Episode 84

Tiankai Feng on Humanizing Data Strategy & The 5 C's Framework for Leading with Head and Heart

TIM: We are live on the Objective Hiring Show. Tiankai, welcome. Thank you so much for joining us.

TIANKAI: Thank you so much for having me. Good to be here.

TIM: It is absolutely our pleasure to have you. It's our pleasure to have you because of many reasons, but one is you're a published author. Congratulations on, on your book.

TIANKAI: Thank you so much. I am not getting tired of it. I mean, it has been six months, almost exactly now since yeah, but it's still feels surreal to be called an author and to be a published author. Yeah. So thank you very much. Appreciate it. it's, it's absolutely our pleasure. And I've got a copy of your book right here. So for people listening at at home, they won't be able to see this. But the front cover is a beautiful, colorful cover with a lot of love hearts. That's kind of permeating the love feeling around the book. The book is called Humanizing Data Strategy, Leading Data with the Head and the Heart.

TIM: What is the book all about?

TIANKAI: Oh, yeah. So the book really is trying to bridge the gap between that everyone agrees in theory that the human side of data is very important. Whereas on the other hand, there's not a lot of actual guidance, practical advice on how to deal with the human side of data. And because of comparably very little literature exists. there are some, but not many about the human side of data. I try to actually basically give it my own attempt to give some practical advice from my experience and from what I learned in my career so far. And yeah, I'm really glad that it appeals to people. I basically break down the human side of data. Into the framework of five C's, which are competence, collaboration, communication, creativity, and conscience. And to say that to each of these elements there are things you can do to put people actually first in your data strategy and how to actually empower people, let them collaborate in the right way, have shared objectives and actually have a meaningful outcome. and working with data?

TIM: Excellent summary. And what prompted you to write this book?

TIANKAI: I mean not as, I would say romantic as it's for other offices. I would say. Actually in 2023 I was translating a book in my free time called Disrupting Data Governance by Laura Madsen. Very nice book, by the way, into the German language. And I worked with the publisher of Laura's as well, which was Technics Publications. And the collaboration went so well that at the end of the year, they had asked me if I had any plans to write my own book. And it was only at that time that I even came to the idea of Writing my own book and I thought about what, what I would write about. And since I've been really specializing and caring about the human set of data so much, I decided basically for that topic and yeah, then the whole writing process started and this is where I am now. Yeah.

TIM: And how do you even get going on something like that? Like it, it struck me as I was reading a little bit of the book today, not having written a book myself. It felt a little bit like a startup. Like you have to create something out of nothing, like take an idea in your head and that's all you have to begin with. And then eventually you've ended up with an amazing published book. So what, what is the process like?

TIANKAI: think, I mean, I would describe the process, at least for me, and I'm sure other authors have different approaches like a mind map that becomes more and more detailed in my head. And actually, for me, it started with coming up with the five C's, like, how would I describe what, for me, the main aspects of humanizing data strategy are, and I kind of said, okay, they all start With C, which is nice. So that means there's something catchy to it and it's easy to remember. Right. Let me use that as the narrative and as a structure of it all. Right. And how would I then describe the details? So from a very high level mind map, it became more and more structured, more kind of path this went off from each of the contents types and aspects. then I basically filled in the gaps in a many ways, unlike thinking what I would want to think about how it all fits together. And then one more. I would say that what's interesting is while being in that ideation phase, I got inspired by my day to day work a lot as well, right? So that was a lot about situations I was in topics that I was facing that I was dealing with that gave me ideas. On what would be the potential solution in a given scenario. So while working full time writing a book is not easy, but it was almost necessary in my case to pick up the speed because this is what gave me the day to day inspiration. To actually have the right content right and coming up with what to write about

TIM: Yeah, I guess when you're in the thick of it and you have this lens through which you're seeing the world, then I guess that helps to highlight the relevant experiences to then include in the book.

TIANKAI: absolutely. Yeah, it also helps when when you experience similar situations and scenarios, like multiple times you identify patterns, right? And you actually notice that probably other people are facing the same kind of situations And if you can offer them some kind of inspiration or guidance on how to deal with the situations that can only make things better, right? Cannot really make things worse that much. So yeah, that is kind of what I had in mind when I, when I wrote the book as well.

TIM: One page I was looking at just previously, which really stuck out to me It's page 16. And on page 16 there's a spectrum that you lay out, which on one side of the spectrum is data focused. So these are data driven decisions, automated decisions made by machines. On the other side of the spectrum, there's experience based experience focused, no data involved, fully reliant on expertise. And this is an interesting spectrum I often think about in terms of hiring. And I've yeah. ask many guests on this show about their thoughts on this when it comes to hiring, like how much they rely on what I would have called gut feel or intuition. I guess here, it's framed slightly differently as experience, but a similar kind of concept versus how much they rely on data.

TIANKAI: Yeah,

TIM: get your thoughts on, on that, how you think about how hiring is done, how you do it. Do you feel like we're going to end up getting more data driven or less? I'd love to get your thoughts there.

TIANKAI: that's a really good question. I would say that in today's hiring, especially in the data space, you cannot really do it all manually anymore, right? Nowadays, you get so many applications to one role. That is impossible to read through all of the cover letters and all of the resumes manually and remembering which ones were relevant and which ones weren't, right? And then having the right mindset for it. So making that screening and that initial kind of assessment of who would actually move to the next stage. Is happening probably nowadays already pretty much more data driven and that's a lot more of keyword focus to the screen resumes also looking into experience with like the right industry, for example Really plays around and that is something that I try to use for help as well other hand, right, the more you get closer to making a final decision, and the more you are investing time into these job interviews, there's the human software element to it. That is important, right? In the end, you need to work together. So the chemistry has to fit and chemistry is very hard to decode into actual. Data elements, right? Chemistry is something that you just kind of feel and you would feel like it would be fun to work with that person and you cannot actually put it into words why it feels like the right chemistry. You can only think of our most experiences in the interview process of when that person was really cool or did something really that you kind of could empathize with. So yeah, I feel like the recruiting process itself is like one of the ultimate Working with human critical thinking and data together elements that we will never fully replace. Because in the end it leads to a human collaboration still, and that always require the human assessment at the very end as well.

TIM: Is there something to be said for being almost pushed or controlled too much by our feelings, which are just sort of hiding some kind of bias that we might have. So as an example, you mentioned there around, let's say, finding someone with whom you might have chemistry and you can imagine working with them and imagine having a fun environment. You could wake up after having slept eight hours. Interview someone and feel great about them. You could have woken up after one hour of sleep. Your baby was crying. You had an argument with your missus as you're walking out the door. The train was late. You spilt coffee on yourself. Then you interview the same person in some other version of the world and you thought they were a dickhead. I feel like there's a lot of stuff that's influencing our perception of something. And yes, although in the end we'll have to work with humans. Is there not? Some way that we should, should we be trying to limit how important our feelings are in our hiring process, I guess, is what I'm trying to throw at you.

TIANKAI: Yeah. No, that's actually a really good point. Just as how important positive feelings are, they have, they can actually have a bad side, right? That we might be biased towards people that are similar to us. We might be biased towards people that have the same experience as us in the past. Right. And they are by default more. Attractive to us, right? Just because we share similar objectives, but objectively for the job, it might need someone actually brings a different point of view. And if you unconsciously focus on same experiences, that's actually not helpful. It's a really good point to think that we as human beings need to acknowledge our unconscious bias. And actively work against it, right. To remind ourselves that we shouldn't actually be as little biased as possible and to actually stick to what is the best role or the best candidate for a job, right. But at the same time, I'm not sure how much data can help with it. Right? So it's one of the things where usually we are facing the facts already. We know that people have different experiences with us or have the same experience with us. It's not like more data can, more of the same data can tell us to be less biased. It's almost like a switch in our heads to say now we have a good feeling about this person, but is it really all valid or are things in there just because I personally. Want to hang out with people that are similar as I yeah I I think I agree with you right at some point this is like a limitation of our own emotions and we actually should try to rationalize a little bit the chemistry and attitudes that we have with candidates

TIM: I'd love to throw an approach at you and see what you think of this. So I, I tend to find and tend to believe that if we just really thought in a lot of detail about exactly what we were looking for in a candidate, and we just had almost like a meditation session to go through and say, like, okay, Let's do a first pass of the skills we want. Here's this list of, I don't know, six technical skills, four soft skills. Okay, that's a good first start. Then you start talking about what you really look for in a candidate. I'll remember that candidate who he hired before. They were really good because of X, Y, Z. And here's what this person's gonna have to do in this job. And you sort of unpack all the requirements until you actually know what you're looking for concretely. And you write it down in a system on paper and everyone agrees to it. That if we did that, that would solve, I think, 99 percent of it because. I feel like where a lot of the slightly more fuzzy feelings based things come in is you'll get halfway through a process and someone will have their own view of what you're meant to be hiring that might not be correlated with the other people's. And they've, you know, use their memory of some other analyst they'd hired who was useless for some reason. And now they're kind of importing it into their framework. But if everyone had figured that out from the get go. We would be able to distill it down to something that's actually words on a piece of paper. Oh, I understand. You want someone with a high attention to detail, who's not going to make stupid mistakes. Like that's really what it comes down to. And so we could like unfuzzy a lot of this stuff. If we really spent more time doing it, what do you think of that approach?

TIANKAI: yeah, no, that sounds actually really good. I think that if you would detail out a lot more of the requirements and the responsibilities of a role that needs to be hired for, that would help a lot, but I feel like. There's the other side of it, right? Where companies, especially bigger companies, they actually try to standardize role descriptions, right? They actually don't want a very diverging role descriptions being out there. And that every role becomes like a hundred percent new description. They want usually it to be like 80 percent same and. High rate managers only have like 20 percent freedom to change what's in there. And that feels like the opposite of what you just described, right? As in that, if I only have 20 percent that I can change of a doc description, that's never going to be enough to actually go to the detailed level of what I need to describe what I want for. And the consequence is. That those job descriptions are not clarified up front, they only get clarified in the conversations and people then get surprised that people are not a right fit or is also maybe especially good fit, but that only appears after like already some time has passed and you had already screened most of the people. To then find out, Oh, we may still need to actually talk about what the job actually entails. And now it's a little bit, no, too late in the process. I, I mean, and that has its reasons, right? I know that from an HR point of view, having standardized job description, everything makes it a little bit easier and more consistent, especially. To have right candidates you actually can also find replacements better and you can Find internal rotations in a better way, right? But again that comes down to then not being able to be super specific hire new people

TIM: Hmm. That's interesting perspective. I hadn't heard that angle before it's almost like if you had like a Almost like a parent job. This is like the general data analyst we're looking for. But then within that, you should be able to say, well, actually for my data analyst in my team, here's exactly what I need. Because as you say, the later on in the process that comes to the fore, the more time that wastes for everyone, the candidates, the hiring managers, the interviewers, if it's only really by the third conversation that you've really gotten down to what you're needing in the role that is just so inefficient.

TIANKAI: I mean, even with the example of a data analyst, it can vastly differ between different business domains, right? Like a data analyst for sales has a very different one than a data analyst for R and D or a simple plug, right? So it's different data types, different stakeholders with different mindsets, with different resistance levels and collaboration attitudes. Right. So I can see it diverging a lot between those two job descriptions of what actually is needed.

TIM: In your book, you talk about putting people at the center of data strategy. Mm hmm. Have you thought about how this might apply to hiring? Like when you've done hiring in the past, if you take an almost like a people centered approach to hiring.

TIANKAI: I mean, I think I very clearly address it. How the Soft skills are also equally important as much as the hard skills, right? And that we are not actually focusing enough on figuring out if people bring the right soft skills to the job or not and Part of it is very simply put that in specific data jobs like data governance there's always a high level of resistance and a high level of misunderstanding of what the Area you're working actually means And that, for example, implies that whoever's doing the job needs to be great at breaking the ice, dealing with conflicts making connections with people, being able to translate complex concepts into simple words and being able to make people understand. And those are skills that are not usually written down on a data profile, right? Usually start with like SQL, Python experience with certain tools, right. And not go on that level. And I think someone's Simon simon Sinek is like one of my favorite kind of organization of psychologists. He said hire for attitude, train for skills, right. That hard skills are always easier to teach and to adopt. Then having a certain attitude or a certain personality, right? And so how do we actually then hire for the right personalities instead of only looking at their hard skills and the technical skills? yeah, so I basically address it by saying when we then hire people and we have job interviews We should probe it, right? Let's ask people more about when they dealt with a conflict and how they actually what they learned from dealing with a certain conflict And just by the answer and how the answer is given, you can learn what the attitude of that person towards conflict management even is, for example, right? Or how, when did you have a time where nobody understood what you were working on and you basically successfully communicated to them and clarified what your focus is, right? And these kinds of things in a conversation can really help to, I think, also reveal how good people are in their soft skills. And that for me is like a more people centric way of hiring for data professionals.

TIM: I've been caught between two minds recently when it comes to the development of AI and what it means for the skills people need. Mm I feel like on one end of the spectrum, I could imagine a time in the not too distant future where super technical things, let's just take coding as an example, in a year, it wouldn't surprise me if no one would ever write a line of SQL from scratch ever again, hmm. that you just put your prompt into Claude or whatever, it's going to write the query with 100 percent accuracy, given enough context so that really you don't, you don't even need to review it like it's now done for you. So now maybe SQL will be redundant. Maybe not. It's been around for like 60 years. So I feel like it's, it's like the cockroach of technologies.

TIANKAI: Yeah. Yeah. For sure. Yeah. Mm

TIM: but, but let's say I'm one of those end of the spectrum. It's like, okay, so maybe Python skills and be, you know, data wrangling, maybe the actual analysis, maybe SQL is slightly automated away so that what's left over is relatively more people or softer skills that let's say a data 2025. But then On the other side, I feel almost the polar opposite, which is like, well, a lot of the work is going to be done by AI, hmm. Mm so your skills to interact with AI maybe is going to be way more important than your people skills. Like if you had the combination of programming and acute knowledge of an LLM, that is an explosive skill set that is maybe worth more investing more in than it is investing in slightly more fuzzy, subjective people skills. What do you think?

TIANKAI: Yeah. That's a really good point. I would be very careful though, to say that AI can completely replace certain skills because in the end still need somebody who can check if what AI generated is the right output or the right outcome. Right. And that will always require human expertise over AI outputs. To, to double check that just for quality assurance other hand, you're absolutely right. I think we're going to explode in productivity. People are not going to write hands on manual SQL queries that much anymore. They can just prompt it very easily and get a whole series of SQL queries from a chatbot or like a copilot, whatever it is. Right. And that's gonna change the way how we do things. So I think for sure, AI literacy or like dealing with AI. Tools is going to be a critical skill in the future to be able to navigate it and use it in the right way. But at the same time, as I mentioned before, we cannot forget to assure the quality of what comes out of AI, right. And just blindly trust AI to then take what's coming out of there to use it. Because if we all lose that skill to check if it's even correct or not, what comes out, there's going to be much worse things happening, right. Which means right. It's a little bit like. more superficial almost about coding and technical skills because you just need to be able to spot mistakes in the future. You don't need to write queries completely from scratch anymore or like write any language from scratch anymore spotting mistakes is different than writing a whole code on your own, right? So, a little bit of a mindset change, maybe a little bit of a skill set adjacent change but the way of how we're going to deal with technology is is changing for sure.

TIM: Feels like. Yeah, just another step of abstraction. So nobody writes machine code anymore. Nobody writes low level languages. It's just one step above that. Exactly. With, with yeah, needing to still be able to interrogate the output of the LLM to, let's say, verify that it's right, part of me thinks, Hmm. There might be like a quality quantity trade off going on. So I'll give you an example. Let's say at the moment you've got like a centralized BI analytics team and they get all bombarded with all these requests from stakeholders. Hey, can you help me with this report or help me generate this analysis? And probably those stakeholders are only asking for their top one request, like their most important thing. But maybe if they had some kind of AI analyst who can produce unlimited work, then they wouldn't be limiting it to one request. They'd be just bombarding them with endless requests, because they can get endless answers. Now, would you rather have like one human analyst produced report that's let's say got 95 percent accuracy, or a hundred AI produced reports that have 90 percent accuracy? In which case that step of validation is just gone because it doesn't really matter because as long as you're right on average, it doesn't really matter if half of the time you're wrong. You see what I mean? I wonder if we might end up in a, in a scenario like that.

TIANKAI: think what you're describing actually potentially happens today already, right? But there's another element to it of what you're describing, which is the prioritization of use cases, simply put, right? Whereas if we all agree on a business use case for data, let's say it's a critical reporting for auditing, right? that decides if we are compliant or not. And if we're not compliant, that's connected to a huge fine or something. Right. there better be some human eyes on it more and more. And it better be 99. 9 percent accurate. So we're not in trouble. Right. Whereas let's say for example, some operational reporting just to check, let's say some like walkthrough times or like some machinery running operational times, right. We probably are asking AI already to create those reports for us. And they're only 90 percent accurate because some sensors are not always reliable, but that's fine for us, right? We're only looking at the bigger picture and if there's a certain outliers that we need to really pay attention to and if certain patterns appear. I think for now it will really come down to how we prioritize higher and lower priority and how much we are confident that AI can take on more and more critical use cases for us. But my gut feeling says that we are going to stick with human in the loop for the critical use cases for a while. And only the more less prioritized ones and less critical ones will more and more be replaced by AI generated.

TIM: I'd be really interested to see what happens with this kind of new wave of AI startups that are like, you know, let's say V2 of the B2B SaaS products and let's say 10, 20 years ago, but now built based on AI to begin with. Because I mentioned a lot of them are coming to the full with some kind of trade off in mind where it's like, well, used to have to point and click around this software to do all this manually. We've now removed you as a bottleneck. So, you know, you don't have to do it anymore. It's not as good as what you would have done, but it's good enough. And it's done automatically. And so, yeah, there must be all these kinds of use cases where it's fine and we would only get better, but we're probably not going to let it fly a plane anytime soon or do the open heart surgery. I'd rather have the surgeon.

TIANKAI: Exactly.

TIM: What about when it comes to hiring? Have you started to dabble with any AI tools yourself? Have you seen candidates use AI on their end?

TIANKAI: Well, I mean to be fair, I, in my organization, when I'm doing the hiring. I'm usually not in the screening part. Right. I only do the subject matter expertise interviews and they are really face to face or like virtual face to face, but it's a conversation in the end. Right. So I don't think there's a lot of AI that is possible besides like a note taking assistant or something, which even for our data privacy reasons, it's not a great idea usually at many aspects. So I'm also not a big fan of it. I'd rather just take a class, traditional pen and paper notes that are easy. Easily destroyable, right? But I think for at least then making sense of my notes, and especially when I have many candidates, I remove all the personal information, but I try to make a little bit more sense of it of summarizing all my notes into like crisp statements for me and for me to then be easier to make an evaluation. Between like a more descriptive and like shorter descriptions, for example. Right. But that's more productivity based, not really making a decision for me. It's just, I, I, it helps me to summarize it instead of doing it alone for candidates. I would say I can only suspect them to use AI or not, but it usually comes down to how their resumes are written when I read them and it sounds a little bit too generic and not really specific where it sounds like somebody just ask the chat bot to Basically adapt and tailor their descriptions of the experience towards something and now it sounds really bland and kind of just uses all the right buzzwords but it doesn't feel like it has a soul almost, right? And that, that happens pretty often nowadays. I usually give people the benefit of the doubt because I know it's hard right now to apply for a lot of roles and not getting a lot of opportunities and if you need to save some time you want to just be able to deal with it all, that's fine. And then it comes down to the conversation. If people are honest about what they did do and what they didn't do and what the experience is, obviously, but I mean, for the face to face interviews, I said, I think there's so much AI currently being involved.

TIM: Yeah, I would have thought for most companies there would be none at the moment at the interview stage. What, what about in a year's time? So again, I feel like there's just gonna be this sudden wave of B2B AI driven SaaS products, including in hiring and recruitment everywhere else. That's going to, I think, radically change things. So that even with a human interview, which personally, I think in a year or two would just be the last step. I don't think there'd be humans involved much before that, I reckon, boldly. But even if the humans there, is there not something to be said for because it's very, one thing with an interview is quite hard to focus on the other person. Keep an eye on time, take notes, score what they've said, like, I don't know about you, but I don't have three processes in my brain, unfortunately. So would it not be nice to have yeah, like an AI interviewer there with you saying, Hey, like, by the way, when you asked this question, did you realize you were being slightly biased or you didn't really? record this one correctly, or you scored the candidate three out of six here. I actually scored them six out of six because of these reasons, almost like having a another interview there just to, you know, nudge you in the right direction. What do you reckon?

TIANKAI: Oh, a really good question. I haven't even thought about that. I think it might work, especially if you can train it and fine tune it yourself, right? That it fits to your style more and it gives you a certain kind of feedback in the way you are. That could be really helpful. I mean, I think it's really nice if it tailors the insights and the support towards your needs as an interviewer. But it doesn't replace your decisions, right? It just gives you recommendations and gives you a good reasoning why certain recommendations are there, but you actually still as a human decision that you make in the end, I think that could really work. That could be a really good one.

TIM: I, yeah, personally hope something like this gets implemented. Especially because I think at the same time, it should be able to solve at least, well, Technically, it could easily solve candidates biggest gripe, which is no feedback, because if the feedback is recorded in a system, it's available for you there, it's integrated, it's like, automatically going to be sent to the candidate, if you've set it up that way, like, in theory, it should solve their problem. What I fear will actually happen in reality is though, that companies will be very reluctant to have it by default, that they're going to tell candidates anything.

TIANKAI: Right,

TIM: Because then they can be worried about, oh, like, What's the feedback? What have you said? What has the EIS said? Is there anything potentially biased in there? So to kind of default towards telling them nothing rather than 99 percent of the time telling them something useful just because you're worried about that one in a hundred time. I feel like being cynical where I might go.

TIANKAI: Yeah, you're right. I think with the feedback you mentioned, that's a really good point. I mean, having been a myself I know how frustrating it is, right? To not get any feedback and you get ghosted. Not even a rejection officially comes in. So you just forever not having any contact anymore. I think though. Would it be better to get an AI generated one or nothing at all? Because an AI generated one seems equally as much, could seem even more, like soulless and you're just being treated as a production outcoach rather than somebody taking the time to actually. you the respect, right? Because in the AI generated method is still not a respect message. And it's in the worst case might be even irrelevant feedback for you. Right. So I'm not sure it's an interesting one. I'm thinking of it from both of innovation enthusiasts point of view, but also like job candidates ex job candidates who is. Yeah. Who are thinking how, how you, I would deal with it basically of getting an AI generated feedback message.

TIM: Yeah, the framing of it would have to be really Done carefully, because as you're right, as a candidate going into the interview, you might be expecting human feedback. So getting AI feedback is actually a disappointment. I agree. Yeah. Even if you realize on aggregate, most of the time you get ghosted. And so any feedback is better than none. I could appreciate how in that moment for that interview, you might think otherwise. I wonder if it could be almost like if we could make the process of giving human feedback so easy. it's like, it's helped you reword and yeah, yeah, yeah,

TIANKAI: And you just have to like adapt it and adjust it to whatever you think. Yeah. That's, that's a good one

TIM: maybe something one, one thing I think that will come about as a result of using AI and hiring is I feel like It will start to shine a light on the inherent subjectivity and vagueness of our requirements. I, I kind of feel like it could play this role of a psychologist in a way where if it actually looked at all the comments of every interviewer of what everyone's looking for across the candidates, it would see, it would show us the incoherence in what we're looking for. It would show us the bias. It would show us like you, do you guys know what you're talking about? I'll give you a quick example. So. It's working with someone six to 12 months ago, and they give us feedback on a candidate, and they told us something like, oh, I, they, they, they weren't red enough. They were a bit more yellow than green or something like using some kind of maybe an HB, HBDI framework, whatever that buckets people into colors like, okay, sure. And then the other feedback for them was, oh, I, you know, I just, I wasn't, they were good, Tim, they were good, but I'm not sure they were really anti Putin. That was the feedback.

TIANKAI: a feedback.

TIM: So what do I do with that? Like, I'm obviously not going to pass it on to Ken. Hey, by the way, you're amazing analyst, but we weren't convinced of your political leanings vis a vis the war in Ukraine. Well, I can't say anything there. So I feel like that's a very difficult conversation to navigate. One thing I think AI is great at, which isn't talked much about, is giving feedback in a non, in a way that can't be defensive because you're getting it from an AI, not from a human. It's like, Oh, by the way, did you know, you mentioned this you might want to consider that cause that's not part of your criteria. Nowhere in the job ad does it say must be pro Ukraine and anti Russia or whatever. Okay.

TIANKAI: Yeah. Wow. That's a really good point. I mean I almost feel like though, because AI assistants need to be trained by human data too. Right. But. If a lot of people are still giving that kind of feedback, then we're definitely feeding the wrong things to an AI assistant to it. Right. Then worst case scenario, an AI assistant just takes care of political affinities and only looks at that and then uses it actually at a feedback email. That would be horrible. Right. For example. So it's a little bit like we need to first role model it for the AI assistants before we can let AI assistants do it, because otherwise they're picking up the wrong things too.

TIM: Yeah. I, I think the, the goal of what we would want would be some kind of as objective as possible system, but yes, hasn't been optimized based on all our neuroses, biases, discrimination. And, you know, the fact that I interviewed a candidate who had, you know, my ex girlfriend from when I was 13 and I had this thing in the back of my head, which didn't like people that name, like this kind of ridiculousness that we must bring to it. I just feel like we have this amazing opportunity to improve this kind of. layer of subjectivity for the better, albeit we can make things much worse as well with AI. So it is a careful balancing act.

TIANKAI: Awesome. Yeah.

TIM: So yeah, we'll, we'll see where things lead us. I think on aggregate, I'm quite optimistic about AI and hiring specifically, because I think the current state of things done by humans has so much. So many obvious flaws to begin with, like it's so biased. And so some of the claims that people would make about AI and the fears, which are legitimate already, they're in spades in the current scenario. So in that sense, we, we, we're hard to make it worse versus other domains of the world where you can imagine AI creating unlimited lies and propaganda in a way that humans can't just because of the scale problem. But in hiring, I don't feel like we have that. That trade off,

TIANKAI: Absolutely.

TIM: what about thinking about your proverbial magic wand?

TIANKAI: Mm hmm.

TIM: So if you could take this magic wand, could be AI, could be nothing to do with AI, and apply it to hiring and sprinkle a little bit of that fairy dust, where would you be sprinkling it? What would you be solving? What would be like perfect, click your fingers hiring for you?

TIANKAI: I think an accurate screening would be nice. Right? I think far too many times still get candidates in the face to face interview phase that are just not a fit for the job and might not be only negative sides where the person being interviewed has not the relevant skill set. It might be also the overqualification side of things, right, where it's just another different type of risk, but still a risk to get someone overqualified for it. noticing it only in an interview seems a little bit late when the real framework for deciding for or against that person. is very strict about it. And I always think that it's related to screening and screening rules, for example, right? But if AI could help a little bit from, by learning from the rejections in the past, and then creating own patterns of just a higher quality of people that go into the next phase or in the next stage of the interview process, think that would be really nice. That would be my magic wand wish, I would say.

TIM: I wonder then. If the unlock would be some kind of new data sets, you know, like any, anyone would say like garbage in garbage out, doesn't matter how good your model is, if your data is shit, then the output is going to be pretty crap. And I feel like in the screening challenge, that is part of the problem, which is we have normally just a CV. Maybe in an application form they filled in, maybe a LinkedIn profile. It's like, how much can you really tell from that? Especially now, given, as you say, a lot of them are written with Chachapiti. It has this generic bland feel to it that seems like it fits the job description, but it doesn't really feel real. So maybe we need a new data set somehow.

TIANKAI: I think so. But as was always with like, have being also working on data governance, right, I'm always a little bit cautious of information that's collected. And shared around personal data, right? And I think it can get very personal when it comes to CVS and everything that's on their resumes, cover letters, people share a lot about the personal motivation on there too. And that could be really quite tricky from a legal risk point of view. other than that, I would actually agree. That would be really nice. And I've also thought about how can we anonymize these data sets and so on. That would be really good. Right. But it's a really tricky, you know, that led, if you, if you anonymize it and it's still just a sample of one that has a unique profile, you can still map it to the one person that has exactly that skill set in the world, for example. so yeah I'd only be worried about let's say the legal consequences of it, but in theory, I would agree that that would really be helpful.

TIM: Funnily enough, I've seen talking about that anonymization problem. The opposite happened recently. I'm not sure if you've seen these posts, but. You know, a recruiter would typically post their job ad in reference to like our client dot dot dot And they anonymize it so that they don't have competition coming after them But I saw a few people using Chachi Petit to figure it out So hey, like can you figure out what the company is here? And it seems to do it quite successfully by piecing together the dots I wonder if you could do the same thing with a candidate Give it a CV, anonymized or LinkedIn profile, yeah, because it might look up the LinkedIn profile and do some kind of comparison, a little bit harder with individuals, I would have thought that a company but,

TIANKAI: right.

TIM: One idea we're playing around with was maybe there needs to be some kind of intermediary between the candidates and the company, because if you're, let's say a data analyst and you go and apply for a bunch of data analyst jobs. The first couple of rounds, the correlation in what you're asked is very high, like you're going to be asked about what you've been doing in your last job, what do you want in your next role, where do you see yourself in a couple of years, what's your skill set, like it's pretty predictable, most of the time I would say. So maybe there needs to be some model where candidates just pre answer in some kind of validated way a whole bunch of common things and then it's just like a web hook for companies to then pick up the bits that they care about.

TIANKAI: Right.

TIM: if something sits in between like that, I don't know, what do you reckon?

TIANKAI: I like that. But at the same time, I appreciate not only the, the answer itself and its content, but also the way it's being trans transported through mimics, facial expression, right? Kind of the voice and the tonal thing, something that you only pick up as human being. And of course you could digitalize it, right? You could do facial recognition, voice recognition, everything. But that makes it also very complicated all of a sudden right that you want to know decode it into emotions and likelihood if they're lying or not and so on which you usually pick up as a person So yeah, there's something more efficient to it. But at the same time, I think it would be interesting to also Get the human touch of it, right? And see how they respond, not just what they respond as well.

TIM: This is something we were always torn on with two different directions with our products and business, which was let's say for video answers from a candidate, should we be on one hand, anonymizing it, which would be getting rid of the video, transcribing it to text and then grading the text as an answer. Because that would be, yeah, then you don't know who it is that you can't introduce a bias around that background or whatever. But then, as you say, clearly you're losing some information as well. So it's like you've lost some noise and you've lost some information, for sure. around how they've actually delivered it. But then is that, is that, is that just another bias? Because, you know, like how much does it matter how they intonated exactly what they said? I don't know. It's a tricky one.

TIANKAI: Yeah. I think I would have to think about that a little more because can get really creative because those are open questions, right? And with open questions, you can get very different answers. And so what is the general comparability across those anyway across candidates, right? And what do you actually compare? Because it might equally as much likely that they all have the same answer where they see themselves on five years Or they have all very different answers where they see themselves five years And then you still need to judge it with more than just the text. So it's it's a little tricky, right? So yeah, not sure. It's a good thought exercise to have

TIM: You mentioned that, yeah, the CVs you've read recently tend to sound a little bit AI generated. Which everyone is saying, and that seems to be the case that now candidates would apply using AI what we're also hearing is that candidates might, yeah, so get the CV written with AI, but also apply en masse to roles with AI, so using some kind of like intermediate tool, which is then creating this weird scenario where there's just like a lot of applicants per role, a lot of noise as well, a lot of the CVs start to look like each other because they've all been written with the same AI. Alright.

TIANKAI: Yeah,

TIM: It feels like a very crowded and noisy market for candidates. If you were applying for a new job now, would you go through the same process? Would you be applying to jobs on a jobs board, kind of fighting it out with everyone else? Or would you try another method? Would you be like going from all the backdoor approach that the networking approach leveraging people, you know?

TIANKAI: I think I would try to do both which means first apply efficiently But then see if there's any way of being connected with a hiring manager to start with an informal conversation, right? And it really depends on the company culture and the hiring culture specifically if that's allowed or not I would appreciate though that if they say i'm not allowed to have that conversation formally with you Please follow the process, but then I did it already, right? I can just say okay fair enough I already applied officially. I look forward to the next steps, right? And so do it the right way But also try to do it let's say through a more Different way, but if the different way doesn't work out You can always still fight it out with the others and you did it officially in the right way, too That's usually my approach

TIM: And is there something to be said also for Like leveraging your existing networks in that scenario, it's like trying to connect to the hiring manager you you don't necessarily know. But also, could your network unlock doors that you might not even be aware of exist?

TIANKAI: absolutely, of course, I mean looking at especially on linkedin, right if there's any shared connections If they can put in a good word for me and everything I think, sure. I mean in that case, I would try to, yeah, see if there's any other ways to put me in the favor of it. It's also one of the things, right. You might think about if that's fair or not, but creating that network and creating that personal branding itself is also a lot of work, right? So it was more about putting the work in and now having that network that you can benefit from and not having to then create a whole network. In the moment of applying for a certain job. So yeah, it's a little bit more of a work distribution, too I would always argue about if that's fair or not

TIM: Yeah, I remember reading a book many years ago now, the details of which I can't remember, but the title stuck with me. And I think it's probably the most important lesson anyway. And it was, dig your well, dig your well before you're thirsty. Yeah, Which is a good one, and yes, you don't want to be frantically trying to network with people while you desperately need a job. That's, you can avoid that at any way possible, then it's worth it, I think. Exactly. Yeah Tiankai, if you could ask our next guest one question, what question would that be?

TIANKAI: I would say And because there's so much cynicism and sarcasm around The data and ai space nowadays, I would ask the next guest What truly excites you right now in the data and ai space? That would be my question Right something where it's not like you just have to be excited because everyone's excited about it But what truly excites you why that would be my question

TIM: Excellent. We will posit that to the next guest, whoever that may be. And in the meantime, yes, thank you again for coming on and sharing all your wisdom and insights and thoughts. And including a discussion of your book here again, Humanizing Data Strategy, and it's available on Amazon, and that's where I bought it. Any kind of good book retailer would have this I imagine. Where can people find it?

TIANKAI: of regional ones. I learned also there are some Libraries that have it now as well. If you could check it out, but yeah it's available on As the most the biggest the bigger regional retailers to besides amazon.

TIM: Tiankai, thank you so much for joining us today.

TIANKAI: Thank you so much for having me. This was really fun