Episode 58

Bharat Ram Ammu on Exploring AI's Role in Modern Hiring Practices

TIM: Welcome to the objective hiring show. Thank you so much for joining us.

BHARAT: Thanks for having me here, Tim.

TIM: It's absolutely our pleasure, and I'd love to start by having a bit of a discussion around AI, one of the hottest topics in probably any or every industry in every country in the world right now, and I'd love to get your thoughts on the use case of AI in hiring, particularly if you've started to dabble with AI on the hiring company side of things or if you've seen maybe candidates start to use it on their side. I'd love to get your thoughts on this topic.

BHARAT: Yeah, so I think just maybe a bit of background on myself, so since my expertise has been working on using AI effectively and efficiently for compliance, which I would say is a more rigid check, so to speak, I think we still try to push use cases that are even a bit old relatively given the pace at which AI is growing with the advancements of LLMs and racks and so on. Yeah, I would say our industry has a lot more to catch up on, so given my context, I would say my experience with using AI for hiring is still something I'm excited to learn more about rather than something I have used in my processes, but I definitely know the potential of what it can bring. To hiring, I think for starters I know One of my friends who's running a startup I know in their company they use just like your AI agents to record every meeting, which means the AI Can't record notes. I also know a use case in another friend's company, I think in the Big Four, I believe, where he was mentioning the other day that they use AI to write critical summaries on CVs. So I was excited to hear, but he also expressed how AI still has a lot of catching up to do compared to how good humans can be at writing critical summaries, so these are some of the things I heard learned from my circle, but I still have to understand and edit the pros and cons from hands-on still.

TIM: Yeah, I feel like most people I'm speaking to in that category are maybe just touching the sides of it, but I don't know of any company that's really using AI properly in hiring yet. What about from the candidate side of things? Have you seen candidates use it, for example, for CV optimization or anything like that?

BHARAT: Yeah, I can't say about others, but I can say about myself as a candidate. I myself, if I'm taking any interviews or any interviews, or if I'm interviewing someone, I'd rather I use AI to prepare myself better. I believe in that; it's all about formulating your questions to AI. That's where the—or, in other words, the prompt Engineering is the key part. As a human, you need to be critical in how you ask your questions. The more specific and critical you are in your questions, the better answers you get. So I think it's a lot more useful as a candidate if you want to learn some specific technical questions, which you know what you're expecting from a certain interview as well. I think that's really, really significant progress that AI made in the last year or two, starting with ChatGPT, where right now the ease with which you can prepare for an interview is much, much more comfortable, and I myself feel much more prepared for interviews as well.

TIM: And so when you've been doing the interview prep, do you use it in the voice mode where it's an ongoing conversation, or do you use it to brainstorm questions they might ask you? How exactly did you use it?

BHARAT: So I think myself, I believe in asking the critical questions. I don't want AI to dump unnecessary knowledge to me. It's important that I get answers that are very specific to what will really help me, so from that aspect I try to iterate the way I ask questions. For example, if I'm specifically interested in learning something in the topic of Python or data science, I try to formulate my questions in such a way that I am only targeting those aspects that are relevant to Python and that are really asked to data scientists in interviews, for example. So because there's so much to learn from Python and data science, it's important to understand that, yeah, sometimes AI will start with pretty irrelevant answers, so it's important you need to iterate and keep looking for giving some examples to the AI as well. Look, these are the kinds of questions I know I ask in my interviews, or I get in my interviews. So once you give examples or once you tell, Okay, these are the kind of topics I'm interested in covering, I think the AI serves me better, so that's how I start again from an experience point of view. I rely on the voice GPT for React just getting my text to speech; that's it. So I think speech to text is better, and then I keep relying on it because I'm more of a reader rather than a listener, so I can read faster than listening, which I think is pretty slow, and I don't really rely on that.

TIM: In an interview, in an actual live human interview, would you consider having ChatGPT or some kind of other large language model running to assist you in some way, or do you feel like that would make it more complicated, or is that off limits? What are your thoughts on that?

BHARAT: Yeah, I think again now I think I'll have to bring my compliance hat where since I work in a domain where security is very important, our experience in my company as well is that we understand there are certain limitations with certain models like ChatGPT where privacy is not respected as much as, let's say, the LLMs from other companies like Google, for example. That's one of the arguments where I think Google is known to serve better for security purposes, but having said that, the performance of Google probably may not be as good as ChatGPT. So there are pros and cons on both sides, but in a certain specific use case like this, where along with getting some value out of the AI in the meetings, it's also important that privacy and security considerations are taken into care and that there is no data breach or privacy breach, that that can happen in the long run after any tool is implemented, and I think it's important that these things need to be really evaluated before implementing AI into production, let's say in terms of having it all in corporate meetings.

TIM: For sure. What about, though, from your end as a candidate? So maybe not thinking about the company and the regulations around data privacy and whatnot, but from your end, if you're going into an interview as an interviewee getting a job, would you consider using a large language model as part of that interview process to help you perform better in the interview?

BHARAT: Yeah, I think I would definitely use it as long as the model is legally allowed to be part of the interview, so I think once I know it's legally allowed, I think I would be comfortable using it because since having used LLMs, both API-wise as well as just as a prompt engineer, I've used it for various years now. So I'm pretty comfortable, and I'm pretty confident in my abilities to extract some essence out of it if that is part of my interview, but having said that, I'd also need to look at a bit more critically what value it exactly brings from the interview. I'd definitely be open to trying that for sure.

TIM: Yeah, I've seen some of those videos on LinkedIn and other places of candidates in an actual interview using ChatGPT on one screen and then they're interviewing on the other, and it looks to me, at least at the moment, like that would be more complicated than just doing the interview yourself as long as you have the relevant skills and experience. I don't know about you, but I would find it way harder to watch one screen with a large language model giving me feedback and then interview with an interviewer answering their questions and translating the chat GPT stuff into my normal words. That seems like an overcomplication to me. So maybe there'll be some simpler version of the future, but at the moment I don't know. Personally, if I were interviewing, I would steer clear of using them in the actual interview itself, even though I've seen some junior candidates try to do that.

BHARAT: Yeah, I think for myself as well, sometimes it's important to be transparent. I think as long as the interviewer and interviewee are transparent about its usage, it's always possible to take a step back where you say, Okay, this is the point where we need Personal conversations where we want to know you or the person the candidate has to be known by the interviewee at a personal level and AI is just noise at that point So I think it's important to have the ability to switch back and forth. Yeah, I think that for that to happen, it's important that both parties need to be fully aware of what is used there, and I think once it's there, it's possible. For example, how I see it technically speaking is, yeah, yeah. I cannot know technical concepts that I want to refer to the candidate or ask some candidate right away, but if I have ChatGPT in the call, I would rather tell them, Okay, look, let me look it up here, because then I can quickly know what I exactly want to ask, because technically speaking, there are so many concepts to learn, and yeah, with the advent of GPT, it becomes much easier to improve the quality of a technical conversation or improve the quality of an interview. So I see that way that it should only be used sparingly and with mutual consent, and I think in that way you can effectively use it, and for that I think it's important to get the approval where both parties are fully aligned and there's a clear process defined on how and when AI is used during the interview and when it's fully personal and it's fully human intuitive.

TIM: Yeah, I feel like that's a good general suggestion as well. For hiring, I feel like hiring could be a lot smoother and a lot better if both parties were quite transparent and upfront with each other about everything—not just whether or not AI is being used, but everything else. Imagine how much time companies could save if they just told everyone exactly the salary on the job, exactly a day in the life of the role, who your teammates are, what your metrics are, what product you're working on, and if candidates were maybe a bit more honest in their CVs as well and weren't exaggerating skills. Everyone was just honest. God, we would save a lot of time, don't you think?

BHARAT: true

TIM: But yeah, we're a long way from that at the moment. I suspect, unfortunately, including with one particular thing I'd love to delve further into, which is what we're hearing consistently at the moment from pretty much every hiring manager we've spoken to, especially in Europe and North America, is that when they put up their job ads, they're being absolutely bombarded with CVs—hundreds of applications, sometimes more than a thousand. for some, particularly like remote jobs in America, where there's just a huge candidate pool, and so they're getting all these CVs, and what they're saying is the CVs seem to be looking better and better in that they seem to match the job description even more than they used to, and they're looking more similar to each other So it sounds like a lot of candidates are using a similar model, probably ChatGPT, to somehow optimize their CV against the job description. What are we going to do, though? Because what companies are also finding is that the CV isn't really reflective of the person; it's even less reflective than it used to be, which was pretty bad. And so now we're left with this document that people are submitting that's just nowhere near reality. We've got hundreds of them to deal with. Like, how are we going to get out of this jam that we're in? Do you have any thoughts on what will happen in the next year or so?

BHARAT: Yeah, I think that's a pretty alarming problem. I would say the more I think about it, the scarier it seems, especially since, to be honest, I'll be guilty myself of using some of it to improve my CV. I tried to use GPT, but the way I do it is I try to specifically get the content I get first, and then I try to look for iterations very specifically in the content I want to improve in the elements of my CV. For example, So that way, my CV still says unique and true to what I want to present, so I think I'm sure if you're not patient enough or if you're not confident or probably you don't have the skills that you claim to be doing, so probably you go the easier way with using a specific generic CV with the GPT, and that's probably where these kinds of problems arise. So I think the key lies in probably it's going to be a problem, I think, especially for HR managers who are filtering out CVs. I think the problem is more from their side, where they have a bigger job to filter out CVs. Perhaps their AI can again play a role, for example, an AI agent that is provided with the right prompts or, I don't know, questions. So these are the things that are provided by the hiring manager. For example, this AI agent can already filter out or give scores to each CV so that or make the AI agent make some remarks on why they think the CV is not a good fit even though they seem similar, so I think Yeah, it's just I think cutting a glass with a glass is like you need to cut the problems created by AI with an AI. again, so that way you already reduce The applications that the hiring manager needs to deal with, but once it gets to that stage, I think with the human level and the ability of people who actually work on stuff, I think it's, I believe myself as well, it's someone who's working on it who is really a good judge of what somebody is saying is plausible or not. And it's possible to judge it by asking the right questions and asking them more critically and iteratively, and I believe most of the hiring managers are capable of that for any role, but I think the first problem is, like you said, it's important that the HRs or recruiters who are not so good at this domain know what the job role they are asked to hire is. Probably they need help from AI so that they can use AI to filter out CVs in a smarter and a more efficient manner.

TIM: Yeah, and I think you're right, and it seems inevitable that's what will happen, so the candidates, the way I view it, are the candidates. They're individuals. They've adopted the technology really quickly because they can just go to ChatGPT and start using it. They don't have any rules to follow, like company policies. They don't have to have meetings to decide how they're going to implement anything; you can just start using it, and so I think it makes sense that they've adopted it quickly, but then that's caused this influx of CVS. It's not just that they're using CVS to change this, so using LLMs to change the CV It's also that they're using them to apply en masse to many roles, so it's the combination of those things that's causing this volume. I'm sure companies will respond by, as you say, automatically screening the CVS and trying to rank score them against some criteria, and that will at least automate the process, make it faster, make it cheaper, make it more objective because no longer humans doing it—that's all great. The problem, I think, is still going to be that the CV itself is not necessarily reflective of the real person, which is already the case, but I feel like it's just getting worse and worse, is my perception. Can you imagine us needing some kind of new screening method like we can't just rely on a CV? Maybe we have to come up with some other data set, something that's more validated. I don't know, do you have any thoughts on where the screening step might go?

BHARAT: Yeah, I think definitely there are potential use cases there in that field where probably an alternative tool to CV needs to be used. Personally, I have my own set of complaints from my perspective on the usage of CVs regarding the comment you made about the candidates using an LLM along with the RPA bot, which is like just a robot. I have seen it myself. I've tested it myself as well. I tried to use it in the open source. I think normally these kinds of bots are only capable of using LinkedIn, so they are not capable of getting into a website of a company and applying through those; rather, they would go to LinkedIn, and all they can do from what I've seen is they can only do easy apply. So my observation is that even if those bots get better, they can only be limited to LinkedIn because the templates for websites of companies would be very different, and it's not possible to automate a bot to apply or auto-apply rather than using a bot and LLMs for various companies. So I think the way I see it, firstly, is that we are going down the lines from my personal experience as well as applying via other or third-party tools like LinkedIn, Indeed, or, I don't know, Naukri. I know from India, so these kinds of portals probably may not be that reliable to companies anymore. Companies would probably prefer candidates who value their company so much that they come to this website, spend the time, spend quality content as well, and apply for these roles, so they probably prefer to filter them more and more. The more false positives I notice here, I think the more they prefer that. I believe that's already happening, but having said that, now I believe this is again making A lot of candidates who are pretty good but who do not have the time or who did not find out how to apply via the website miss out on certain relevant roles, so I think this is where probably a better tool rather than applying where a CV is needed and I think it's probably up to the HR teams where the usage of AI can really help them, probably the recruiting teams, and recruiting HRs need to have and need to use AI tools to find the right candidates, probably using platforms like LinkedIn but also using AI to also get better scores. Maybe ask AI to ask the right questions so that the score is not just based on the skills. because I believe that's way too superficial the way I notice, for example, classifying a data scientist by just Python, SQL, R, data science, or some buzzwords related to data science, and if you find those keywords, LinkedIn normally already ranks it as relevant, so that's how simple currently the LinkedIn algorithms are. Again, LinkedIn is the platform I know I use, so that's my take on it, but I believe there's a lot more potential to make it more in depth, and a lot more attention to detail can be brought about by using LLMs in filtering and If HRs or recruiting managers can have AI tools to use and to look out for candidates, probably instead of relying on CVs, that's the way to go that way. They can also find the right candidates, yeah.

TIM: themselves, or they've gotten a chatbot to write something about them, but it's not in any way like you can just put anything you want I could say I'm a rocket scientist, and no one can really stop me from applying for a job at SpaceX. That'd be an interesting experiment, actually. Could I get an interview at SpaceX as a rocket scientist? I reckon I could just by hacking together a CV and getting past the initial CV screen, but who knows? We'll try that one out, but I feel like we need something that's closer to the truth. We actually have to interview the candidates using an AI interview. We have to give them some kind of skills tests. We have to have some certification process that means they definitely know what they're talking about. That's to be, I feel like some fundamental change away from just relying on a CV or anything that is just taking the candidate's word for it, so it'd be interesting to see if anything like that emerges over the next couple of years.

BHARAT: Yeah, I think to myself that what I noticed is the change, or I see there is a lot more potential to improve, especially on HR teams, especially the non-technical recruiting teams, who need to understand or use, if not understand, at least use AI tools to understand the specific requirements of the recruiting managers. For example, what I notice is that sometimes, a lot of times, even now, I keep getting roles that require C; for example, where C is something I only worked on 10 years ago or something like this, I've never worked on it, but I keep getting it, probably because C is probably part of my skill set in LinkedIn, which, of course, I added because I just worked on it 10 years ago. doesn't mean I would ever take a job where C is the core skill needed. Yeah, you can see it's completely guided by LinkedIn's just rule-based skill set, which the HRs are just using to approach candidates, so now I believe that's the same that's going to happen, and what I noticed is to filter or to get filtered in the CV is something also probably a bit more difficult. Not so sophisticated, yet the filtering process of the HR teams really needs a little bit of help; even AI can already make the process better where the CVs I think I believe CVs are still useful, but I think it's the technology or the tools on the HR side that is lagging. And once that there is an enhancement on that side, I believe the CVS bluffing using CVS can easily be tackled with that.

TIM: One thing I've always thought, to be honest, is a bit of a flaw with the way hiring gets done is that so much of the process we're expecting talent acquisition people, recruiters, and HR people to manage for roles they themselves have no experience in at all. And so you just mentioned a point there where often it's the hiring manager trying to coach the talent acquisition person. These are the things I'm looking for in a talent acquisition person; they will do their best to understand, but they're not really going to understand the nuance of the difference between a data analyst and a data engineer. It's something that might be obvious to us but is two very similar things to them, and I feel like maybe our expectations of HR talent recruiters are unrealistic that we should even think that they could comfortably hire for roles that are vastly different from their kind of skill set, so maybe, yeah, as you say, maybe if they had this AI tool that almost enhanced their skill set because they now didn't necessarily need to know any of the technical things, maybe that's an unlock. But I also feel like even with that, is it just not a flaw fundamentally to have people recruit for jobs that they don't know themselves? What do you think?

BHARAT: Yeah, I think I fully agree with that because the more I think about it, the more I start to question why recruiting teams or HRs are asked to do that rather than the recruiting manager themselves, the original manager who wants it, but I think that is where probably there's an inherent expectation that people who have, I believe, somebody who did their MBA in HR is expected to understand these nuances. On requirement, as in not really knowing it upfront a prior but rather understanding them or studying them as a requirement comes, for example, if there is a lawyer role that is required to be posted, probably that's when the HR team member needs to do that research probably, but at the same time, it's like you mentioned how much research they do; they would do that. probably not be able to distinguish the nuances between 500 CVs, which one differs exactly how much, so in this case I think there's a clear and a very easy and interesting use case in my the way I see it using an AI model to just score the relevance and have something like a relevant score based on just matching the job descriptions on both sides Of course, that can also mean that the persons or the people who have faked their CVs could get them closest to relevant scores, but at the same time, that's where probably The human eye comes in where normally, from my experience, someone who's good at what they do, they're pretty good at also identifying false positives or people who fake CVs. So I believe After using this AI score to find the relevant CVs, I believe the further or fraudulent CVs, if I can say it that way, can be filtered out again using the human eye of both the HR teams as well as the hiring managers so that, yeah, they're left with only the relevant candidates to go for further rounds.

TIM: And what about from your perspective? If you could suddenly have any stage of your hiring process automated away using AI or any other technology for that matter, which stage would you love to automate away and why?

BHARAT: Yeah, I think personally from the recruitment process point of view, I believe the first round is where I've done myself as well with a certain number of series—not too many, I would say probably I just received less than 20. Probably I did not like that bit of work as much as the work I did later, which is understanding or understanding themselves personally and so on. That is the part I liked, but filtering out is something that required me to go through all the CVs, make critical summaries, and then rank them based on what I thought was important for that role and so on, so these steps also led me to a lot of subjectivity and a lot of emotional, I would say, biases, or probably more like emotional, yeah, just getting emotions into the picture because I might find an okay I like this candidate, but they're probably ranking-wise my objectivity says they are number six, and I only want five candidates, so these kinds of things are probably from both points of view where I don't need to get subjected to emotional biases as well as subjected to going through the CVs in detail where some of them are mostly written by GPT as well these days. So from this point of view, I think this is the round out; probably also want to automate using AI, or rather smarter AI, which is customized

TIM: Yeah, I'd love to get rid of that stage as well, and I feel like it's where arguably the biggest and quickest win is. I would say if you're getting a thousand candidates in, selecting the right five to interview is very important. At the moment, it's very inaccurate the way that's done, so loads of great candidates are missing out, and then everyone's experienced interviewing people that they wish they didn't have to. So, lots of kind of false positives are creeping through as well. The false negatives are really interesting, I think, because these are the candidates who should have gotten an interview but didn't. That is a very under-analyzed set of people because no company is tracking this at all because, at the moment, basically, if someone's CV gets rejected, there's just, It got rejected. There's no backward analysis to go a year later. Where is this person in their career? Oh wow, they've been promoted. They've gone to this other great company. Jeez, maybe we missed out on a great candidate. I feel like we can reduce the chance of that a lot by having, as you say, this objective measure that removes the emotional bias from it. done in a consistent way, even things like I don't know about you, but if I have to read a hundred CVS, I'm not going to read the hundredth one with as much effort as the first. I'm going to be very bored reading a hundred CVS in three hours, so it's even just that that level of bias could creep in so easily, and an AI doesn't get bored, thankfully. We can pay it one cent an hour, and it does a very consistent job.

BHARAT: Yeah, I think that's a really great point, and I think another point that really I feel is personal is filtering using CVs or false negatives is something I believe I would be one victim of that, I believe, but again that's based on my personal experience where if perhaps I was not that good at formatting my CV as I've seen other candidates who normally get interviews, they're pretty good at designing their CVs. they know how to Update their skill set to make their skill sets, like you mentioned, relevant to the job description. I would say I like doing my job or answering questions about my field rather than doing that part that is a part that I did not like so much, I would say, but I had to force myself really to do that in order to get filtered. and I've noticed the difference myself, so having done this, I believe, yeah, the false negatives are the candidates missing out because they lack these skills or they lack interest in these skills is again a cost, and that's probably where I believe currently with current systems among most of the companies it's probably pretty simple. They call it AI, but it's probably not just AI; it's just probably a few rule sets that are filtering out CVs, and definitely that is not the way I believe, so definitely there's a lot more, a lot more nuanced AI that is needed to filter out CVS in the first round, and I think that's the need of the hour: both reduce false negatives as well as reduce false positives.

TIM: I think it has the potential to improve this even further, so I'd like to share an example with you around some studies that people have done when they apply to jobs using CVs and just varying the names that are on the CVs, so I've seen these experiments in America, Australia, Canada, the UK, and Germany. all sorts of markets one in Australia from I think in 2021, the University of Sydney got about something like 10,000 CVs and split them into three groups. In each of these three groups, or the three groups were similar to each other, the only difference of which was the names on the CVs, so for the first group, it had Anglo Saxon first and surname; the second group had an Anglo Saxon first name. The third group of Chinese last names had Chinese first and last names, and so the researchers then applied to all these different jobs in Australia: different industries, different domains, and different seniorities—just thousands of them, like 10,000 or something. They then measured whether or not they then monitored whether or not the CV got a callback for that job, either a callback on the phone or an email back to set up an interview, whatever it was, and then they measured the callback rate among those three sets of CVs. The first group, which had the Anglo first and last name, had a 12 percent callback. The third group, which has the Chinese last first and last name, got a 4 percent callback only, so all else equal, if you apply to a job in Australia with a Chinese name, you have only one third the odds of getting a callback as if you apply with an Anglo-Saxon name. That is very unfair and just, let's be honest, complete racism. And the researchers are very clever; they thought of a lot of other variables that they controlled for, like reasonable differences that might've explained that, and the only thing that was really left was the name, so they concluded that was just complete discrimination, so I know some people listening to this and some people in talent acquisition might be fearful of AI, and there are certainly guardrails we have to put in place to make sure it doesn't exacerbate things, but the current process is already so unfair. I feel like there's so much upside to improving it to make it more objective by using this kind of scoring mechanism in a consistent way like you've suggested.

BHARAT: Yeah, I think, of course, there is a possibility of having discrimination; there will be inherent biases as well, probably a bit from the hiring managers themselves as to what they are looking for. Sometimes they want international teams; sometimes they don't. These factors Probably will be reflected in such processes, but having said that, I believe the key lies in providing the right questions. So I think I just want to come back to your proposal where you talked about the need for CVs. I think the fundamental need for CVs needs to be addressed. I think CVs just serve the purpose of answering the questions that the hiring manager is looking for, and now these questions are different for different roles. So I think the key lies in the hiring manager defining what questions to look for in each role and probably providing these questions and looking for answers based on this. One way, rather than what I've seen in Europe, especially, is there is a standard template of CVs where Europass is one template, for example, where there are only certain questions in a certain order so that there are not many differences in the way a candidate pitches their answers. I believe having a standard template where you're asking the same set of questions, so instead of relying on a CV, why don't you just rely on an application from rather where you get to know exactly the question and the answers to the questions that you think are important for that role? So I believe that's a more specific approach as well, and it can be done at that level firstly, but at the second step, once the hiring manager communicates the exact set of questions that CV needs to answer, the same can be communicated by AI so that the HR team, the HR team that starts the nontechnical recruiting members, can get answers to those questions from each CV. So based on that as well, they can make a decision so that then I believe it's, yeah, that translation from CV into the answers that hiring managers are looking for. I think that translation can be bridged by the use of C AI.

TIM: Yeah, that conversion from unstructured to structured data is just going to have such a big improvement on so many steps of the hiring process, including interviews. I could easily imagine now transcribing interviews and trying to rank candidates based on how they answered questions in interviews, providing feedback as well. So this is one thing I want to ask you about, which is, I'd say, candidates single biggest complaint is a lack of feedback in the hiring process or getting no feedback at all or the feedback being useless. What about the scenario, though, where candidates are being rejected, but maybe they haven't really been rejected for a very black-and-white reason? So it's not like they've done a skills test, and they've scored 40 percent, and another candidate scored 80 percent, and then it's very clear why they haven't progressed in that step. What about if sometimes it's more around like a cultural fit, team fit, or something a little bit more subjective? Have you experienced that yourself, either as a candidate or hiring manager, in that kind of scenario? Like, how should we deliver the feedback? What should the feedback be in your view?

BHARAT: Yeah, I think I'm probably more experienced myself in the past when I felt my CV was not in the standard format. That's what I believe is the answer, but most of the time what I notice is that the answer is almost no, we cannot tell you why, so I think Having an answer like Yeah, there are better candidates or just saying you're not a fit I think this is the that's that for me. It just means no information; it may mean subjective information; it could mean anything, like you said; it could be an inherent bias; it could be any other thing that you notice, but you don't want to tell me. Because I have noticed myself even sometimes requesting recruiting teams to provide me with some specific feedback that can help me in my next interviews, for example, so that is what I haven't noticed answers to them as well, so that clearly suggests to me at least I used to take it personally at that time, but when I look back now, I think the process is defined in a certain way that sometimes it's just difficult to answer exactly why. at a certain place, and that means there is no logical algorithm that is ranking the CVs; it's just a human handpicked process that is guiding the interview processes or hiring processes in most companies. That's my understanding, and that probably can be made more transparent and consistent with AI. So that is probably what this can address as well, yeah.

TIM: Yeah, I think so, and yes, the root cause of some of this lack of feedback is the fact that companies themselves don't necessarily have very clear criteria on who they're hiring and why. It's not like they scored a candidate at each step—the CV, the interview, the test—they don't necessarily have those numbers. I feel like that's one of the biggest benefits of having a more objective, data-driven approach because you automatically have numbers against different criteria, so it would be easier to say to a candidate, We compared you against the three other candidates at this stage. You did better on these three skills; you did a little bit worse on these four skills overall. we've gone with this other candidate because we place more value on communication, where they got like an eight out of 10, and you've got a five out of 10. I think companies could do it this way, but they just haven't, and yeah, maybe that's one of the reasons why they can't give this feedback: it's because they don't have the numbers themselves to support it even internally. Again, maybe AI will help to enhance or make this easier because you could have that AI interviewer sitting in to do the transcription to do the scoring. It's interesting to see where this goes; maybe there'll still be this cultural fit thing, which maybe can never be measured; maybe companies won't even try to measure it. I don't know what you think.

BHARAT: Yeah, I think cultural fit is a very tricky one. I think I used to be a bit more skeptical about this cultural fit in the past, but I noticed myself as well in the recent experiences that cultural fit also plays a role, especially if I think the mission and vision can really align sometimes. But when I talk about values, for example, most of the time the values are redundant from various companies, so we also, for our company, we did our values exercise where we try to prioritize the top five values from the company, but any day we have one discussion for an hour, the values can just go upside down. So the value that is sitting on number 13, for example, from our company's point of view, can all of a sudden come to the top five; it's just one or two good points or one or two members really having a strong opinion on that day, so that's how subjective values can be, I think, but at least about mission and vision, probably I cannot say the same. So I guess cultural fit, if it consists of mission, vision, and values, at least when the mission is a strong fit, again it's a very difficult way to measure how a mission can be a strong fit as well, so yeah, overall it's pretty debatable and pretty subjective, this cultural fit assessment of candidates.

TIM: Yeah, I agree, and I've always been a little bit suspicious of cultural fit interviews because I feel like they sound good and sound reasonable, but so often I think they can be used ultimately just to reject a candidate that you don't like, and you don't want to say you don't like them, or there's some reason why you don't like them that's unfair, and So you're just going to, I call it the joker card of hiring, you just play the cultural fit card. They're not the right cultural fit, and that kind of kills the conversation, at least in my experience. But yeah, hopefully AI could make at least whatever this interview is called a little bit more consistent and objective.

BHARAT: I think along with the mission vision values, perhaps AI can be used to define again questions. I believe that the value of almost all this lies in the question, so probably I've seen certain interview processes as well. There are some questions designed to assess cultural fit that are not so easy to lie about, which means you really need to be honest in order to answer them convincingly. So I've noticed that only in certain companies with good HR teams who have been happy to do that, I think it was a lot easier if your client just hired a third party to provide this questionnaire. And in that case, I also personally enjoyed filling those questionnaires as well, where these questionnaires help you understand what your strong values are and so on. so I believe If the questionnaires are objective, probably yes, maybe you can use them to, so if, like you said, if the questionnaires are actually defined and the candidate answered them, and you probably did not find the score suggesting that they're right, probably that transparency is already available in the report. It should be fully transparent even if the candidate disagrees. Okay, look, we have the results. It shows that you're not a cultural fit. Culturally, you don't fit because of these reasons. Yeah, I think, yeah, and then the same can be brought about using AI as well. Yeah, so it's pretty important that this is evaluated objectively and more transparently. Yeah

TIM: Yep, objectivity and transparency. I think these are things that AI can help us implement in hiring, and I'm personally excited to see what takes place in the next couple of years. All right, it's been a great conversation today, wide-ranging and very interesting, and thank you so much for sharing all of your thoughts with our audience.

BHARAT: Thanks, Tim, for having me. I really enjoyed sharing my thoughts and knowing your thoughts as well on hiring an AI.