Preparing the Workforce of Tomorrow | Shay David from Retrain.ai

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Shay David 0:00
data, of course is not a panacea. There’s no pixie dust, you’re gonna spray some data on this problem is gonna go away. But without data, you’re definitely not going to be with solid then of course, data companies algorithms.

Alexander Ferguson 0:14
Welcome to UpTech Report. This is our apply tech series UpTech Report is sponsored by TeraLeap. Learn how to leverage the power of video at Teraleap.io. Today, I’m very excited to be joined by my guest Shay David, who’s based in United Kingdom. He’s the CEO and co founder at Retrain.AI. Welcome, Shay. Good to have you on. Thank you. Great to be here. Now Retrain guys in the talent management platform, folks vary for enterprise and government agencies wanting to assess their workforce, their knowledge, qualification skills, using AI machine learning, helping them to retrain and recruit.

Alexander Ferguson 0:50
So help me understand what was the the challenge the problem that you initially saw and set out to solve with return.

Shay David 0:57
So today, there’s a huge problem that we call the skills gap problem. And when I and my two co founders, Avi Simone and Isabel payflow, yourself, joined forces in order to solve this problem, what we were looking at is a skills gap problem, which is not only very large right now, but it’s also growing really fast. And that skills gap problem is increasing, because there are two opposing forces that are happening at the same time. On one hand, there is a very significant change to demand, the nature of work is changing what it means to be a teacher, for example, you’re changing what it means to be a work in a warehouse is changing what it means to be a pilot and nurse programmer, but also to work in a supply chain, or anything else that you would touch potentially is changing. So the demand for skills are changing on one hand. On the other hand, there is a significant shortage of skills supply, because the education system as we know, it does not train people with the skills that employers actually need. Those two vectors combined mean that the skills gap problem is only expanding. And you’d think that you could solve that problem with a bunch of tech. But as it happens, HR tech is one of the most updated areas of innovation right now. But for years, it’s been very much neglected. And for the most part, HR tech is very dated, and just cannot keep up with the pace of changes. So we think that there’s a real tsunami coming to the world of work. And a lot of what we’ve seen with the skills gap probably is only the ripples before the big waves hit in that both individuals, departments or organizations, and sometimes even states and countries are really right now below the flood line. And our objective is to build technologies that’s going to take them to safer grounds.

Alexander Ferguson 2:53
You paint this picture of the skills gap issue. And, of course, people when they think of the future and AI the think of robots taking people’s jobs, and you described in our earlier chat as well. And now to have a tsunami mindset that it’s happening, but it isn’t here yet. You start to see the flood lines going back, but you’re describing a very near future where this gap is only going to grow. Is that kind of am I getting that right?

Shay David 3:20
Yes. And I think that that assumption is based both on hundreds of conversations with business line managers, CEOs, Chief people, Officer, Chief HR officers around the market, and also coming from research. Particularly if you look at research by people like Professor Carl Frey, head of the future working centers at Oxford, for example, together with his colleague, Osborne initiated a paper looking at the impact of automation on jobs. And based on research like that validated again, by hundreds of conversations with it, we understand that is some segments of the market, maybe 90% of the jobs are going to wait. Right. And we’ve seen we’ve seen that happen with previous technological revolutions, right, or the gas lamp lighters are no more, as well as the print offsetters as well as many other professions that have disappeared. And the same is happening today. But this is only beginning to happen. So in some segments, it’s really 90%. In other segments, it might be 10 or 20%. We’re using a conservative number, suggesting that 30% of jobs are going to either disappear or completely morph in terms of skill sets. And if you look at the world workforce of just over 3 billion people, it’s 30%. We’re talking about a billion people. They need to be upscaled and retrained over the next decade. So this is really truly one of the biggest problems facing humanity right now. I would say

Alexander Ferguson 4:46
you’re seeing that as 30% conservative number of the global industry. So a billion people are going to have an issue with this tsunami that’s coming and I actually you earlier had mentioned the words the the symptom is unemployment. The real problem though is unemployment ability unemployability where people just don’t have the skills to be employed, how to manage that? How are you then solving this? we’ve described described the problem that we’re in, but what we’re How are you approaching this issue?

Shay David 5:14
So we think that, you know, we’re all geeks, data geeks, direct means data, Kings. And we think that a lot of this could be solved with the right sets of data. Data, of course, is not a panacea has no pixie dust, you’re going to spray some data on this problem is going to go away. But without data, you’re definitely not going to even solve it. And of course, data comes with algorithms. So when we analyze the problem, we’ll look at the matching problem, we’ll look at a supply and demand problem. The first way of the solution is to understand demand. What are those jobs of the future? What are the jobs of the prison for that matter? What are employers hiring for right now? Most organizations, most large organizations don’t even understand their own hiring practices. Because it’s hard to know. You know, many departments, for example, are hiring in parallel. There’s very few organizations that have true centralized skill requirements. Certification qualification. Definitely, if you look at the state or country livers is no good knowledge was updated, granular enough information to understand what the job market looks right now. So the first order of business is to develop more modern labor market analytics software that looks at real world data of what are employers hiring for right now. But when we say what are employers hiring for right now, we’re particularly focused on skills, because we’re making a strong assumptions, again, validated with a lot of research and with hundreds of conversations from customers prospects, ecosystem partners in the market skills are the new currency is a very significant shift in the way that people hire. They’re hiring for skills, rather than for degrees. They’re hiring for skills rather than for job titles. They’re hiring for skills rather than for demographics. So being able to understand the skill requirements of employers right now, in real time, is the first order of business. How do we do it? We have automation AI, using natural language processes that analyzes millions of job posts every day, in order to understand what is it that the employers are hiring right now. So we get signal, and we separate the signal from the noise. And we’re creating a profile of an occupational tree that helps us identify what are the pockets of occupational opportunity? There is that one thing,

Alexander Ferguson 7:36
let me let me just recap to understand if I understand correctly, I’m following along. You see that there’s there’s this shift and you’re tracking, then all the the job postings out there using natural language processing to say, Okay, what are they posting? What are they asking for? What are they also expecting as far as experience and then you’re basically taking all that information back to then look through it and provide an answer.

Shay David 7:59
Yes. And we are putting that in a new type of data structure called the Knowledge Graph. Of course, we didn’t invent knowledge graphs, but we’re applying knowledge graph theory to that problem. So imagine that we can create this huge knowledge graph, where we start identifying demand and placing points on the graph to understand demand. Here’s what an ideal driver looks like, based on a million ads for drivers. We have both the demographics, the geographics, the skill requirements, etc. And that’s a hard problem to solve. Because if the ad looks like two forklift operators needed five years experience and willingness to work on the weekends, and natural language processing engine needs to take that very broad language and be able to understand what if this is about skills? What if this is about requirements, what it is about qualifications, etc.

Alexander Ferguson 8:53
that was gonna be my actually question to you is as job postings can be so unique and different, even if it’s, they’re doing the same job, the way people throw up a job post might be different, but your training your crawlers, your engines to be able to go out to and come back and distinguish and understand that it is truly this is the same thing, even if it’s described,

Shay David 9:12
that’s part of the magic is exactly that to be able to take that huge amount of unstructured data and put some structure in it and put it in our Knowledge Graph. All of that is step one, which is understanding demand. Step two, is about understanding supply to be able to place people within that same knowledge graph. And for that we use a series of assessment mostly coming from partner technology to be able to say, Shay is a forklift operator. And here she has credentials in here is she has experience in here she has skills, qualification, knowledge, personality boots, basically we assess for competencies, and we try to understand what the candidate can do the innovation and what we do is that we’re using a unified data set to be able to create It doesn’t matter. So we start in the first layer of our system, we just look at those raw signals. But in the second layer, we’re already creating maps, that on one hand, identify pockets of occupation or opportunity, identify demand, but immediately are able to place individuals within those same maps. So think about what we’re building is imagine we’re building the Google maps of the labor market. So step number one is create the map. Step number two is show you where are you right now on that map. And that takes us a very significant step forward, because now we’re able to already identify immediate opportunities for matching that could be used within an organization to start identifying talent that organizations could use for jobs that might not have considered for example, or it could be used by employment services around the world to start creating opportunity for job seekers, to be able to understand which jobs they are able to do. The difference between that and kind of existing systems is that most existing systems use Boolean search, you basically type in some keywords, set your geography, check a few boxes, about educational attainment, etc, and search. This is a new type of search. This is based on semantics. And based on knowledge graph, algorithms that might give you dramatically better results in terms of matching, because you might not even know what the dimensions of the search that you’re looking for are, if you’re just using the old system of going to Boolean search. So that’s step two.

Alexander Ferguson 11:34
If I if I can the same way to, to see if I understand correctly, you’ve the first part, pull in all the information of the job posting to get this knowledge graph of what is out there, and what a company is wanting. The second part is looking at the job market, the people that have the talent and figuring out, alright, what could they be doing over here and being able to connect the dots for both employers to say, Oh, this is could be a good candidate, even though they may not realize that or I didn’t realize it before, but you’ve figured out the data points to connect it or for them themselves. If they’re searching for something, they could search on this knowledge graph and be connected to job posts, that traditionally they wouldn’t have realized that their skill sets apply to, or if anything, it shows them how they could apply to it. Am I catching that?

Shay David 12:20
Yeah, and I’ll give you a couple of examples. We’re working with a big HMO in Israel, for example, and one of the things that we looked at is that when we look at that knowledge graph, naturally, many of the occupations in terms of skill requirements cluster pretty naturally. So in an HMO, or doctors would be clustered or the nurses would be clustered administrative assistants would be clustered or whatnot, where interesting results are beginning to show up in areas where the skill requirements might be close. But the job titles might be very different. For example, think about genetic counselor, and think about a community nurse in a traditional job hierarchy, those would be two very different professions. But if you look at the skill requirements for some of those jobs, they actually might be pretty similar in terms of their ability to communicate medical information with the community, the capability of showing high levels of empathy, the capability of working with young adults, etc, etc, whatever the excuse it might be in our system identify that automatically. So if one of the challenges of every HMO in the world, for example, right now, because the population is aging, so fast is defined, say community nurses have geriatric nurses to work with local communities, a system like that allows him from within their own talent pool to be able to understand who are the people that could feel this very high in demand job, for example, could be used people that maybe were hired for something else, or maybe they’re candidates, even for another job within the recruiting system, and actually offers them quick training to be able to meet a job that is in high demand. And the beauty of this system is that this could be done at an organizational level, but could be done at the cross organizational level as well. So that’s very significant. And that is step two. Importantly, there’s also a very critical step three, and that is the capability to map out training pathways within that same knowledge graph. So every person could do something, and most people would be close enough to some points in the graph to be able to get a job. Some people would not be close enough to get any jobs. That’s what we started talking about earlier about unemployability. Unemployment means you don’t have a match right now. unemployability and graph theory in this case means you’re too far away from any other points in the graph. Because you do not have the skills and the prerequisites to get close enough to any of the jobs. That’s a real problem. And the solution to that problem is to identify what sort of skills do you need to get close enough to those closest points. Every person has the closest Right. But the question is, are they close enough to those closest points or not? Sometimes they are below a threshold as I just couldn’t get improved. And that is what creates long term unemployability. And I was actually just writing about it, you might look at the the article I just published on Forbes is HR counsel. To look at the data, we looked at the Department of Labor Statistics Bureau of Labor Statistics data in the US. And we see that while overall unemployment in the US, for example, is declining, the share of long term unemployment is rising. So we’re starting to see real unemployability happen, this is not just a theory, the data actually shows that the proportion of people that are unemployed for the long term is rising, and rising significantly and rising fast. And that is a real problem that this system is a cluster. So

Alexander Ferguson 15:48
that third important stage is is to show that individual, I actually don’t match any of these, but here are the skills that you need to learn that can take you is the gap that you need to learn and train to get to this next stage, or the pathway for lonely, the thought that came in my head is almost like degrees of separation of of connecting with anyone in the world, they would say seven degrees of separation, it’s almost like you’re showing the degrees of what you need to learn and to get any job you want. Just start learning these skills and

Shay David 16:22
absolutely right. And I think that metaphors have kind of six degrees of separation is absolutely correct, because that game is also a type of a knowledge graph. And I think that if we discussed that, within that knowledge graph, to use Knowledge Graph terminology, the the nodes on the graph would be the jobs and the people, then the edges of the graph, the connecting lines between those nodes, would be the training path within the career pathways. So when I’m jumping from job to job, this is my career pathway, when I’m training to get close to a job, that is my training pathway. And graph theory gives us a lot of algorithms to be able to think through that problem. So the very modeling of that problem within the conceptual space already takes us, I believe a step closer to the solution. And importantly, I think that there’s a misconception in the market. When we talk about the use of AI or NLP to start solving this problem, a lot of people have a very basic misconception, these things that we’re talking about the fact that because AI, and automation is so prevalent, what we need to do, is to place everybody in jobs that are related to AI and automation. And that is a total misconception. Because potential is a would be a lot of jobs in AI and automation, there will be a lot of data scientist jobs, there would be a lot of robot operator jobs, and those jobs would exist, and many of those are jobs in the future. But in all likelihood, there are going to be less of those jobs in the jobs that are being destroyed. At the same time, many other automation related jobs are being created, might the old joke is that every grandmother a few decades ago, wanted her son or daughter to be a doctor or lawyer. No grandmother Ever wondered her son or daughter to be an Android developer. But it turns out that probably in 2020, Android Developer was probably the fastest rising job in 2021. Probably contact tracer was the fastest growing job. Nobody definitely no grandmother Ever wondered her son or daughter to be a contact tracer. But that’s the real power of AI and NLP is to be able to identify those pockets of opportunity not to place people and turn them into Python programmers or data scientists, even though those are definitely jobs of the future, but to really identify what those jobs are, could be and are becoming everyday.

Alexander Ferguson 18:52
Now, this is only about a year since since you started the organization as I read that about April of last last year. Now you’re No, you’re no. You’re not new to starting a venture or in the tech space. I mean that with your previous venture called tour, that was 1516 years. And I’d be fascinated to hear your journey. So stick around for part two to hear sheis to be able to hear that journey and insights. But for you, what do you see is kind of the next steps of being able to roll this out. And what can you share? Well,

Shay David 19:27
of course. So as you mentioned, we’ve been working with it for about a year. And where we are with this technology is we are rolling it out in production with our first design partners. Our first project is a project we’re doing with the Israeli government, and they’re kind of a HR tech incubator in organization called the JVC. to that and what we’re helping them build we are part of a big national data project about the labor market called our data. So we’re helping take that project to the next level. Part of the challenge for government related projects around the labor data is that the data itself is sometimes dated. But using our technology, were able to take the database from probably a 24 months average age of the data to 24 hour age. So a dramatic shift, taking data that is two years old and turning it into data that is one year old, because we have this AI capacity to scan so much more data. And if in a country like Israel, after they flattened the COVID curve, so rapidly now their main mission is to flatten the unemployment curve. And you’re using three year old data, you’re not going to get those million people back to work anytime soon. You need the data, it goes back to the point we were discussing earlier, you need the data and you need the algorithm to start solving that problem. Those are not sufficient conditions, but are the necessary conditions, if you want to get back to full employment, in this case with the controller, so we’re doing that. And we’re working with commercial partners to start deploying the system in production. So HMOs big banks retail supply chain, in order to start helping organizations both with internal mobility and with recruiting. But all of that is based on skills and skills is a new currency, managing those skills inventories, managing those skills was understanding that skill based hiring skill based onboarding, skill based training, retraining. upskilling is kind of the way of the future.

Alexander Ferguson 21:24
I the the future that you paint, it seems so golden to be able to know the right person, the right person knows what their job is. And to connect the two, though, I imagine pushback there could be some people think well AI can truly understand help in this messy field of human resource management. It doesn’t take a real person to understand an individual, what would you What would you say to something like that?

Shay David 21:48
First of all, I would agree with it. And I think that I think that, as I mentioned before, AI is not pixie dust, you know, you can just spray it in hopes that the fairies are going to show up and save you, you definitely need people in the loop. And we are approaches we call it organization in the loop. Because the biggest challenge in any AI system is training the AI system to get better over time. And for that we definitely need human intervention potentially while the system is learning and potentially overtime to also help it improve. And we’re not afraid to say the objective is not to get all the HR people out of the jobs. In contrast, this is a decision support system that gives HR professionals business line managers, CEOs, the data and the algorithms they need in order to take decisions all the way through the cycle of HR, what we call in our kind of HR tech language from hire to retire. Think about recruiting, hiring, onboarding, performance management, company benefits, scaling, upskilling learning and development, etc, etc. The entire lifecycle of HR could benefit from having good data infrastructure that is common among its elements as it could be used for smart decision support systems. So if the questions attend is, what are the job trends in my industry? Should they recruit externally? Or should they use internal talent? How do I make sure that if I want to upskill I’m actually upskilling? the right people for the right jobs? How does that connect to my learning and development systems? Right? How does that affect the comp and benefits? Could all these people continue on the journey? Or are some of them retiring? And that’s fine? Should they use freelancers? Or should they use full time labor? Where do I actually get the knowledge that I need to bring into the organization if I’m going on a learning and development journey and upskilling the people etc, etc. There are many questions. And today, for most organizations, these are solved with gut feelings with hunches. And some sophisticated organizations with some sticky notes in the fridge in the cafeteria, but for the most part, really without resistance in place. And I think that’s really kind of the next jump within that market.

Alexander Ferguson 24:03
I think technology that the power of it is enabling us to make more educated and insightful decisions. We’re still making the decisions and doing the work. But we have it at our disposal to be able to make those decisions. Thank you so much for for sharing this. For those who want to hear more about the journey that she’s going on, stick around for part two of our discussion. Also, for those who want to learn more about retraining, you can go to Retrain.ai. And they can request a demo over there. What is that a good first step for folks to take is there absolutely. There’s

Shay David 24:30
a free demos and free trial accounts. And we want to partner with as many people as we can. Our objective is to help 10 million people find good jobs in our first five years of operation, we’re well on our way for that. So we might actually increase that goal. But we need your help. So if you’re a large employer, in HR tech company in the ecosystem, or professionals within that field, this is one of the biggest problems that we’re facing today. And anybody that wants to collaborate with us is very welcome to contact us and see what this is all About,

Alexander Ferguson 25:00
I love the clear goal, how 10 million people or more find better jobs in the next four or five years is that what? Four years now? Time is ticking. So I’m excited to see that journey right. Thank you again, stick around for part two. And we’ll see you guys on the next episode of UpTech Report. That concludes the audio version of this episode. To see the original and more visit our UpTech Report YouTube channel. If you know a tech company, we should interview you can nominate them at Uptechreport.com. Or if you just prefer to listen, make sure you’re subscribed to this series on Apple podcasts, Spotify or your favorite podcasting app.