we can take it to do bad. I really hope as humanity

TIM: yeah,

NOA: to the better directions.

TIM: yeah. Yeah. Even if I think about my nephews and the way they use technology broadly, there's such a yin and yang good and bad to it because they sit there on YouTube watching a lot of content. A lot of it is crap. But a lot of it isn't so my eight year old nephew is explaining to me what a red dwarf planet is. I don't even know that. And he's learned that from YouTube and he knows how to do all these specific things because he's just had access to such a vastly larger amount of content information that I ever did. So there's all this upside, but also there's downside because then you see like the attention span is a bit lower scrolling through this YouTube shorts every three seconds. And like, what does that do to your brain? So I'm, I'm personally so torn on kids with technology. What, what, what about you? Like, are you, are your kids on the phone for hours or do you have a sort of limitation?

NOA: Yeah. I really feel what you're saying. My daughter, all she wants in the world is a screen time and candies. Those are the two things that make her super happy. I think, you know the, the she does see a lot of, like you said not educational material. They're more exploring than I think our generation. Like, it also goes with the I, I forgot how you say it, but the, the fact they cannot concentrate on the same task for a long time. I think they can, they can also not go deeper into things maybe than, Our generation, like speaking broadly, but their exploration is much quicker. Like they can identify good, like relevant content from them rather than non relevant really quickly. yeah, they, they can, sometimes she watches them in like two videos at the same time. I don't know if your nephews does it, like the YouTube splits up into two and she, I don't know how she's, her brain even processes this. Yeah. And obviously they, a lot of kids cannot stand still any longer and sit in the classroom listen to the teacher, in the past.

TIM: Well, that would be exciting if this technology created an evolution in our brains that we could multi thread. That would be cool.

NOA: That

TIM: Okay.

NOA: that will be really cool.

TIM: what about then bringing you back a little bit more to a shorter time horizon, a little bit more into the work environment? Like what do you think then is going to happen in the next couple of years with data roles, given how quickly large language models are developing, like what is a data analyst going to be doing in two years versus now what's a data engineer going to be doing? Like, how do you think those roles are going to develop?

NOA: Yeah. Those were really evolved. I think since the LLM revolution you see more and more titles of LLM engineer. Prompt engineer and, you know, positions that weren't there. Just a few years ago, I did my NLP thesis in 2017, when the whole revolution started with the Attention is All You Need paper that was back then, you know, just a paper. And I experienced it myself, like how my field is experiencing a revolution throughout my thesis. And I think it will leave more time for other things. So if I'm talking about the data scientist, more time for modeling the data, evaluating and some projects that would have been closed with a whole team of data engineers, data science, and data analytics maybe can now be reduced to one person writing a prompt with a little bit of rag, and data engineers, I think the other way around, like now we need them more than ever. more data to process, the vector databases their role became even more significant. And as for a data analyst, also, they now have the ability to automate their query writing produce much quickly visualizations. But it's, for me, the data analyst at the end, good one department strategies to become data driven. So still the interpersonal skills off, you know, looking at the data, explaining the data, making sure that the stakeholder takes the decision based on the data. soft skill is still needed. Hopefully again, it will be more time for them to focus not the technical SQL writing, but more of the rest of their work. But I would say in general, data becomes more and more and more central in companies. So you see now and departments in every company. Sometimes they are very central. Sometimes they are something on the side, but there are everywhere. So that's nice.

TIM: Is there also something to be said for maybe non data roles becoming a bit more data-ry in the sense that now if, I don't know, a marketer or an accountant or a executive now has access to this groundbreaking technology that could be like a gateway into them doing analytics. Now, there's the kind of almost more self service analytics as well. Everyone becomes an analyst, do you think?

NOA: yeah, I, I seen more with like the product role. So I just had an AI product workshop with product managers that are building AI tools. And also the usage of AI tools. Like they can build an M a mock or an MVP without even coding, you know, just describing what they need. And they have this website up or this app up and running. We have the coders at Fido that, you know, they're writing code with co pilots and, and other AI tools that, you know, The the director of engineering just got to me like we have a full pull request that was written by an LLM and the the the question was like, how do we review it now is one human reviewer sufficient or we need two and is the future, you know, having a different agent as a reviewer. I think it revolutionizes all of the tech departments. And like you said, even outside of tech, everyone, we gave training in Fido for our customer. experience team. So how you can use chat GPT or any other LLM to make your responses to our customers more efficient. You know, it's, it's everywhere.

TIM: Yeah, the upside seems almost limitless. The other thing that I hear people talking about it would seem to happen. I would think it would happen would be that. AI may be a very deflationary effect on technology and software, because if the kind of marginal cost of producing software is coming down so rapidly, as you say, if an LLM can produce an entire PR successfully, get it passed, get it code reviewed, then you can spin up software so much quicker than you could before. So what is, what is that going to mean then? Like, yeah, we can do things much more efficiently, but We can't charge as much money, probably for the software as we used to, I suspect. And then we're gonna have to have way more customers to make the same amount of revenue. Any thoughts on how that's going to play out?

NOA: Yeah, it's an interesting point. I guess like always your product needs to meet a certain real need for developers, for people. Once you give value to the people and they can't live without your product and they will pay Whatever I think for us, you know as entrepreneurs or people working in startups and it just makes Our lives easier because we don't need maybe as much human power as we Used to well, we can develop things much quicker so that i think it's a good place for us

TIM: Yeah, and even for, as you said, the data roles that are obviously going to change and evolve, I feel like so many decisions in the average business currently aren't really that data informed. So I feel like there's so much upside for making better data driven decisions. If we had access to more data sets, if we had some tool to do all the analysis automatically rather than relying on an individual analyst to do it. I feel like, yeah, the kind of aggregate demand for data skills surely is going to go up even if each analyst is now much more efficient. What do you think?

NOA: Yeah, I agree with you I feel like You know, this is my mission as the director of data and AI at FIDO, like to make all of the departments, even the operational ones, or, you know, the really non tech ones, more data driven and, and now that they have the tools within their hands, they, they can be a boost to, to to the data departments as well. It's a challenging question, like how you scale it? Also outside of the data and still make sure that it's the quality of the of the deliverables you intend, because making decision based on data, but the wrong data or the wrong processing is even worse than, you know, making it based on intuition or something else. So we do need to. I think we will be able maybe to scale. But we need to make sure that the quality is still, you know, the best

TIM: I'm One thing I was thinking about on this recently was maybe there's a bit of a trade off where it's like at the moment. Let's say you've got non technical non analytical stakeholder, you might go to the BI or analytics team. Ask for a report, ask for a dashboard, ask for a bit of an analytics and, you know, you can only do so many bits of analysis. It's only so many people's and so many things you can ask for. Probably that stakeholder is only asking for like their top couple of requests. Whereas if they had some kind of like AI analyst, they could ask anything they wanted to. Maybe they'd just be asking them a hundred things a day rather than one thing a week. And so even if the accuracy is only 80%. Better to have 80 percent unanswered than no answer at all. And so I suspect we'll be in this middle ground initially and then chip away at the accuracy through time, perhaps.

NOA: Interesting. Yeah, I can see that happening. But let's wait and see.

TIM: Indeed. Yeah, we can make as many predictions as we want, but nobody is Nostradamus. What about on the hiring front? Have you personally started to dabble with AI? In hiring, for example, to create job ads or review candidate CVs or anything like that. And have you seen candidates on their side start to use AI at all in the hiring process?

NOA: So in terms of me using AI, we say in Hebrew that the shoemaker walks barefoot. So it's like an expression that I'm not using AI to, you know, to screen CVS or to write the job descriptions. I, I, you know, in, in, back in 2016, Amazon published the CV screening tool, an eternal tool to, to screen CVS within Amazon. they had to shut down the project because it was heavily biased towards discriminating women. So yeah. So when done needs to be done very carefully not to induce biases. I still review CVs manually or the recruiter reviews them manually with the job description. Maybe I just use it for, you know, spell checks and things that I haven't missed in my English. I will say in Hebrew, it's, sometimes publish in Hebrew, not like the job description, but like in, in Facebook groups, et cetera, the, the positions that we have open, we have now five open positions. And in Hebrew, there is a challenge of writing in a general neutral language, because it's a very genderific language. I wrote a, a, a chat a custom chat that you put a text. In any language, let's say English or non general neutral Hebrew, and it converts it to general neutral Hebrew because it's very, very hard. You're like, you need to really stop and think and, you know, use a lot of tricks so it can be read like both genderless female and male. So that's maybe when I do use AI to write my posts on Facebook and LinkedIn for recruiting. As for the candidates They use AI, you know, LLMs heavily so I know like in the past I was in companies that used to do home tests and we at Fido we do give on site tests to the candidates. I take into account that those tests will be done with LLMs. So at that point, like, you know, I don't want to look at your coding level because your coding level is probably chat DPT level, which is amazing. I'm looking more of like, did you understand the data? Did you know what we want to get out of it? Did you understand the problem? Which tools have you used? But, but yeah, it is challenging also in the teaching, like how, how do. Like I give students an assignment to write themselves a decision tree algorithm, but they can just Give it like a get it by chadu PT, and then we missed out the whole point So I think like we need to adopt ourselves as interviewers as teachers as whatever to the fact that you know LLMs are there in their work as well So if they can get a very good work done with LLMs, I'm fine with it. So I need to like shift we, how we test the candidates and in my opinion, you know, just let them use it if I want to test skills or, or SQL skills or, or data science skills. I, I verbally and one on one interview, like a small questions you can answer online. Yeah, but, but it is it is harder, especially with the remote interviews that, you know, one can just, yeah. Even open the chat on the side and and type the question you just asked.

TIM: It's so tricky, isn't it? Because it's like this technology has improved so rapidly. And candidates have adopted it so quickly and certainly quicker than companies ever could because they're individuals and they can just Do whatever they like. They don't need to wait for software to be built using the technology. They don't need to wait for approval or anything. They can just kind of do it. And it feels like we're in this weird in between stage where the hiring process has kind of been a little bit broken because as you say, you're giving someone a coding test, what does that even mean anymore? Because yeah, you don't want to know Chachapiti or Claude's ability. You want to know their ability or do you, because maybe in a year they wouldn't need to code at all. And so we're in this kind of. Hazy in between stage. I wonder if a year or two down the track, it'll become a lot clearer. Maybe the technology kind of settles into some almost normality where it's like, okay, we now understand these are things we will never do again ever manually. This is 100 percent AI. That's it. It's, it's domain. Nobody's gonna, I don't know, write code from scratch ever again. Or maybe they will. I don't know. But I feel like we're just in this. Confusing in between stage. We just need to get to the next point and then we'll know what to do.

NOA: Yes, I agree

TIM: and what about for the candidates? Is there any bit of the hiring process that you would really rather they not use a large language model? So you kind of accept it as a given. They're going to use it to do coding. Fair enough. They've probably written their CV partly with it. There's anything you would prefer they not use it for.

NOA: when it's a I have a remote interview a candidate because we hire Both here in Israel and also in Ghana. I remotely interview a person in Ghana and And I speak to them and I look into their eyes I don't want Chad GPT to be in the room with me. And unfortunately I did have events and you know, trust in the work in every relationship, but also in a work relationship so important that also for the candidates, not worth doing like an interviewee can recognize that when something Is weird that you pause too much to think that you muted yourself too many times that the internet went off too many times, you know, like it's stuff that we we we read between the lines and it's just not worth it. I want to know how you think, you know, at the end, an interview is to see that we both fit also the candidates fits the requirements and the culture and whatever. And that, you know, I see him fit as well. Yeah. It's both sides need to, to learn it through the interview.

TIM: Yeah, I think I'm with you that using it in a live interview, if you're saying you're not using it, so you're kind of not meant to be using it, but it's clearly there is probably not a great idea for candidates. I think because if, if nothing else, it would be to their disadvantage, unless they really knew nothing about the job and they had just applied hoping for some weird chance, they'll get a foot in the door. But like, if they actually know their stuff, I think it would be more difficult to sit there reading the output of a ChatGPT and trying to think, hang on, how do I put this into human words, as you say, muting themselves, all this kind of bullshit tricks, it's probably easier just to face the interview.

NOA: a i never understood how to cheat on tests in school, in university, especially not in an interview. So for

TIM: Yeah,

NOA: Then you know, wrapping their heads around how to do this stuff, it's like Just answer, you know, s like, you know, sql you don't need,

TIM: yeah.

NOA: to, to, for, to answer a joint question. You don't need chatGPT Like just stop being, become and you'll get it. You don't need more than that

TIM: Yeah. And let's, you know, thought experiment on this a bit more because I see in the current state, it makes no sense really to try to use AI during an interview. And some people have said, well, yeah, because once you get into the job, yes, you will use AI while you're doing your analytical task at your computer or with your phone or whatever. But if you're having a conversation with an executive, you have to actually speak to them. You can't deviate to Or whatever, like you need to actually have something to come out of your, out of your mouth. That makes sense. But how far are we away from maybe that not being the case? Maybe we have a wearable LLM device. That's okay. There's like a neural link. That's maybe 10 years down the track, but even before then, maybe we have a device that's just. helping us in real time with real humans and then maybe it won't be so much of a problem. I don't know.

NOA: I don't know about you, but when I brainstorm about a problem, I like to brainstorm it with a human. I, I can also brain brainstorm it with chatGPT and sometimes I do it, you know, but, but, I think, like I said, with the data analyst role, I think for every role there's the human aspect of it. Connecting the dots in, you know, areas that weren't connected before. Convincing the, so the data engineer convincing the backend engineer to go a certain way, that a lot of the soft skills. Are still needed at this point, maybe, you know, 10 years from now, they won't be needed anymore. And if I want to hire a chatGPT I don't need to hire anyone. You know, I have it with my premium enrollment to, to open AI.

TIM: The 200 a month well spent potentially. Well, that's a good point then, isn't it? So the soft skills are sometimes called human skills, which obviously are needed between humans, but Maybe as more and more work is done by large language models, those assumptions we'd made about like stakeholder management and convincing people and selling like of all those decisions, just being made by an algorithm, then there goes the soft skills, I guess.

NOA: Yeah, that'll be, that'll be very interesting.

TIM: what about from the hiring process? Is there any, any bit of that process that you would love AI to, to do for you, anything where you think there's like a big value add, because it'd save a lot of time or make a better decision. What do you think, you know, that tech's going to go in the next couple of years?

NOA: yeah, I think the hiring process is very time consuming and what maybe I don't like about it is, you know, it's a very precision oriented process. At the end, you want a candidate that you, it's the perfect candidate for you. It's not a recall oriented process. You miss out on a lot of great candidates. this process. And, you know, using this platform to say to a lot of people. I'm sorry, like, I'm sorry that we never got to your CV that you never got to an interview like I know that we're missing so many great people out there because of the way this is a funnel. And, you know, you always prioritize with to interview today because the schedule get busy. So, yeah, I think, you know, I do believe Despite what I the anecdote I said about Amazon, like a CV screening tool or a good test is somehow not cheatable by an LLM, maybe can help, you know, reaching those candidates that we sometimes miss out. And yeah, and I think, you know, the CV writing is already there. Like, I think it's very useful and I, I do recommend. Everyone to use it when they write their CV and to adapt the CV to the exact job description of the company. And job descriptions, yeah, we can use it as well to check the test. So if we did give a test to check the test, I've used it. So like if I didn't have time to go over the code, just for the call, give me the notes. And then it makes my life easier where to concentrate. So it is speeding the processes now. Yeah, but, but in terms of, like, changing things, I wish it would help me get to the candidates I, I'm missing out on.

TIM: personally feel pretty optimistic. And I, I think that with the current state of LLMs. A lot of the hiring process could be automated away. I think humans often maybe do more damage or do more harm than good in the hiring process. Not necessarily through any maliciousness, but just because we're so biased inherently in so many different ways that are. Basically impossible to remove no matter how well you're trained, just, just in your brain. And I think in theory, a well designed AI system could solve a lot of that stuff. For example, there's some really interesting studies that have been done around measuring. Discrimination against people from different backgrounds based on their name. So these studies that apply on mass to hundreds of different roles and then measure the rate at which CVS get called back where the only difference is just their name. And it's really staggering to see how much discrimination there is in different markets. So yeah, the, the Amazon CV screening example is interesting because they. Yeah, I think from memory, they were trying to optimize for similar profiles to what they already had in their company, which was predominantly male dominated. But maybe a different approach might work better. So I'd love to ask you about this actually, because of your expertise here. So if we're talking about, let's say a CV screening tool, if one basically approached it as, hey, get the CV compare it to the job description. extract all the relevant bits of information, come up with some kind of matching score, bang, here's the top matching candidates according to what they've said on the CV. That's not really about picking I want a similar candidate to this set of candidates. It's just using the text. Would that kind of approach be a way to do it in a quote unquote biased free way? What do you think?

NOA: think, you know, it, it doable, and kind of does it now with a lot of positions. You see, like, the ticks the requirements you've met. I think maybe what the candidates don't know about us, that we sometimes as the people who hire Are not exactly sure like what are the parameters the exact parameters we're looking for? So the job description, you know what one candidate can tick all the boxes but you know that you didn't even know that before that you were only looking for certain universities or That you were lacking in your team and expertise in a certain area, and then, you know, of time in this process, we kind of iterate on what exactly is it that we're looking for. At least on my end, I wish I was more perfect and, you know, know from the start, this is what I want. But if we have all the criteria and we know exactly what we want, then definitely it's a very easy, not easy, but a very doable algorithm to do that matches the CV to the job description. You mentioned the biases and the example of We talked about Amazon, so there's also the John and Jennifer research, you know, in Yale University, have you heard about it?

TIM: I haven't.

NOA: So it's a very cool research So they gave a bunch of students CVs to review they asked the students Would you hire that person into specific position and what would their salary be? And they divided the students into two. One received a CV with the name Jennifer. And the other one received the same CV, same last name, with the name, a personal name, John. And not only did John get hired more frequently, he also got a much higher salary

TIM: Lucky John.

NOA: So, you know, you said about those biases and of course we'll need to really mitigate them in the algorithms as well. And removing the name doesn't cut We need to do much more. But I think at the end, like I'm going, if we know what the features are, like what exactly we're looking for, then yeah, it's very doable. It can be very helpful. I feel for me, you know, there is, there there is never the ideal candidate, even if there is the ideal candidate, you find out that something is still missing. So that's when it gets trickier. Like if it's not a perfect match, like what is the distance metric that will still make me satisfy? But, but yeah, I think HL tech is there. And, and it's, I, I do, I'm like you, I do see like. Very few years from now recruiting process end to end. Like maybe the last interview, just, you know, because we're still communicating with humans nowadays, still, like something to check, like, who are you working with? But they passed, you know, a very basic test, a coding test you know, they matched everything if the recruiter know what they exactly want. So that's the key thing.

TIM: Yeah, that's such a good point and one we sometimes neglect. So we think of maybe on the candidate side, the CV being a pretty weak data set to make a hiring decision because it's document they've written about themselves. Now it's a document AI has written about them. And so the feedback we're getting from people in the market is a CV seems to be even less predictive than it used to be. Of whether or not the candidates any good, and it used to be terrible. So I can't imagine how bad it is now. So that's like one side. But as you say, on the other side, it's that job description is only tells you so much. As you say, also, sometimes it's iterative process through time to kind of refine what exactly you're looking for. So like with any data system, isn't it? If the data is crap, then it's going to be hard to have a good algorithm. So we need to fix the data, I guess. But. What would that look like? Like I was talking to a few people about this going beyond the CV, like, are there new data sets that we really need to unlock about a candidate to have a better prediction? For example, you know, Extraction their task management system for the last three years of their work or their real projects, or I don't know, like conference material that they've published that we can now analyze with a large language model or something like maybe there's some unlock there on the candidate data side of things. What do you reckon?

NOA: Yeah, I, I'm, I'm really up to what you're saying. I do know a lot of colleagues that. Have as a very initial step a screening test, so that kind of, you know, collecting data to make the better decision instead of the CV. Some, some of them even don't ask for a CV, just, you know, complete the test or do this Kaggle thing. And the top candidates, I will interview them. So I think the one hand, it's a very like fair starting point, I would say. On the other hand like at least with my consultation with the people in Ghana, it's very prone to cheating. So, yeah, I think, I think, but, but I think, like you said, I didn't, up until you said it, I didn't think about collecting, you know, true data the candidate, not just, you know, a test for the sake of this interview. So there is the GitHub repository that sometimes we're being asked to provide as candidates or, or it's in your CV and then we review it that can be a good signal. LinkedIn profile to see the extracurricular activities. Yeah, it could be, yeah, maybe, you know, like in FIDO, we ask our clients to give access to their data because we give We give credit to we give loans to unbanked in africa, so we don't have transactional data about them What we do have is access to their phone data and out of this we try to predict who will repay back the loan or not I will say like it's a very like So you do see a do you see candidate like giving you Full access to everything you have maybe maybe if you're google or like Corporate that people really want to work in.

TIM: Well, that's really interesting because you've just given me an idea there. So yeah, so you're saying you collect phone data and that's somehow I'm not going to ask you exactly the details how that some it helps you predict whether or not someone's going to pay back a loan, which someone is an outsider is like a non obvious that there's a correlation there. So then suddenly I'm thinking about recruitment. Yeah. Like what, what would be some kind of weird proxy for. on the job performance. Like if you shared your, I don't know, Strava running data or your health app data on your phone or God knows what else. Yeah, the probably find some strange correlations between job performance and other things out there, but it's also sounds like a data privacy nightmare. So yeah, we'll, we'll leave that one alone for the moment. You'd, you'd mentioned in, in passing that some of your teams based. In Ghana and some in Israel. I'd love to hear a bit more about like some of the challenges in hiring across those different markets and what you've noticed as some of the main differences.

NOA: And first of all, it's very different recruiting in Israel and in Ghana. We do have a office in Ghana, so like technically the process could look the same if we wanted to. But but although the we're looking for similar qualities, let's say now, for example, I have a data analyst position open in Israel and in Ghana, and we're looking for the same, you know candidate more or less. think the first difference is the difference is the CV screening. You know, things that will in Israel be perceived as, you know, red flags or things to avoid. Can be like completely advantages in Ghana, for example, the role switching. in Ghana, it's very common that you're a mobile engineer that moved to back end and later to data and then moved back to being a tech lead back end engineer. So The role transition there is much more common, I learned from my data set and you know that this signal is actually identifying good candidates in Ghana, where in Israel it might like be perceived as you know someone who's uncertain about their career and usually they're very like immature in every field. So the same signal in the different countries led me to, to, you know, different candidates. Also, you know, of course, there's a cultural difference that you need to be aware of when you interview because when you interview also look for the interpersonal skills. And I will say that because we are A hybrid group that communication skills for me is like, or the interpersonal skills even more than the important and technical skills. I do want, you know, a certain level of technical skills, but I would say like 51 percent or more than 50 percent is also the communication skills, identifying that it's someone that can work with people from a different culture than theirs. And, you know still be able to embrace it and, and, and work together well. Yeah, I, this is on top of my head.

TIM: And I would say at least in my experience, the stereotype is Israelis would be considered quite direct communicators, I would say on average, that's at least what I hear. And maybe Ghanaians are different. I'm not sure. And is that part of the challenge?

NOA: Yes, it's true. Like Israelis are much more upfront. And with my Ghana employees, I need to work harder to get their genuine feedback, if it's a negative feedback, even more, and they're, yeah, it's a, it's a different culture in so many levels and even I don't know if to give this example I will give it, like, even wishing someone happy birthday in Ghana. You can later learn that in their tribe, they don't celebrate birthdays. not that this wish offended the person, but you know, a lot of assumptions, you need to delete them and really learn the person that you're talking to and working with.

TIM: Yeah, it's amazing how all these things do vary by, by culture. I know I've got a few A few friends from Eastern Europe when they moved to Australia, the first thing they commented on is like, why are strangers asking me how I'm doing and how my day is going? Why are people talking to me and lifts and smiling at me? Are they unwell? They just, they couldn't really comprehend that. So yeah, it's not right or wrong. Like if you asked a Russian how they're doing, they're going to give you a really honest answer. They're going to actually be truthful as opposed to some, Oh, I'm doing fine. Thanks. Kind of response. That's it. Yeah, it's yeah, always amazing to see how these things vary across markets. Actually, I was reading a book just at the moment, this one, I'm not sure if the camera can pick that up.

NOA: Yeah, I see.

TIM: Which is a really good one. And yeah, the author Kim Scott was talking about managing cross cultural teams and the difference between managing their American and Japanese team when she was at Google and she, she needed to really. Like encourage the Japanese team to give honest and open feedback and she had to really double down to get anything out of them in a way that that she didn't necessarily have to do with the Americans and it was interesting. Hearing hearing her stories. So yeah, that's sounds like a challenge, but an interesting one. What about managing across these different markets? Like not just the hiring aspect, but managing particularly in these kinds of emerging markets, any particular challenges you've come across?

NOA: I think the talent there is maybe harder to find. But once you find, find the good, like the, the great people, they're like an unpolished diamond. You just need to polish them a little bit and they really, really shine. Yeah. So I think, you know managing is harder. It's like the. The advanced class of management and I have a lot of tips on how to to improve remote managing, but obviously the challenge is that I don't see the person. I don't see how they feel today. I don't know if even working today or what they do. And the cultural difference that we discussed. So, I think, you know, Corona and the fact that we all became remote managers at some point really helped speeding this up. And I do have a lot of I do in my day to day to help this. One is like being, being very transparent and accessible. So, you know, have my calendar open. If someone slacks me or want to huddle, I do it straight away. Second thing is to overprioritize the people that are not in the same office with me. So now if I have spare time and I want to check in on a project or to check in on a person, do it at the person who sits right next to me, but actually, you know, accessing them and asking them questions. Visiting Ghana, so that's also fun and also very important to build a relationship and hopefully them visiting us at some point. And yeah, so I think, you know, it's a lot opening the cameras. Maybe it's trivial, but a lot of seen a lot of managers and a lot of employees that communicate with closed cameras. I do want to see their face. I didn't want to see, you know, there's so much the body language and other things that, you know, we want to see as managers and and avoid hallway conversations. So avoid like, Closing decisions. Now at the office where we're on a coffee break. No, if this person is a part of this project, it should be in a meeting or a slack channel or something that you know is more. Everyone had the same opportunity to give their opinion. It is harder. It's a hard challenge. I said it's management advanced class.

TIM: Yeah. And I can imagine the hybrid scenario does add an extra layer of complexity, as you say, to make sure that everyone's. I'm not treated equally, but everyone has the same on the same playing field as much as possible. That must make it especially hard. I would have thought one final question for you today. Noah is if you could ask our next guest one question, what question would that be?

NOA: I have a very specific question that has to do with hiring. If this guest had experience with hiring multiple candidates to the same position at once, what do they think of, like a boot camp strategy? So, like, to hire a lot of people at once, maybe they don't, like, perfectly qualify for the position, and train them so that, you know, after weeks, they become ready for the role. I was playing with the idea myself and I would love to hear thoughts about it if they have any suggestions.

TIM: well, Noah, it's been a fascinating conversation today. Thank you so much for sharing all your insights and wisdom with our audience.

NOA: Thank you so much for having me. I had a great time and you know, you opened up my mind to so many ideas about recruiting and in general. So thank you.

TIM: Thank you so much.