00:16
Hello everyone, my name is Josh Bersin and I’m privileged to interview Ashutosh Garg today to hear about the role of eight fold in this incredibly interesting world of AI in HR.

00:27
Thank you, Josh. I’m actually kind of a co-co-founder at PayPal and AI estimated us to have this conversation today, Josh.

00:35 -Josh
AI has suddenly become one of the most important issues in the world and certainly in the role of HR, what role do you see eightfold playing in this transformation?

00:45 – Ashu
We started the company six years back, and our key thesis was employment is the most fundamental thing in our society. And our thing was that can we actually use the power of AI and machine learning to understand people’s potential, not what they have done, but what they can do next.

So that we can live in the world of skills, we can really understand the skill capability of every individual. This is a full as a last one year as he has become an automotive industry. There are two things that are happening. One is the rate at which AI is maturing, enabling a number of processes that can be taken over by AI. And in that case, it fully AI is playing a leading role in assisting HR professionals in doing their job in a more consistent, unbiased fashion on one side, but second is with the changes in AI coming in. Every role in the world is changing, how I’m doing my job, how you’re doing your job, how are you in essence doing their job, and the implication of that is now HR professionals have taken the front seat in guiding the organization in this changing world of AI.

01:53 – Josh
How do you maintain privacy and security and all of these really important decisions that need to be made?

02:00 – Ashu
At the heart of this is it’s not about an individual. It’s not about who’s John or Mary. It is about skills. What is skill a person has, how these skills are developing over time, how this is reflected in the success they are having in their job, and what is it leading to their future development? So it’s really all about that. So as it would when we say we talk about a billion of people we really think of it as billion career trajectories, how different people have progressed, right? So we anonymize all the PII stuff about these people, and we focus on their capabilities, their skills, their potential and their career development.

02:37 – Josh
All of a sudden, there seem to be dozens of companies that are trying to analyze skills. tell us just a tiny bit about what you do to analyze skills that’s unique.

02:47
Six and a half years back when you started a company everyone was like, why a second question was what is a skill? how do you guys define this? is a skill is it a capability? is it some other key word artifact right. And what did the video from day one. We were like, let’s understand what people have done with the primary purpose. First of understanding what they can do next. So, it was less about what you have done, what is because you have shaped saying in your resume and as very simple analysis working through the details right is we look at every person what is because this person has in ear one, what is because this person has an ear to what is because this person has in year three, using that tool understand what is because in year one reflected into the learnability of his cousin yet what is cause in year two reflected in the learnability of this because in year three, so this way when a company is trying to hire this person, or trying to promote our internal mobility, they can say that okay, this person has had these crystal date, what it means about this person’s skill next year, so it has always been about skill learnability and future skills and skills potential.

04:01 – Josh
You were already thinking about a much deeper level of understanding from history from context, from career trajectory from projects from relationships people had with each other. And today because the skills problem is so ubiquitous, every vendor claims to have some form of skills technology.

My assessment is that you guys are at least one generation ahead, maybe two generations ahead of most of the technology in the market that’s trying to do this. And I think it’s because of your AI background. And I think it’s because of the problem you’re trying to solve the depth of the problem. And just the way you guys have been thinking about the problem for a long time.

04:43 – Ashu
When you think about skills. Another thing we think a lot about is the skill context. What is the context in which you’re talking about this skill? The other thing which has worked out well for us, and important for this industry is from day one, we were all about diversity, that how did we reduce the bias? And one of the conjectures that we have had going from day one was anything that is not relevant for the job should not be part of the resume.

05:12
And it started out as just ignore information that’s not relevant. That’s not relevant. For the job, right? And simple thing was that your name is not relevant. Your age is not relevant. Your gender is not relevant. Your race is not relevant. Your ethnicity is not relevant. In fact, it does not even matter whether you work at Google versus Facebook versus Microsoft. What is truly relevant is your skills. And your learnability. So it almost came out of thinking about like what is relevant for the job and focus on that.

06:44- Josh
You right Fears of AI, regulating AI laws passed in New York State and other places.

What do you think we should do about regulation in AI? And what role does Eightfold play in all this?

07:02 – Ashu
I think like any important technology, any important development in our society, as they say, with great power comes great responsibility. And AI is one of those things. AI is extremely powerful. It can do a lot. But if you’re not being responsible with it, you may not go too far with it right.

Now there’s debate going on. Around who is responsible is when they’re responsible. Is the buyer of this technology responsible or someone else?

And frankly, the answer is all of this. As a candidate, it is my responsibility to ensure that I’m providing the right data. As an employer. It is my responsibility that when I’m using these systems, I’m using them in a way that we they are designed and not misusing them, using it to reduce the bias versus perpetuate the bias.

But these are complex systems. These are complex technologies, expecting everyone in the world to know the intricacies and the details of these is high.

Take for example, open AI. I’m using chat Jupiter, you’re using chat up, neither of us know what it is it was trained on. And what are the number of parameters of this model and who designed it or any of those things? Right.

So in that case, it is the responsibility of the vendor who’s also betting these systems to take the ownership of what the systems can and cannot do. So I see that these regulations as a good thing provided did not hinder the development and innovation over here. Ai should be developed a with transparency in mind. Ai should be developed with right analytics so that you can see how these systems are behaving. What are they doing is the intended use of of these systems the way they are designed, but then all three of us like vendors, users, and employers should take the responsibility to make sure that these systems are designed.

08:44 – Josh
So today, Geoffrey Hinton was in The New York Times, talking about the fact that he Abreu apparently is sort of one of the founders of the neural network. He believes that AI could end the human race. What’s your position on these kinds of conversations?

09:04 – Ashu
If you look at the history of last 60 years, 70 years is the longest stretch we never had any deadly war. And the reason is nuclear weapons. Nuclear weapons have stopped not create the war right? So if you think from that angle, right, these two countries are extremely powerful.

Can they be misused? Yes, they can be misused, right? You can really train an AI system to do extremely bad things. Yes, you can. Which is true for any technology out there. So AI is not the only thing that can end the human race, 1000s of other things in the world that can end the human race, including humans who are the most capable of handling the human race.

The world is still full of full of a lot of problems that need to be solved. Six months back, I lost my dad to cancer and I wish there was a treatment available. And what I was told by every doctor was don’t even bother third treatment was gonna be painful, and our commodity I wish the AI is advanced enough to come up with a solution to those kinds of problems.

The weather climate everywhere in the world is a mess right now. I wish AI was advanced enough to solve for that. I wish even though I live in California 10 miles from the ocean. And we don’t we struggle about the clean water. AI was advanced enough to solve this problem.

So I think there’s a lot AI can do in a positive sense. So my suggestion would be less focus in building this practice that are aligned to solve the world’s problems, including employment and HR. And let’s be responsible.

10:35
There’s a regulation in New York. There’s a couple of regulations are one in Chicago, but relative to hiring, of course, the EEOC already has laws that prohibit discrimination. I think you brought up a very interesting point. The EEOC is already here. Right? And they have been regulating for years what humans can cannot do in the employment process, right. And the way we look at AI systems is really nothing as a state intelligence, how the laws already exist, the laws already exist right here to control to ensure that right things are happening.

11:07
Buyers are the solutions that people are deploying these solutions should ensure that these systems are following those laws adopting adjusting to that. And overtime regulations will come. Those regulations will come in terms of transparency in terms of the data in terms of algorithms and graphs of bias. And as these are coming in, this is going to be an ongoing thing. This is not a one time thing. So companies should adopt it. Just keep looking at it. But the talent is today’s follow. You can’t wait for next two years, next four years for systems to settle this to settle before you adopt these things. It challenges your column today. You need to solve it today. And you need to do what is the most important thing. So we are fine, who is already there. Let’s make sure that we respect those and solve the next now we have new skills,

11:54
just flooding the world of business all the time and think about all the skills in AI that have come. Do you think the velocity of skills changing is something that even leads to more need? For AI or what what what is your, you know, thought about that topic? Your point, whether you have that skill or you don’t have that skill?

12:15
And typically we have thought about skills in isolation. Do you know Java or you don’t know Java? But let’s take it even simpler example right? Can you lift 50 pounds of weight or you cannot lift 50 pounds? of weight? Well, if I give you a pulley, maybe you can lift 100 pounds of weight but if without pulley you cannot lift even 20 pounds of weight right? And that is a new world of since 25 years back when I was doing my PhD internet was built. So you have to go to libraries. You have to find the paper journals to read and study and it was not available. It was not available. Today, the context has become extremely important. So that’s one access right? The context has become important. So you can’t think of a skill as a binary concept. In the context with the assistance available to me, do I know this thing or not? So with a brand new skill, it’s going to be very hard to find context. So you’re going to be building signals on context very quickly. So second thing that is happening actually, is what I call we are living in the world of super specialization is programming a skill is front end programming a skill is angular I skill is not a JSF skill, where do you draw the line? So the adjacency helps you at GCSE learnability and available systems around you.

13:27
So when I was coding at Google, because of the libraries that were available to me, coding and C++ was way faster than I could code in Python. But if you take away all those standard libraries, and coding and C++ becomes extremely slow. So thinking that context that adjacency that learnability has introduced multi dimensional to every scale, there’s no excuse or no longer single dimensional saying that you have the skill or not. On a scale of one to five we have to lie now. Well, let me ask you another question about skills. How have you

13:57- Josh
Well, let me ask you another question about skills. How have you guys been doing on soft skills?

14:03
I would say soft skills on both the most important is because of our skills are the most important. It’s the set I needed to get the job done, but also the skills that are most at the risk of creating a bias. We find people who are like us a lot more comfortable. We assume that if I have a power skill, someone who’s like me they will also the power is good right? So at eightfold what we have done is we have instead of folk thinking about whether someone has a specific power skill or not. What we have done focused on is what is required to be successful in this job. Learn from the people who have been successful in a similar job, what kind of trajectories they have had, what kind of experiences they have had, and how they have reflected into those power is because that have led to success over here.

So for example, I’m hiring someone who needs to be really good at public speaking, have they given interview or maybe they have not? Given many interviews, the analogy part of TV, but maybe they’ve been very active in extracurricular activities, which may have led to the development of our schools over there, like public speaking, or they have been doing internal tech talks. Now, here’s a very interesting example for you. When you think of public speaking, right. You would think of actors, journalist and key persons, all those things will come to your mind, right. But what would really come to your mind is a researcher. You will never think of machine learning researcher as a public speaking person. But think about it. This is the person who is supposed to publish four or five papers every year, go to faces conferences every year give keynotes and talks over there every year in front of an audience of hundreds to 1000s of people. So being a researcher can lead to Public Speaking at a scale and that is what we try to improve automatically.

15:45 – Josh
I think one of the big questions customers have about vendors as credibility. How do we know that you’re as good at this as we think you are? It for the eyes credibility comes from multiple places.

15:56 – Ashu
One is years of experience. I personally started doing AI 28 years back when you do things like Markov models, speech recognition, hidden Markov models, Bayesian analysis, and so on. And over the years working across multiple domains, it gives us a sense of how the data works, how these models learn, what is the generalizability of these models, what will work well in the field or not.

Second thing is the scale of the data that which which you are working. Today we are working with the data of more than a billion people across the globe. It’s not about one person here, one person there one industry or not right? It comes from the unique data sets of outcomes that we have collected over the years. So anytime when we go to customers. We try to collect as much data as we can, of course anonymize it, to preserve the privacy and security to so that we can learn from every outcome that enterprise has.

It comes from not only using the best algorithms out there, but actually pushing the envelope pushing the frontiers of advancement over here. A simple example is focusing on equal opportunity algorithms that enable systems to learn across every protected class and ensuring that the behavior of the system is same for man versus woman. Young versus old, no matter who you are, right. It comes from the analytics of people. It comes from the transparency that we bring to everyone. It comes from the patterns that we have filed over the years. I personally have more than 60 patents to my name. At eight fold. We have filed 20 Plus patents already multiple books you have written on this topic. During my PhD I published 40 Plus conference papers on AI. So have been doing it for years.

And I’ll tell you one central point is when we started the company everyone was you should just focus on tech sector. Or you should focus on startup or you should focus on English language. Today we have customers using us across 25 plus languages across 100 plus countries across 20 Plus industries. And the reason why we went with that is not just that the market is large, or the problem is widespread. But there was a very interesting insight over every large organization is pretty much hiring for every role. Google has doctors, neuroscientists on the payroll and so does our system administrators and salespeople. So it Salesforce. So it’s me okay, that difference is Mayo Clinic will have lot more doctors and Google will have lot more in my working with companies in healthcare, SPS were able to go deep in health care data, which learning we can bring to a company like Google on the so every company may have a narrow understanding of the segment they’re working in. But by bringing the data of everyone we can build much richer models.

So when we started the company, I had no idea what HR is. I didn’t even know what is ATS. And I would struggle to figure out what is HR is versus HRMS. But the interesting thing because of that was we never thought of the actual fragmentation. We never thought we are solving a talent acquisition problem or a talent management problem or a diversity problem or succession planning problem or lnd problem or career development problem or a payroll or performance problem, right? We always thought it is a talent. Enterprise need the best talent that can do the world wherever that talent is. So you can’t think of talent as a silo and the reason why that is important is once you start cutting through this entire lifecycle that is when you have the best understanding of the data. So that now you can think of who I am attracting who I’m hiring, who I am promoting, when growing in my company, one of the skill inquire I’m retaining over time right and what is because tell me that story what is consumed me that is story. So that has been our big focus area and a different approach to solving this problem.

19:49 – Josh
So as you know, I should tell workday, SAP Oracle have all been investing in AI. And they claim to have software that’s sort of equivalent to a fold in some respects, but they are fundamentally transaction systems. What do you think the role will be of their AI in the market versus yours?

20:13 – Ashu
First of all, I’m super excited that everyone is talking about AI and everyone has accepted that it’s no longer just a buzzword, but it’s a must have to solve the challenge. The challenge is, where are you starting first? Is AI an add on to your transactional system or is transaction and add on to an AI system? That is the fundamental difference between these incumbents and solutions like eight fold? We always thought that AI is the fundamental and it’s not even AI actually, in the talent space.

There are only three problems to us all. Understand every role and as part of understand the task and skills and everything with you understand every person and mesh that we think of that as the foundational problem that needs to get solved. While understanding that the data is going to be noisy, it’s going to be incomplete. People are going to misrepresent themselves. Job descriptions are going to be noisy, inaccurate reflection of the actual task that is needed to get them in the world of that noisy data. Can you solve these three problems understand people understand jobs and and mash them?

And once you solve that, everything else becomes a transaction on top of but on the other hand, if you start with a transaction system, you get limited by the data that you you get limited by the constraints that you impose. So for example, when we go to our customers, the incumbents are always saying give us the clean data. Give us the pristine instance is because have to be structured and limited. There has to be a strict ontology and if you remember the days of banking, Open Directory Project is a directory structure and at that time, it was very clear everything has to be that directory structure. Are you a healthcare company or an end computer science company or at a telecom company? Well, is Google a consumer company or an enterprise company? Is it an AI company or a search company? Nokia? Right there’s all of those things right? In fact is a healthy company or not. With verily they are a healthcare company as well and an automobile company as well. Right? So those boundaries are gone. In that world. You can’t live in silos. You have to assume everything is noisy, partial and incomplete. messy, but can you make sense out of it? Because that is the real world.

22:25 – Josh
So you started the company primarily focused on hiring initially, although I know your vision was much bigger than that. And now you’re getting into the whole talent process. Tell us the story of how you got there.

22:38
As I said earlier, we started the company with this primary problem of matching. And my thesis was if you solve the matching, whether you’re trying to hire talent, you’re trying to develop talent. Whether you go to a doctor and looking for witches who’s a good doctor for you, whether you’re trying to hire a patent attorney to see who they could borrow an attorney for you, you’re trying to find a plumber. Each of these are just current matching.

So our initial thesis was less focused on that magic problem. And if you saw that, as I said, I didn’t know HR. So it was not about hiring. It was not about HR. It was about solving that talent matching problem that we started the company around.

But then as the world has evolved over the next 10 years, the disruptions that we have seen in our society are greater than ever, probably between 2017 to 2020 of these were only focused on having so there were some of our customers who are using us for talent management, or people were like build our talent management and that they I already have a price I don’t care about that. I really just want to hide.

So that led us to focus more on building capabilities on top of our matching engine for hiring the purpose. As COVID happened. Initially, there was an aggressive slowdown, but very quickly companies realized that they need to retain the best and the focus shifted from hiring to talent management. Then over the following year, the focus expanded because not companies were not able to hire but they wanted to hire they were growing rapidly. That they are like I mean, I need to retain and I need to hire. So help me do both things. Now as those things are happening, and AI is growing really fast.

The third thing that is happening in our society is the rate at which the search engine is faster than ever. We say the half life of a skill is five years, maybe now close to three years or two years right. In that world companies realize that they can no longer hire people who have skills, but they’re productive. So as a result, we ended up focusing a lot on career development, career pathing, learning development for people as well. sufficiently many became a big thing. And now especially this year as these disruptors are reaching the peak. We’ve also launched workforce planning because now if you really think of researching organization in real time on an ongoing basis, because whatever tools it is technologies you have access to today, will change tomorrow and will change again. So you have to really rethink what is the nature of your workforce is it adopt to get the best out of these systems or not?

25:04 – Josh
In some sense, a fold is followed the path of the economy, your product has really evolved to deal with the newer and newer business problems as the economy has changed. You get measured on how bad

25:15 – Ashu
you are at predicting the future. And for that, first you have to understand where the world is. likely going to be next 10 years from now. Not today. Every year you have to inch the world towards that take people along with the right for example. Two years back no one was talking about skills. We like we have investment skills right last year. You’re like let’s start a conversation on upscaling which later last year early this year became the hottest topic. This year. We have we’re talking about workforce planning, strategic talent planning. Hopefully next few years you will see a lot more focus over here is skill based. Everyone is talking about skills, but skill based compensation is around the corner. That is what will become the hottest thing we our last few years. You have been talking about how you should think about full time employment versus contract worker in a single system single experience because there’s so we’re all about total and it’s no longer about the silos. I think our next three years we will see a lot more investment in that space as well. So constantly thinking about where the world is headed a little bit inching forward towards that. And so, that has been a focus area. How do you think HR people should explain?

26:21- Josh
How do you think HR people should explain AI to their function, their peers?

26:28 – Ashu
The simplest way I would say is think of AI as a human being who can read all the text that is out there in the world, make sense out of it can access it in real time and help with decision making. But at the same time, the person is a human being. So we’ll make mistakes. So we’ll have flaws in their thinking and approaches. highly scalable. Human being is how I think of it right? Or I think of AI as assistant I guess not artificial intelligence, right? It’s not magic.

But it can do is now and here. It is here to help you scale and do your job consistently at scale on an ongoing basis. Help you adopt access to changing realities. The world is moving fast. And if you don’t do it, you’re going to be left behind.

The other thing is it’s also a great normalizer. For me, personally speaking, I’m good with math. I’m good with algorithms. I’m good with programming, but I have a severe handicap in terms of my accent. English being a second language for me no longer first language. Now what wait till two years back. I could not make use of devices like Alexa and google home because they would never understand my ex. I would keep falling at&t And they will keep saying keep PCB, please, please repeat. But now, these devices understand why they’re able to understand my language. And that helps me do my job much much better. Or I can use various tools to fact check grammar check my English. So I’m no longer scared of writing a long email because at least I won’t make a fool of myself by writing a poor English, right? So in that sense, it’s actually a great normalizer it’s also breaking down the boundaries that we have across the geographic locations. Today, simply add eightfold. Our system is used in languages that I have no understanding of myself, but it’s not about what I know. It’s kind of learn from the data and deploy it.

28:35 – Josh
What do you think HR professionals and leaders should be telling the rest of their organization about AI that matters to them?

28:43
So first, AI is not going to take away the jobs. But AI may take away the job of the people who don’t adopt AI always the people who are up there because the expectation of the organizations are changing as a employee in a large organization, you are expected to perform much more now because of the availability of these tools and technologies, right? And even if you don’t other people in your organization, and even if people in your organization are doing it, other people in the documentation are doing it right. So at some level you don’t have a choice. Yeah is now in here.

I think the challenge that is happening is many people are afraid of AI is going to be my job. Is it going to destroy the world? Or the gloom and doom but AI is already in our lives. The reason why we have this reporting available today at a high contrast right is because of the here’s the reason why we are able to get the best medical treatment out there is because of AI AI is already here in our lives. And once we understand that, then it’s already there. Some fear goes down. There’s no longer that.

Second is no one is building these AI systems. Typically your job that’s the goal is to help you scale and do your job much better. If you ask any one of us right? And I’m having people the very first thing is are you ask a software engineer, right? Let’s just take a stereotypical example of that. They love system design. They love algorithms, but then they end up spending 95% of their time just coding and testing the software. EA can help you do that. With AI systems such as turning on the data. They don’t have the smartness and intelligence of humans, but then they can bring all the data to Limelight make it transparent, right.

And so, my ask to all the challenges will be don’t be don’t fear AI. Be thoughtful about deploying it. Focus on educating your teams. Focus on educating a peer group around what it can do. Don’t expect magic, expect assistance from AI systems, but then also put enough safeguards in place to track how the systems are behaving? Aren’t they’re making your life better or not? Take it delivering the outcomes that you desire.

31:01 – Josh
What do you think the role of it should be? In the selection or implementation or adoption of technology like a phone?

31:08 – Ashu
It needs to evolve from being the office of CIO, which is Chief Information Officer to Chief intelligence officer. Now the AI systems are actually not designed to solve the CIO’s problem at some level they are designed to solve the business problem. Now the focus is on solving business problems versus just solving that is relevant. Now, CIOs can play an extremely important and critical role in understanding how the systems are developing, how they’re being deployed, what data they are consuming, what metrics are being potentially, and if they do that, they can be the change agent for the rest of the organization and bring the business value.

So by solving it not just for themselves, but by solving it for the business outcome of the organization, CIOs can get in the front seat. One of the biggest problems to solve in society is employment. If you provide right career to everyone will make the world a much better place. There was a time when I used to think, what is the purpose of HR.

But over time, as my career grew, I realized that as HR leaders, we help everyone get the right job. We enable career and that is the biggest contribution we can have to our society. So if someone were to ask me, which is the single most impactful job in the world, it’s not doctors. It’s not even your teacher. It’s not law, law enforcement. It is really HR help people get the right career because if once we do that, everything else follows from the ashes.