How to Build a Skills-Based Organization Using AI
In this episode of the HR Leaders Podcast, we speak with Ying Li, Head of People Analytics at PepsiCo, exploring how organizations can leverage AI and data analytics to build powerful, skills-based organizations.
Ying shares her unique journey from electrical engineering and data science into people analytics, highlighting practical methods to understand skill levels, measure skills agility, and strategically plan talent development.
🎓 In this episode, Ying discusses:
Accurately assessing employee skill levels
Using skills scarcity data to guide talent decisions
Building a skills-based organization through analytics
Practical steps to maintain an updated skills taxonomy
Measuring and leveraging skills agility and learning propensity
Deel is the all-in-one payroll and HR platform for global teams.
Deel helps companies simplify every aspect of managing a workforce, from onboarding, compliance and performance management, to global payroll, HRIS and immigration support.
Deel works for full-time employees and independent contractors in more than 150 countries, compliantly.
And getting set up takes just a few minutes.
Chris Rainey 0:00
Hey, Ying, welcome to the show. How are you? I'm doing great. I'm good. I can't believe it's already March the year exactly, it's already gone so fast. I don't know how the years disappeared. How things were you are
Ying Li 0:17
things been going pretty well. Had a busy start in January, but a little bit quiet in February, but now things are ramping up again.
Chris Rainey 0:27
Nice, nice, yeah, before we jump in, like, tell everyone you have such an interesting background. So tell everyone a little bit more about your background, and then your journey to where we are now, and the role at PepsiCo.
Ying Li 0:37
Yeah, happy to Well, I would say that I have a very non typical path to my current role as the head of people analytics at popsicle. I have a PhD in electrical engineering, and then I started my career as a research staff member at IBM TJ Watson Research Center, which is based in New York. So I spent 13 years there, doing research in various areas such as E learning, computer vision, computational creativity and the lay down, spanning into digital marketing and the finance. So I did a lot of different things, yes, yes, yes. That was really fun. And then I spent one brief year with RBM global business service unit as a principal data scientist there. And after that, there was an opportunity to lead the data science team in the people analytics organization at corporate HR. So I just jumped on that opportunity, you know, without knowing much about HR business at that point, yeah, well, I should say I knew nothing about HR at that point, yeah, except that, you know, we know HR checks healthcare to pay people, but that's all I know about. HR, yes, so, so that's that's very interesting, but you know that, not looking back, like almost eight years later, yeah, I was really grateful that I took that opportunity and transitioned into these people and analytics space, because it's such a fascinating area. You know, there's so much you can do with data and insights. Literally, the sky is the limit. Yeah? So, yeah, so that's kind of how incredible, yeah, I would say, okay, so, so I did, like, I spent, like five years there, leading the data science team at HR, at IBM, HR people analytics, and my area, like many, focus on developing advanced analytics and AI solutions to really help attract, retain and develop our employees. So that was really good five years, you know, I gained a lot of HR to my knowledge. I wouldn't say I knew everything by HR, but I did learn a lot about the HR business, and then also used my technical skills to develop solutions that really touched upon people's lives, you know, which really gave me a very profound sense of self accomplishment in that perspective. I love that. So that role ultimately led to where I'm now at pabsco, leading the people analytics team. Yeah, well, great
Chris Rainey 3:38
journey. I mean, yeah, it's so fascinating, because this funk, this this role, didn't even exist right until the last maybe 10 years. That's right. So you were really the first wave of people analytics lead. Imagine when we look back years from now, it's pretty it'll be pretty cool to think that you were part of the first generation of people analytics leaders, right? You know, when I started, it was HR analytics, now and then and then, obviously people analytics, and it's kind of evolving now, but that's pretty cool that you were part of that wave of innovation, and it continues now with the age of AI to can only, only get even faster, and the things that we could do now, we kind of dreamed of five years ago. That's now. I'm sure it must be really exciting. What's really top of mind for you right now? Because there's so many topics that mean you spoke about offhand. But where do you want to start really, what's what you most excited about? What's really top of mind for you right now?
Ying Li 4:38
Yeah, so like in these, you know, eight years in this people analytics space, I've done a lot of things you know, which really touched upon almost every stage of an employee's life cycle within the company. You know, from talent acquisition to retention. Engagement, development, performance, performance management, right, the whole journey. But one topic that is always, always very dear, very dear to my heart is the skills. As I firmly believe that we need keep refreshing our skills. You know, given how fast the technology is evolving, right, and how quickly the companies are trying to catch up with all those changes, right? I'm sure you have heard the saying that skills is the new currency. Yeah, you know, so, right? So all of us need to have the necessary skills to remain competitive in the market, yeah, yeah. Let's
Chris Rainey 5:39
start from the beginning then, because I know you're going to tell me the importance of the foundation. So let's start with the foundation, right? Yes. How do we understand people skills, using data analytics and AI talk me through the foundation. How do we build that first?
Ying Li 5:56
Yes, that's a that's a great question. So I, I'd like to, like a comment on, like a four different areas that kind of builds the foundations around skill and that will really help build a skills based organization. Or, people have been talking a lot about skill based or skills powered organization, but really, you know, we should start from the foundations. So one thing I think it's very important is really to understand people's skills, depth or skills, you know, the level of their skills. You know, when you talk to people, you know you will hear people say, Oh, I have skills in machine learning, for instance. But it's really hard to know that. Are they, you know, do they only know machine learning as a beginners, you know, they they probably just took a couple of courses, and they think they have the skills of machine learning right? Or they actually have a lot of experience of using machine learning in their jobs. Have developed machine learning based solutions, right? Or some of them even could be, you know, thought leader in the in the machine learning or AI space, right? So you can see there's a wide span of different levels or depth, you know, for a particular skill, right? So, and knowing about people's love of skills is super, super important, you know, to me, because that will really help us kind of plan on rescaling upskilling people, right? You know how to match people skills with the particular job positions, you know, right? Especially internally so, so that that's why I think this is really one of the basic things that company needs to really start with.
Chris Rainey 7:44
Yeah. How do you do that presently? Yeah,
Ying Li 7:48
yeah. So, um, there are various ways to to do that, especially using data analytics and AI technology. So, um, so let's start with what kind of data we want to tap into to to understand people's depth, right? So, So, generally speaking, you want to really tap into people's digital footprint within the company. So for instance, like you know, you want to look into the internal social network like it's a viva engage, or the slack where people they communicate, they have the discussions. They talk about their work. They do the knowledge sharing, right? So they, you know, they jump to the problem solving. So that definitely give you one aspect into what people know and how much they know about specific topics or specific things, right? And then you can also look into people's profile or resumes. You know, I'm sure companies have resumes, but about their employees, maybe not 100% complete. But when you apply for the job, or you got the job, you have the resumes in their database, right? You know, and also, in your learning platform, you have people's, you know, profiles there, right? So that's another data source you can look into. And also, you know, looking to their learning and a training history, you know, you can sort of see what courses people have taken. What are the skills related relate to those courses, right? What are the training programs they have, you know, enrolled in so, so that history is another good data source to look into. And then you have all the badges, certifications that definitely speaks to people's skills, right? And then publications patterns, you know, like, if you have R and D, right, you have research divisions, definitely, that's another really, really good data source for you to look into, understand people's skills and their levels, right? And then other things are like performance assessment, Project assessment, that's, again, you know, another one. Ways to understand people skills and how good they are, right? So you have, like, all these, you know, really a very diverse set of data you actually have within the company that you can really tap into to understand people skills, and then you build, like a machine learning model on all the data that you have, right? So the model can then eventually speed up, kind of classifying people skills into, let's say, from level one to level five. Level one could means Entry Level and Level Five could be expert or thought leader, right? So that's kind of, you know, like very high level ways of really understand people's skills that
Chris Rainey 10:44
love that. And I suppose, because you're looking at it through the skills lens, you're finding skills across the business that perhaps you didn't know exist, correct? That's right, yeah, like you, you weren't in the HR function, you had the skills that were transferable to the people analytics role, right? That's exactly right. Yes, yeah. How often do you do that? Do you do, like a skills depth analysis? Is that what you is that, what you'd even call it, the skills, right? I've heard of inference, but does inference also mean depth? Because they're kind of too Oh, is that the same thing?
Ying Li 11:29
Not necessary? But yeah. But in this case, we are inferring people's love of skills. You can think, okay, that way, right? And the inference also gives you a sense that you have science behind it. You know, you love to see the technology to infer, right, infer from the data about people's skills levels. Yeah, so, so to your question, How often do do we do it? It really depends, you know, depends on, like, your sense of how fast people skills will change. I mean, people, do you know people skills do progress right over the time, but do you think they will change, like on a monthly basis, quarterly basis, right? Or, you know, annual basis, right? I mean, you certainly do not want to do it on a weekly basis. You know, that's too much change. But you know, maybe quarterly refresh makes more sense, right? And, yeah. And then another, another aspect I want to point out is that the validation aspect, you know, Chris, you might be kind of wondering, how do we know you influence result is correct and accurate?
Chris Rainey 12:40
Yeah, that's exactly one
expression. So
Ying Li 12:46
one way to do that is the validation piece, right? You know, you I mean, AI model will never speed up 100% accurate information, right? So, so there's definitely a validation phase. So, I mean, you could do like in various ways. One way is that you could, you know, share back the inferred skill levels with the employees. And you can tell them, Hey, Chris, here's the list of skills we, you know, we mind from all the digital footprint. And then here is the level of skills we inferred along each dimension of skills, right? You know, go take a look and tell us you know whether you agree or disagree, and if you disagree, you know, give us your self readings, right? So that's kind of self validation. And, I mean, it's good, you know, you are making it transparent to employees, because essentially, you want to leverage this skill, steps data for some downstream application, right? So you want to kind of make it transparent to employees so they know that you're actually using all the data, and, you know, for good reasons, right? But a caveat of self validation is that as an employee, I do not have, like, a big picture, or global picture of who are the other people, having, say, the machine learning skills in my company, and how my skills compared to other people's skills on the machine learning, right? You know, you probably don't know everything, so you may feel very good about your own skills. I think I'm an expert on machine learning. Yes, right. Yeah, right. You may not necessarily know. Actually, there are many people who are better, you know, they have more deeper skills on machine learning than you are, right? So the AR model, you know, has this global picture, because we feed all the data about our employees into the model. So the AR model is sort of staying at much higher ground, if you will, right? So can have, you know, more fair assessment. So there's a pros and cons, I would. Say to to ask in price, to self validate their their skills. Are
Chris Rainey 15:06
you just leading your trust? Are you just leading your trust at that point? Yeah,
Ying Li 15:10
that's, that's exactly right, you know. I think there's definitely merit there that you share back the data within price. Have you know us, you know, have them? Have a voice, oh, yeah, right, a voice, you know, in this whole process, and you, you know, you take that, take their input with a grain of salt, right? You know, integrate with other information you have. So, for instance, you can also ask their managers to validate their skills. That's a good way. That makes sense, yeah, right. So you could have multiple sources coming to you, and then you can make a final decisions. Do
Chris Rainey 15:43
you have that as part of your skills based talent marketplace? Do you feed? Do you feed? Feed the data from take data from there, and because I'm sure people updating profiles within the talent marketplace, yeah,
Ying Li 15:54
yeah. So for now, um, I would say pops, not there yet. Okay, yeah. So this is more, this is more related to the work I did back at RBM. You know the skills inference work, because the fun, what is another layer of foundation? Because you need to have, like, a skills taxonomy to be able to do that. Yeah, yes, for doing the inference. Now at PepsiCo, we are on the journey to have that skills taxonomy. Actually, we call it global job architecture. And I, you know other people kind of frequently refers to that as well. So we are almost towards the end of building that job architecture, and once that's ready and launched, then we can do a lot more things on top of that, right? That's really level zero of the foundation. I'm glad
Chris Rainey 16:52
that you mentioned that, because that's something that many people miss, that I speak to, that aren't, that don't have the experience that you have. There's no There's no point having the talent marketplace if you don't have the skills taxonomy for the talent as well. What about because I noticed you mentioned there's four things. I think the second one you mentioned to me was around skills scarcity. Could you talk about that and how you share the skills both within the organization and also make sure that you can leverage that across functions. Yes,
Ying Li 17:27
skill scarcity is really about understanding how skill the skill is, both within and outside of the company. You know, you can consider this as like, how rare, right, or how little these skills is in the market. Now, understanding skills scarcity is very important, especially when you want to, like decide on whether I need to buy, build or borrow skills, right? Let's say, you know now we want to, we need people who has machine learning skills right now, if this skill is very rare or scarce in the market, then you'll probably have to pay a lot in order to hire that skills into a company, right? So in that case, you'll probably wonder, Hey, maybe I will be better off just to build these skills. You know, within the company, I can do rescaling and I can do upscaling, right? So that might, you know, be a cheaper solution, right? And then, or you may just want to go with the borrow option, if that skills is only needed, you know, for a short period of time, for instance, right? So that that kind of definitely will be an effective that you need to consider about and also understanding the skill scarcity can help you, like, make a differentiator strategy on talent retention, yeah,
or compensations. You know, if snow about talent retention right, we might say, let's say you have a product based, meaning that you will give salaries to people that you want to retain. Right so now, in this case, you have limited budget, and then you need to decide whom I really want to retain. Right now, if skills is very important to you, then definitely you want to retain people who have the skills which will be hard to hire from the market, right? So then the skill scarcity index, if you will, will become a key component when you decide or prioritize whom to retain, right? And even with what is a percentage of salaries, and that's all related, right? You. Yeah. So
Chris Rainey 20:00
when you, I know that was a question for later, but then maybe we could touch on it now like so when you, when you feel about linking skills to compensation, part of that is through the the skills scarcity and analyzing that and understanding these are the skills that we need to retain. Therefore, we're going to adjust our compensation for that
Ying Li 20:22
Exactly, yeah. So compensation, yeah. So another app, I mean, retention, so increase, that's a part of the compensation. But separately, you know, you have this annual salary planning, right? You know where you do the marriage increase to the employees, and then, you know, based on your compensation philosophy, you know, I guess most of companies probably is performance based compensation philosophy, but if you have the skills based compensation philosophy, then you would want to, like, invest in employees who have those essential skills to the company right now, how do you define those essential skills? And then definitely the skill scarcity is one key element there, right? You know, I want to invest and essentially, you want to retain those people, sure, within the company, yes, as well. What we
Chris Rainey 21:18
haven't spoke about yet is, is is skill agility, and what people's propensity is to learn? Love to hear your thoughts on that. Yeah,
Ying Li 21:27
yeah. I'm just, I'm not sure, Chris, have you ever heard people talk about skills, agility, propensity to learn in our Euro podcast? Not
Chris Rainey 21:36
that much, a little, but not enough.
Yeah, mainly from the chief learning officers that I interview, less so from the people analytics or these, yeah,
Ying Li 21:48
yeah. Okay, I guess, um, I mean, because of my background in data analytics, people analytics, so I tend to approach to those topics from more analytical way. But, you know, as you can imagine, that we do collaborate with, you know, with all the COEs, like learning, skills, talent management, right to work on that. So skills agility is exactly the topic that we kind of CO explode with our learning team. You know, people, we always say curiosity is a great trait, right? But how do we quantify people's curiosity? You know, it's, it's a little bit subjective, I would say, right? But then understanding people's curiosity or their propensity to learn is actually very important. You know, when you think about the skills planning, let's say you have a rescaling program. You know, you wanna have more people have the skills in cloud computing space right now. Let's say you have two week, two weeks of rescaling program, but you only have 20 seats, right? But then you have, let's say you have 100 people who are qualified for this workshop. Then it comes to the selection problem, right? You know who? Who are those 20 people that you want to select out of these 100 candidates? Right? Now, of course, there are many factors you can consider improvisation, but one element could be their skills, agility. You know, you do wanna read. You know, re skill people who have a high propensity to learn, and who are more eager to learn new skills, to apply the new skills into their job, right? So that will actually bring makes, you know, will lead to a higher success rate, of course, from the risk program. You know, not everyone will graduate from the RE scaling programs, right? You know, some major up, some may just forget about the new skills they learned, right? If they do not apply them in their in their daily jobs, right? So now that's kind of where the agility aspect will become important.
Chris Rainey 24:10
How do you measure for that? What are some of the key data points? Yeah,
Ying Li 24:14
yeah. Great question. Um, actually, we tapped into many different data points. So for instance, one is, you want to look at their the growth of their skills over the time. You know, we we talked about a skill step, yeah, so you can sort of see for their existing skills, how their skill levels have been progressing over the time, because you need to invest in yourself in order to improve your skill levels, right? So that's like one aspect into people's propensity of Learn to Learn, and then now we can also see, are they gaining new skills over the time? That's another great element to look into
Chris Rainey 24:59
that also shows. The Curiosity right to look beyond seek different types of skills. Yeah. Okay,
Ying Li 25:06
yeah. And you also want to look at are those new skills aligned with the company's strategic directions? I mean, people could be interested in many different things, but not all of them are important for for the job, important for the company, right? So, you know, to some extent, you can also look at like those, you know, the the hot skills or key skills that advocated about the company. I, you know, back at IBM, we have all these list of hot skills or strategic skills, it's like very well, clearly defined, so you can easily see how people skills aligned with that hot list
Chris Rainey 25:49
was that also shared with the employees? Yeah, they knew, yeah, hey, these are the skills that are in demand. Yeah, exactly.
Ying Li 25:57
Yeah. Okay, yes. It's all transparent. And even for each learning course, it's tagged with, you know, hey, this learning course will help you again, skills in these hot skills. So it's everything's very clearly communicated. People know exactly what skills they are, you know, building,
Chris Rainey 26:18
yeah, how I know you're going through the process right now with PepsiCo, but how are you have you looked at skills adjacency in the past? Because obviously there's you've got similar skills. It could be machine learning. It could be called data mining, you know, etc, data encryption. You end up having 20,000 skills in there. How have you approach that to making sure that your skills are taxonomy, you know, duplicating things.
Ying Li 26:46
Yeah, that's that. That's another great question. So, um, I guess you know, again, that goes back to the skills taxonomy, right? I mean, skills taxonomy needs to be refreshed over the time, because now you, you know the company with the new strategic direction, you may have new skills added to the to the repository. So that's definitely need to be, like, refreshed on a periodic basis, so that you have new skills to be added. And then probably some skills become obsolete, so you want to clean them out as well. Yeah, what we found
Chris Rainey 27:20
to be the best approach. Like, you know, some some leaders, your peers that I spoke to have, like, gone to the actual teams in those functions and asked them specifically, what are the skills, and then mapped it out that way by literally being hands on. I there's so many different ways that there's existing, there's existing skills taxonomies that you can kind of start with the predefined like, what have you found as the best way to define this and make it as small as possible? Of course, yeah, and not have too many
Ying Li 27:49
adjacent ones. Yeah, great question and very practical question. So back in our BM, at that point, we had career and the skills team, and we actually have had multiple career and skills team in a supporting different business unit. So those teams, they, you know, they were very close to the business unit. They know exactly what were the skills required to, you know, to support all those business growth. And those people they are, you know, they really help with, like a key, you know, keeping the skills taxonomy up to date, you know, clean out those absolute ones, you know, including new ones, right? So that's kind of how it was done at that point at the RBM, now with popsicle. Now, as I said, we are still building this global job architecture, and we involved a lot of, you know, domain experts from different functions, and they kind of helped work together to define the list of taxonomies, skill skills for the taxonomy. But it, you know, you could imagine it's pretty manual, yes, and I would say, yeah, once the initial version of the job architecture is built now for refresh purpose, you know, we can definitely leverage the analytics and AI technology. You know, we can look into all the digital footprint like I mentioned earlier, and we can kind of identify new skills that popping up, and then, you know, and see, you know, less frequent mention about, you know, certain other skills. And we can kind of use that information to update and refresh the skills taxonomy, you know, on a pure other basis, but, of course, for the human supervisions, yeah, so I think that might be a better way, you know, to save the cost and energy and effort, right? Yeah.
Chris Rainey 29:53
I'm wondering, I was speaking to a wrist. Do you know a wrist, the company, the. LXP platform. Do you know them? No, no. So they're like an AI. They create learning pathways and courses using AI. And one of the cool features they have is that the AI will message all of your employees inside of teams and slack and ask you what your challenges are. And based on the collective challenges, it will go and build the AI learning Co Op pathways specifically. So it goes out and does that. It'll go to your sales experts, your your R and D experts, and it will basically ask questions, and based on that, then it will go and build the learning pathways based around the skills that they also need to acquire. So I'm wondering if there's ever going to be, if there's, maybe there's already a solution that does this, where you could do the same thing, where you can ask employees around the questions around their skills, and within teams, within slack, whatever systems you use, yeah, and then the AI can then bring all that information back and then start to build some of the skills taxonomy with that information. Unless that even makes sense. I've just kind of just threw it out, yeah? Like, it kind of goes back to your point of trust, because you're trusting they're gonna say the right things and assess them. But it could be cool to do, yeah,
More from the HR Leaders Podcast
Anisha Thomas, Head of People at Inscribe.