Episode 34

Husain Mustansir on Leveraging AI for Efficient Hiring and Business Operations

TIM: Husain Welcome to the Alooba objective hiring podcast. Thank you so much for joining us.

HUSAIN: Thank you for having me here.

TIM: It's a pleasure to have you, and I would love to dive in straight away to everyone's favorite topic at the moment, which is AI, in particular its impact on hiring, and so I'd love to hear your thoughts and experience of what AI might already have contributed to hiring, either using it yourself on the business side or maybe some impact you've seen from candidates using it during the hiring process. And then I'd love to also deep dive a little bit further on the screening stage in particular and have a bit of a discussion there around the usage of AI in screening candidates, so to begin with, I guess, yeah, have you used AI at all in hiring so far?

HUSAIN: Yes, as an agency, it is our objective to use AI in as many business functions as we could, so we keep on experimenting with different tools that come up or the different features we particularly use on specific platforms more than others, like, for example, ChatGPT. So we have been with ChatGPT since it launched, and every new update we keep testing it for its different features. Then we have automated a lot of our internal operations as well using these tools. It's still, but like what I've seen, I think with what many of the other leaders and many of the other practitioners in the technology field must have also noticed is that AI is not the answer to everything right now, and it still has a long way to go before it can be used as a proper business solution. Other products out there in the market, which are like established, which predictably do exactly the same thing in and out hundreds of times, but AI is yet to be there, so in our hiring process we have started using AI, but it's still—we take it with a lot of salt, I would say not a pinch of salt, so particularly the particular area in the hiring that I use, I like using AI to formulate job descriptions, so usually what we have seen is—and we are a very small agency, small as in lean. So we don't do mass hiring, and the structure of the company is that it's spread remotely across the globe, so it's 100 percent remote, and all our work gets done through the cloud through automation, so most of the heavy lifting happens through automation and the cloud, so it becomes very important for us to, when we source a candidate from the market, We don't just put in textbook descriptions out there or textbook job titles and say we need a data analyst, we need a technical marketer, we need a data engineer, or a web developer, so we have to tailor it to be very specific as to this is the kind of skills that we're looking for, and they need to have so, and it needs to be very detailed. and the problem with that is that what I've seen, I've experienced not in this company but in my previous companies as well, where I was part of the hiring process, I was part of the interview process, and I was actively doing interviews with candidates. I was actively hiring, and I think I've been supporting this recruitment function for I think more than eight years now in different roles in different companies. What I've seen is that it's very difficult to come up with a job description for a particular job role. AI can help with this tremendously. You put in a prompt, you get the job description, and then you go through it; you edit it to make sure that it conforms to your requirements; you maybe elaborate on a little more topics. and then you put it out, it gives you much better results rather than putting a boilerplate job description saying I need a data analyst, and they need to have XYZ capabilities, need to know statistics, need to know data wrangling, and things like that, but those are pretty standard stuff. What do you really mean by data wrangling? What do you need as part of this particular job or as part of this particular company? Different people do different stuff. What kind of tools experience do you need? Yes, AI helps them.

TIM: And does the value add that it is because the job descriptions would just take a long time to write manually? Is it that it expedites it, or is it going into more details about the different things that you need and elaborating because I guess it doesn't know what you need until you tell it in a sense?

HUSAIN: Exactly in the beginning I said it's not like a solution to everything, so there are a few schools of thought that are developing around using AI, especially among the leaders, like the people who are not hands-on with the AI right now. One is they think that they really don't understand this technology; it's like alien to them. So what's going on? We are better with traditional stuff, so they would put a person and say, You have one hour to write this description, two hours to write the description, go on the internet, search, prepare a word document, and then show it to me, and I will review and sign it off. So those are the people who are like, Afraid, and I don't know what is going on, and I'm a bit overwhelmed by what is happening. Then there are other sets of people who say it's a silver bullet: you just tell it, Give me a job description, and it spits it out. It doesn't work that way. Both of them are wrong. The right approach is you start with a prompt, so it's you are teaching somebody to do something. Okay, so you need to treat AI like that, so every new task that you give it, every new job description that you give it, you need to treat it as if it is the first time it is hearing about this, so it's like you're explaining something to a child. The child doesn't know what it is; the child is hearing it for the first time. So what would the reaction be? The first thing would be it would make a mistake, so you need to expect that it will make a mistake. If it gives you a good enough job description, then you know you're lucky; it's great because it has that knowledge now with all the large language modelers that are improving every month, and they are sending out updates. The good thing about it is it has learned most of the stuff that the business stuff, especially that we don't need to train them, so if I give it a prompt saying I'm looking for a data analyst to be on my company, this will be their typical day in the job, like these will be the typical activities that the person will do. and I want to post it on LinkedIn, and I need a job description generated for this so it gets an idea that, okay, the platform you're targeting is LinkedIn; you need a data analyst; this will be his or her typical day, like so this way it is much better equipped to generate the right job description for you. Still, it can make mistakes, so you need to go through it; you need to edit it; you need to add a few cultural things to it that you think are important based on your company, and only then can you publish it, but yes, it saves lots and lots of time, like somebody who had to do research on the internet, somebody who had to actually think about it. Okay, what does a data analyst do? I had to ask around people in the department. Okay, the person will be reporting to you directly. What do you expect that person to do? All those things are now gone, so you save hours generating a job description.

TIM: Yeah, and I think you make a good point there, which I wasn't really thinking about, which is a lot of the people who are writing the job descriptions aren't the hiring managers; they'd be talent people who then have to research it, whereas I was thinking more of myself writing a job description for a role that I'm hiring. I obviously know what I need, but there's often a disconnect there in the process, so I guess for those people, then ChatGPT might help even more.

HUSAIN: Yes, because in most of the companies, the person who is writing the job description and the person actually conducting the interview and the person who is actually finalizing the process—they're not the same person.

TIM: Yes, which is probably another conversation about some of the systemic issues in hiring and those kinds of disconnects between the hiring team and the talent team. I find that quite interesting, and I feel like that's a root cause of a lot of the issues of hiring, but I don't want to go down that rabbit hole at the moment. If we could stick to AI for a little bit longer and think about the screening stages of hiring, have you tried to use AI to decide which of the candidates you should interview if you contemplated that or had any thoughts around the screening problem?

HUSAIN: Yes, and I would like to be a bit of a critic here so we can see or hear that most of the companies nowadays use a system called ATS, or Applicant Tracking System, so what happens is whenever you apply, the application goes into that system, and then that system decides the filter, so it decides whether the resume should pass on to the next stage or not. and the ATS stays in the loop till much further the resume travels, like maybe all the way up to the interview call or even after the interview call. The ATS remains active in the loop, screening all the candidates from stage one, round one, to round two, to round three, and they keep track of the... So I think it started with somebody saying we need a database to keep track of all the people that we interview or all the people who submit the resume because right now they may not be a good fit, but who knows? In the future we may have a requirement, and we don't need to spend lots of hours again to do the work. So then somebody was smart enough to build an ATS, and now there are so many ATS in the market. The good thing about it is that it's a good tool to prepare a database to give you the candidates that are in queue, the candidates that have not been hired anywhere, or it's like an active database of, okay, six months back, whom we hired or who has submitted the resume. We have an opening again. Can we look at that so we save a lot of money and save a lot of effort trying to post that same job again on the market? The bad thing is that this thing has started to decide which resume should go ahead and which should not without human intervention. Now, for some of the roles, it could be good if it's a factory role. When I say factory role, it's something which is like detailed out to the dot. Exactly this is exactly what the person needs to do. Somebody needs to fix a bolt on an assembly line. This is exactly what that person would be doing eight hours a day, so that is like being 100 percent specific: this is what is needed, and this is the kind of experience we're looking for. You don't need human intervention over there, but for technology roles, be it any technology, not just data and AI, or not just anything to do with analytics or any such thing, yeah, it's more abstract; it's more complicated, so you need a person who needs to do a lot of things, and you need a person who is good with a lot of tools. You need a person who has experiences with different processes with different systems with different frameworks, and the list goes on and on, so putting a system that is autonomous and just makes decisions without having any consideration or having its own logic to apply, and the human says it has helped me out. It's like a myth. I would say they're living in a myth, so I think for certain roles you can start with ATS, but then pretty soon you need to be very involved in the hiring process. You need to scan through the resume yourself because, and what I've seen many times is that People who are talented people who are good with technology are typically not good writers. Like we see in software documentation, we see if we ask developers to document something, we know what kind of content they are going to spit out, and the people who are good with words or who are good at articulating stuff are typically not the people who are technically inclined. If we get a combination, great, that's that. That's a rare breed. It's great, and if you find somebody, just blindly put them on your team because that person is going to be highly invaluable to you, but then, beside the point, there is a disconnect between this somebody who is technically talented and who is also able to write a great resume, something that can pass through the systems and then reach the recruiter's desk and say, Out of thousands of the applications, these are the 10 I have filtered, and I think these are the best fit. I think it's too much to ask, so while the systems are there, while AI is there, I think a lot of these ATS now implement AI in some form or the other, but yeah, I think we still need a lot of human involvement while scanning the resumes.

TIM: Yeah, at least in my experience of integrating with a lot of these ATS, like Smart Recruiters, Greenhouse, Workable, and Workday, and speaking to a lot of companies until very recently, like in the last few months, these companies, though, wouldn't really use the ATS in any kind of, from what I could tell, automated way, like they were still doing 100 percent manual CV screening with a human. They would maybe have application questions that would automatically reject candidates, but they would be for more like logical reasons, like do you have working rights in this country? What salary do you need? There wasn't really much automatic selection of candidates going on. Maybe now, as of very recently, some of them have integrated into GPT and are maybe doing some CV scoring, but I haven't seen ATSs go rogue and actually start making decisions over candidates. Maybe there's one that I'm not aware of or a particular market use case, but I personally haven't seen that. I have to say.

HUSAIN: So now I think there are courses available on the market, like on social media, that teach you how to trick the ATS to actually not reject you. I think I've seen something that said use AI to generate your resume so that it gets passed a hundred percent on the ATS. So people are coming up with these hacks.

TIM: Yeah, I feel like there are some snake oil salesmen as well, unfortunately, who are selling this bullshit at some level.

HUSAIN: Exactly, and this is actually creating a lot of issues in the entire hiring process. Some of the things we don't do because we know we can see it, like we know the limitations of AI because we have been using AI for many of our business operations as well. We tell it, we give content to it, and we say, Can you give us a better version of this? We give it a content and say what kind of keywords do you find in this, typically with we do it a lot with our SEO optimization work; we do a lot with content copy work, so we understand its limitation understanding prompts or trying to decide, but what I have not seen so far is how can I, so if I get in, if I'm getting in a stream of resumes. Let's say I get 100 resumes a day as a recruiter. What kind of workflow optimization should I put in place using AI so that these 100 can really be brought down to a good enough 10? Something like it's not excluding somebody talented out of that list, or those 10 that it has, it is highlighting me to look at are really the ones that I need, so I will be saving time. I haven't seen that kind of workflow yet, but what I've seen is that the ATS will filter it out, so typically you're right because in many applications the questions are upfront, like whether you have rights to work in a country, do you need a visa, are you eligible to work, or do you have W4 authentication if it's a US role, and things like that. but many of the applications don't ask these questions, but they still go through the system, and the system then filters it out.

TIM: The system filters it out, or a person using the system, unless there's been some very recent development, it is humans sitting there reading a CV, and they make the rejection decision. 99 percent of scenarios that I have personally ever seen in five years

HUSAIN: I hear you on what you're saying because there is one person who is designated to do that in the company scan-through, but if you are asked to do that day in and day out, how motivated will you be to do that? So if I'm getting a stream of a hundred resumes every day, I would say I'm getting a lot of junk. and then after that what I'll do is I'll probably, if the ATS has the feature, I will just ask the ATS to filter things out, and then one is it's being frustrated, and second, it's being lazy, because once it started giving you even 50 percent results, you start believing in the system, and you say, Okay, it's doing the 50 percent work. so might as well I don't care because I'm being saved the frustration of the routine, the grind that I have to go through a hundred times a day for each resume. Yeah, it's a very gray area, but if AI can come up with some kind of a workflow that can filter the invalid ones out, I would say invalid ones are the ones that really don't apply to the job. like on LinkedIn I think if you go to the job listing on LinkedIn, almost for every job role, you will see a hundred-plus candidate applicants. You don't see a number that is below that, like the job is posted one day ago, and you have 100-plus applicants, so now I don't think everybody's in the job market looking for jobs right. And then for roles like, Okay, we need a CTO," it is posted on LinkedIn, and then there are a hundred-plus applicants in one day. I don't believe that there are a hundred people looking for a CTO role in a particular country, and there are so many people that after one day of job posting, there are a hundred applicants. So what is happening is the platform is open, the job is there for anybody to apply, and everybody's just clicking on Apply. Right? So this is a problem for the recruiter because they keep seeing all the resumes that are not relevant for the role, and they are pressed for time, so their boss says, I need something in a week. They are scanning through all these resumes that don't make sense to them. If they have a system that gives them, I will do the filtering; you give me the criteria, and I will do the filtering. That is one second; I give you the confidence, and this is important: I give you the confidence that whatever I'm going to filter and send to you is the one that you need. Now AI has that promise because through large language models and the different learning patterns, it can understand exactly what is needed. It may make mistakes once or twice in the beginning, but after it has understood the context fully, it can probably go through all the fluff, remove it, and then give you really good ones all the time. But that is the confidence that AI has to put into the person, but till that time I haven't seen a system that comes out with that, so the closest one is the ATS, so I think that is why the people are inclined to just go to the ATS for filtering.

TIM: Moving on a little bit then, so what about for data professionals themselves, so data analysts, data engineers, data scientists, these kinds of characters? How do you think their skill sets are going to need to evolve over the next couple of years? Will the definition of what a data analyst does, for example, change drastically? And yeah, how do you think AI is going to impact their role paths?

HUSAIN: So what I think is, traditionally, or even right now, if somebody—so one good thing is this field is open to anybody, so it doesn't need a math degree or a statistics degree right now; it doesn't need a computer science degree. You need to know programming a little bit, but then those are some of the skills that you can acquire. You can learn those skills if you're willing; the field is quite open. In other fields in technology, you need a certain degree; you just need to have a basic understanding of algorithms. basic understanding of math if needed Otherwise, you don't even need that. It started with that. I think that's the traditional requirement. What is happening now is, and what will happen in the future is The data analysts right now are thinking this is the data I need to work with; this is the data system that I need to work with. and I have a program for doing this. I have a program like these that are like my workflows. I need to know Python. I need to know Pandas. I need to know NumPy, and I need to be able to manipulate data in a certain way, being good with statistics. I think in the future what is going to happen is They don't need to write a lot of code because AI is going to write code for them, so they need to really become creative about telling AI what to do because the more creative they become in telling AI what to do, like generating the prompt, that will be their skill, and that will be the value they will be bringing to the company. Right now, if you look at Microsoft CoPilot, if you look at ChatGPT And I've seen Copilot; I've used Copilot in many of my works previously, and I think recently as well. It is extremely good at generating code, so if I want, if I have a huge database and my boss asks me to generate a certain insight out of the data or understand a pattern in that data, I would be spending like, okay, one hour, two hours. writing that code and then running it, testing it, and saying, Okay, this is what I think, and then my boss would say, Okay, tell me how you did it because I need to understand the process just to be double sure or a hundred percent sure that we did not miss any of the requirements with Copilot. I can run the same thing, the entire workload, in minutes, and I can come up with a fail quickly, fail fast kind of thing, and I can come up with a solution in minutes, so the data analysts of the future need to understand how to work very well with these tools, like how to understand Copilot and understand its limitations. understand its capabilities understand ChatGPT, understand its limitations and capabilities The better they understand their limitations, the more creative they can become working with these tools, so this is definitely going to be a requirement that companies would look out for. It does not mean that they can forget about understanding statistics, understanding how algorithms operate, or understanding how to use certain frameworks or libraries in Python. Yes, but those will be the fundamentals everybody needs to know that because what if it's like flying a plane nowadays? A plane just flies itself; the pilot is only needed for landing and takeoff, but what if something goes wrong midair? Then you need the pilot's experience. It's not like you are on the train on the simulator and then you directly go flying; you still have to go through the grind. Most of the time, the computer takes care of it. Your experience and your training will be useful in that 0.0001 percent probability that something happens midair. I think that is now true with all the data jobs that are there. and it is going to be an increasingly useful requirement that data analysts, data engineers, and data scientists use these tools effectively in their resumes and in their projects. They claim that this is how we use these tools, this is our creativity, and this is the kind of result we can get. They need to be vocal about that.

TIM: I love that analogy to the pilots in the plane; that's really helpful to paint a picture, and this advice could probably extend to any person in any role, really, like needing to really understand how to use these tools, the limitations, and also probably keeping up with the progress because they're changing all the time. so the capabilities the strengths and the weaknesses are consistently changing, and I feel like it's always challenging when you've been doing things in a certain way for such a long time. Questioning something as fundamental as, for example, imagine you've been doing it for 20 years, and every time you do anything, it starts with a ticket. In a backlog system with a description, you open up your code base. You have been figuring out where in the code you need to write it; you're adding some code; you're testing it locally; it gets code reviewed; it goes into dev. Like, it goes through this consistent workflow that you've been so used to. There's something now fundamentally changing where it's no, you're not meant to be writing the code. You tell a large language model, and it writes the code for you. I think it's hard to when things as fundamental as that change, it's hard to adapt quickly if you see what I mean.

HUSAIN: Yes. I can share a very recent experience with you, so what happens is when somebody has been doing something in and out for many years and they know it works, we are not questioning their system. Their system could be flawed. When I look at it as an outsider, the system could be flawed. I can say I see a few improvements right up front. Why don't you do it? But the person who's actually doing it for many years, that person is comfortable; they are in their comfort zone, and it's always a human tendency if you challenge the comfort zone. The person takes it personally, and they think of you as an enemy, and they think that this person is out here to take my job. or if I follow their advice, suddenly I'm becoming obsolete, so they get this fear, and they start building this bit of a hate kind of a relationship with this entire process, so I can look at my recent, pretty recent experience when there was somebody on our client team, and that person was doing things in and out consistently for many years. and this whatever process that person followed, they used to give them consistent results, and it happened that that person was following the process blindly because now for many years they're trusting it. I looked at it, and I saw a few improvement areas, and I said, Why don't you do this? It needs to be done in this manner. So I was using my experience and my knowledge and giving the person the guidance. I never thought that it would be taken in a way that you are challenging my work. It actually hurt the person's ego or something. So the same thing happens with Copilot. I'm taking an example of Copilot because you mentioned software development. So imagine that the testing team locked a bug. You received a bug in the morning, and earlier you used to go through the code base, scan that particular area, and then test it out, and then whatever conditions they have put, I'm not sure exactly what it is called, but there's a term that says this is how you reproduce it. So the test is usually put in those steps to reproduce, and then as a developer, you go and you exactly do all those steps to make sure that it is really being reproduced, and then you start working, so this whole thing used to take many hours, and I myself did that because I was a software developer for many years before coming into data so it would take us days to actually fix small bugs and then wait for the release, and that's a whole different story, like when the release happens, but yeah, fixing one bug would be like from one hour to one day to two days; it goes on, and then you have all the history and version history and things like that. imagine that you are scratching your head, thinking of the algorithm doing mental calculations This is what happened: this is what happened. Okay, I'm missing a condition here; I'm missing a condition there. I need to… There's a variable with a wrong value coming from somewhere, and then maybe this particular loop is just going It's actually blocking the entire code, so something is timing out because of this particular loop. Maybe inside that loop I'm doing something; I'm scanning the array of objects; I'm going through a list of objects, and then Copilot, so I need to really think about how I can optimize that. So if the issue is like the code is bloated and you need to optimize it, the issue is like I may spend days with Copilot; it says this is the optimized code, and it gives it in an instant, and it just scans the code and gives it in an instant, and when you test it, it actually is optimized, so suddenly what happens is you feel as a developer that you are not valued anymore; you're not needed anymore. because now the AI is telling you this is the right code, so what happened to your engineering? What happened to your academics? What happened to your experience? Everything is suddenly nullified, but I don't think that is the right way to look at the current situation. I think the right way to look at this would be that there is a web developer who has been paid X amount of salary every month and is expected to do Y work with this AI with this copilot or ChatGPT or whatever you're using. If the developer who earlier used to take, say, 10 hours to complete the job can now do it in half an hour or one hour, the company has nine hours of buffer, nine hours extra for this person to use somewhere else or in something different, so this is one. Maybe the company can use the same person for multiple projects. They are actually optimizing their resource usage or optimizing the cost. This is one way of looking at the situation. The second way is the company says, I don't care what you do; it needs to be done in 10 hours, and the developer says, I can do it in one hour with Copilot. but my manager doesn't care, so fine. I have now nine hours. What do I do with it? Either I could play a game on Xbox PSY, or I can do my own self-development and improve myself. Maybe I can learn a new skill; maybe I can learn something different; maybe I can try for a role that requires slightly more challenging areas or whatever. so you can focus on personal and professional growth I think the web developers should start looking at things with a different perspective rather than saying the AI that is coming out is their enemy. I take it on my ego because it can write code better than me, and I just hate AI, so I don't think that's the right approach. and for companies, I don't think it's the right approach to say I don't need web developers because I will just hire an intern, teach them how to use Copilot, and then let them run.

TIM: Yeah, that's a recipe for disaster, but yeah, you're right; that's a great reframing, and surely if you have a tool that might soon be able to give you a 2x, a 5x, or a 10x boost in productivity, that is a good news story for you because if you could be 10 times more productive and 10 times more valuable, you've created 10 times more value. I guess the challenge is if, let's say, a developer in this example, if they've tied up in their identity writing code and that's what they do, as opposed to producing business value or producing a product that solves problems or helping people or helping the end user. I feel like that's the problem as well. and there needs to be a reset in their mind that what they're writing is just a means to an end, the code itself Despite what some people might say, maybe it isn't the valuable bit; it's the product that then solves the problem for the user that then gets revenue for the company that ultimately pays them their salary.

HUSAIN: Exactly, and in my hiring process, like when I am doing the interview, there are a couple of questions that I always ask, and people usually don't—they are not prepared to answer because they never thought about it this way, so the first thing I ask is why did you do what you do? Why did you write that particular code? Or why did you do the data analysis in a certain way? What prompted you to do it? Mostly the answer is that was a requirement. No, so the problem is people don't think; they think that, okay, they believe that my job role is my identity; they believe that my job responsibility is my identity, and I think that is the problem because that is how traditionally the industries have been set up. When you ask somebody, What do you do? I'm a data analyst. I think that that thing came all the way from, I don't know, the 1960s, the industrial revolution time or whatever, when people had set up factories and everybody was given a role, and what do you do on the assembly line? I do this. What do you do? I fix this so it's like nobody, so that was that time, but I think it has carried on, and it has. We are conditioned into believing that data analyst, data engineer, software developer, and web developer—even like you have in software web—also you have front-end developer and back-end developer. People associate all these roles with themselves when they are telling people about what they do and who they are, so if I say, So what do you do?" I am, so they don't say, "I do data analysis.They say, I am a data analyst.

TIM: Yes.

HUSAIN: So I think the problem is at the core, and I am not sure exactly how we are going to solve it, but it needs to be solved now because more and more, with all these tools, I am a data analyst is not going to fly because the tool is a better data analyst than you as a human at what you are doing. So I think everybody—the people who are coming out of college, maybe doing internships for the first time, or even experienced people—they need to really rethink their responses and rethink what they are doing or what they want to do very seriously in a very different way. They need to just shatter the current mindset that is there, and they really need to rework the whole thing.

TIM: Yeah, I think that's a great analysis, and you're right. We are too tied up to our jobs, which are inherently temporary anyway. People change professions; they change roles; they get promoted; they change careers completely. It is perverse to say the phrase I am dot job because that's surely only a limited part of who you are as a human. and we've overvalued it; maybe you could argue we've overvalued our profession relative to other things in life, perhaps.

HUSAIN: That's a different area for discussion. Maybe, maybe a different show

TIM: Perhaps, yeah, what about coming to hear a bit more about your agency and particularly how you approach hiring and whether there's any kind of connection between your business culture and your unique approach to marketing and whether or not that influences how you think about hiring into your company?

HUSAIN: Okay, so basically the growth agency is the name, so it's a Belgium-based digital marketing agency, and it's a completely remote agency, so all of us are spread out across the world, and we are 100 percent remote. I'm working from Dubai; my colleagues, some of them are in Belgium, and some of them are, I think, in Indonesia. Yeah, and then we interact, so what we do is we use it. So how do we do it? How does 100 percent remote work? So many people, after I think with COVID, people started one thing—one good thing that COVID gave to us is to know, or for a fact, that things to do to get things done, you don't need to travel to a place that is called a workplace. A workplace now can be anywhere. Okay, it depends on the kinds of jobs; like if you're a doctor, you need to go to the hospital; you cannot bring the hospital into your home or wherever, or on a beach. You need to go to the hospital because a hospital is the infrastructure that you need to practice, and that infrastructure you cannot generate anywhere else. So there are some jobs that require you to travel to a certain place, but with that as well, yeah, what COVID has taught us is that consultation can be done online. You don't need to visit the doctor to do a consultation; it can be done online, so I think people have realized that all the jobs have some part to a certain extent, and that extent depends on what function you're performing. Sometimes it's a hundred percent; sometimes it's 5% or 10%, but some of them can be done remotely, and some of them can be automated, so technology-wise, what we believe is that if you are a digital marketing agency, if you are a software development company, if you are into design, UX design, it can be done 100 percent remotely. and we have been doing it for, I think, many years. The company is 10 years old. I have been with the company for the last two years, and we are doing it really well, so we use a lot of tools to collaborate to make sure that the work gets done. To discuss, we lose use a plethora of tools to help us with that. And our approach to digital marketing is that we are a data-first agency, meaning that whenever we say something or whenever we put something out, saying if we tell a client to change the strategy of their branding campaign or of their any marketing campaign, we start with data. So we don't base everything on our experience like we think in the industry; 20 percent of something is a standard, so we say you are working in a financial industry. We think 5 percent of ROAS is the standard, so we would actually target that. No, we don't do that. What we do is we say we will look at the data, the historical data; we see what the trends are, we see what the patterns are, we see if you are impacted by seasonality, how much, and when. And then what did you do about it, and did it have any impact? Did it have any result? Based on this, we prepare our strategy. So we do a lot of data thinking and data analysis before we come out with any strategy, so all of our decisions and all of our conversations with the clients are data backed, so we never say to the client, "In our experience, this happens. No, we don't do that because it's not true. Yeah, and even with the same client, we don't say last year it happened like this or this year it will also happen because trends are changing. We always look at trends; we always look at what has happened, what can happen, and what the forecast looks like now based on this new variable, and that is why data is at the center of whatever we do in the agency. and we use a lot of Python to actually do the analysis for almost everything and then come up with all the numbers, and even our pitches are data-driven because we start with what has happened with this client or what has happened in the industry when the client wants to go there; this is their objective. this is a goal How we can bring them there and what does the data tell us? So many times we tell them this is a goal that is a bit far-fetched, and we support it with these data points that are there, and with AI, what we are doing is we are trying to automate as much as we can of all the routine work that earlier we needed to really wait for somebody to come back to us. and most of the improvements that we are seeing are in our SEO and content. We are the ones that we started with AI-based marketing campaigns. A lot of our creatives for ads are AI-based, so earlier, what happened is earlier the clients, or even we, the client, didn't have the resources, so we would do that. We hire actors. We do a video shoot, and we say, Okay, it's a five-second ad, so we do the video shoot; we do everything; we hire a crew, and they get things done. You have the actors and things like that, but we started launching similar quality or better quality ads just using AI videos. and it really got a really good response; we got a tremendous return out of it, and it has been made as a use case in one of the clients, so for the one we use, it became a use case where they wanted to show it to their entire global group, saying this is the AI-generated creative versus this one that was created manually. and see the kind of impact the AI-generated creator has, so now they are coming back to us with, Can you help us with it? and then there are other people, other institutions within Belgium who are taking interest in this and saying, How can we use AI in this kind of marketing as well? Yes, it has helped a lot. It doesn't mean that we had asked our designer that they no longer need it; in fact, they needed more because now they can generate AI-based creatives quite fast, but they can also control the quality, so we need them more now because earlier they just used to do the creative, but now they actually are more involved in the process. So what AI has done is it has boosted the value of people in our agency rather than diminishing it.

TIM: Yeah, it's a great leverage tool, isn't it? And really getting more out of what you have, I sometimes think of AI as a multiplier, so if you're trying to use it for something that you know nothing about at all, multiplying zero by 10 probably doesn't get you very far, but if you already have knowledge in that area, you have a knowledge of five; you multiply that by 10, and suddenly you've got 50. and that's, I think, one of the amazing powers of AI. One final quick question for you is if you could ask our next guest a question, what question would that be?

HUSAIN: Interesting, so in the same AI space

TIM: in any space you like, and you don't even know who the guest might be; it could be anyone.

HUSAIN: How did they bring about the data culture, or how successful were they in bringing about the data and AI culture in their organization? I would like to know how they did it; for us, it was our CEO who already has that mindset. It's not like we are cultivating it. It's not like we are a 50-year-old company. We started with traditional stuff. We have bosses who are like 50 years old, and we have an army of 500 people working for us. No, we are like a team of seven, eight, nine, or, at times, 12 individuals, but then we function lean; we function very optimized. For us it is very easy to actually say this is something that we need to change, and we change it as a company culture, but so the two things help us: one is the size, and the second is the mindset, so the mindset is already there that we need to be experimentative; we need to be agile, so everything then comes together. But imagine that there is a traditional organization, and you want to bring data culture or the AI culture into the organization. As a leader, how would they do it, or have they done it, and then how did they do it, and what was their success rate?

TIM: A great question that we will level at the next guest on the Alooba objective hiring show, Hussein, it's been a great conversation today about a wide variety of things, almost delving into some philosophical areas of life as well, which is interesting, so thank you so much for your thoughts today.

HUSAIN: Thank you so much for having me on the show. I really enjoyed speaking with you, and I hope not all of the things that I've said resonate with everybody, but even if they find something useful out of it, then I would consider it that yes, I have brought in some value to somebody.