Episode Transcript
[00:00:00] Speaker A: Okay, everybody, today I am really excited to be with Ashutosh Garg, who is the founder and CEO of Eightfold AI and another company I'll let him tell you about.
Ashutosh and I have known each other for a long time. I think he's really one of the pioneers of AI in all aspects of hr, although it's been around a long time. I think Eightfold really brought the concepts of AI to market early on in what we called talent intelligence at the time. But Ashutosh, thank you for joining me.
[00:00:32] Speaker B: Thank you for the opportunity, George. Super excited to have conversations with you today.
[00:00:35] Speaker A: The first thing I'd love you to do is tell us about your background and your experience because I think you have a very unique set of skills and background that you bring to this whole space.
[00:00:46] Speaker B: Absolutely. So first of all, by training I'm an engineer, got my undergrad from India and after that was fortunate to get my PhD in machine learning from University of Illinois Urbana Champaign. Well before AI was popular. So this is in a timeframe of roughly 98 to 2002 is when I was at Illinois Urbana Champaign, get my PhD and during that time worked on the core machine learning algorithms and the application to variety of things from bioinformatics to computer vision to speech recognition. And later on in year 2003 I joined IBM Research where I started doing the deep research around understanding people, understanding web search. I was there for a year and then left that to start join Google research team. And that is where I ended up spending quite a bit of time in understanding personalization recommendations, search engines, how to analyze the content, how to understand people's behavior, what they are doing, what they are liking, how do you evaluate them? So my formative years were all around machine learning and applying machine learning algorithms to the problems of search recommendation, personalization.
[00:01:55] Speaker A: So what inspired you to leave Google and then go get into the HR tech space?
[00:02:01] Speaker B: So actually I left Google in 2008 and at that time I started a company, Bloom Reach. Actually that was in the E commerce personalization space. So grew that company to now that is a few hundred million dollars revenue, thousands of people, company across the globe, doing very well and serving some of the largest enterprises across the globe.
At that time the thesis was I know search, I know personalization, everything is moving to the web. How can we provide the best digital experiences to people, something that is very personalized. This is still the early days of Amazon, Netflix of the world, right? How do we bring the same kind of an experience to everyone?
Fast forward in 2016 started thinking that what's next for our society? What can I do to help people? And I always think of three pillars in our society. Education, healthcare, and employment. And given my experience both as an entrepreneur, as an AI researcher, as someone who has worked in large enterprises like Google and IBM, it was very clear to me that my calling is employment. How can we help everyone have the right career? Now, what made it fascinating is connecting the adults that if I can recommend people what they should buy next, or what video they should watch next, or what news article they should read next, in the world of employment, we are sticking to what they have done, not what they can do or what they would like to do. So we said, can you flip this problem on its head and say, start thinking of employment and career as about the opportunities we can offer to the people that match their potential versus just the work they have done. But that requires training, learning. So we said, can we actually start looking at what different people across the globe have done, how they have progressed in their career? And what it does is that it gives us insights into the learnability of each person.
So a very textbook, simple, trivial example would be, if you know Python, are you likely to learn Java quickly? Or if you know language C, are you to learn Java quickly? Now, for humans, we can do it for one person, one function, one role, one skill set, or few skill sets. With AI, we can learn across hundreds of millions of people and then do it for pretty much every skill, every role across the world. So that became the foundation of Eightfold.
[00:04:19] Speaker A: You know, when you first built Eightfold and I first met you and it was a long time ago, and you described it as the Google of people at the time, as one of the ways to describe it. There are many ways to describe it. And it was this big AI intelligent data system of personal profiles or professional profiles. Did recruiting occur to you as the big use case, or did that come later? Because I know you were always interested in helping people find the right job and find the right career. How did it turn into a recruiting tool or recruiting system?
[00:04:50] Speaker B: So I would say thank you, Josh. Two fascinating things. When I started Eightfold, I had no idea about hr.
[00:04:56] Speaker A: I know.
I mean, maybe that was the best of all because you built something that could do so many things.
[00:05:03] Speaker B: I didn't even know that there's a concept of applicant tracking system and hris, believe it or not, I didn't even know that there's a company named Workday or there was a company named Teleon. Like I was that oblivious to this space.
The Thing that I always found ironical was as a hiring manager, anytime I talked to any hiring manager across no matter which company, every hiring manager was complaining that they cannot find good people.
On the other hand, every friend of mine was complaining that they cannot find a good job.
And I'm like, something is broken in this process.
Then the second thing, again, because of my entrepreneurship experience before that, it was very clear that even when we think of career, you cannot solve this problem at an individual level. You have to really connect them to enterprises who can offer the job. So that led us to start thinking that how can we help people get the right opportunity? So we started talking to the HR leaders and that nationally gravitated towards recruiting and internal mobility. On day one. Whether I'm trying to help people from inside or outside, help me understand who's a great fit, help me see people's potential, help me understand what skills they can quickly learn so that I don't have to just seek outside. I can train and upskill these people.
[00:06:18] Speaker A: You know, the, the, the interesting thing, and I remember vividly when we first met and you came to my house and we spent a lot of time together, you were way ahead of the market in conceiving of the problem and the solution and understanding how AI could be applied, because nobody was really even talking about AI very much at the time. And then, you know, you successfully kind of created the category of talent intelligence and really captured the. I mean, most of the vendors in the market were afraid of Eightfold in the early days. I know that because I talked to so many of them. But now we sort of move ahead into the world where AIs embedded in everything and we see skills, inference or at least inferencing happening on most of the platforms we use all the time. How, where have you gone as a pioneer on this? Relative to the frontier models and the open LLMs and so forth that are available today, the way I have always
[00:07:15] Speaker B: looked at the problems is try to understand where the world is headed, not just where it is today. And now, especially with generative AI20 models, the world is moving so fast. If I just think of today's problem, I will be left behind in a year or two year from now. So you have to constantly keep on out innovating yourself almost, right? That's the first part. I think the core problem, if you take Frontier Motors, right, what I am telling is that these days intelligence has become a commodity. In fact, anyone can have intelligence for $20 a month. What is going to differentiate is still the human judgment. And human accountability.
What it also means is that the people and talent that means are 10x100x lot more valuable for enterprises going forward than they were even in the past because the X judgment will matter even more. AI will be able to do a lot, but the human judgment still retains with humans, if you make the wrong call, things fall apart. If you make the right call, things move forward. Right? So having the right talent is even is more important in future than it has ever been. Now, the second part of this is what does it mean from the perspective of frontier models? We have been using LLMs from day one, whether it is to understand passing the resume to understanding the job descriptions, to how do we bring these two things together to understand the skills, skill gaps on those kind of things. And we have done enormous amount of work both on the research side and application side on that axis. The second part with frontier models is can it enable us to do that was not possible in the past.
I would say one very simple example is one area where we have gone now deep is AI interviewing in the thinking over there works. Today, let's say 100 people apply for a job for one role, which is a typical industry average number.
As an enterprise, you can barely interview 5, 7 people because you don't have time and resources.
So 95% people never get any opportunity. Now, 50% may not have clarity skills, but 30, 40% may have some skills that are relevant.
So with 20 models now, can you build an AI interviewing solution that can help you assess another 35, 40 people, which can open the aperture, give lot more opportunities to a lot more people. Right. And help you better understand what other talent exists in the market.
So what we are doing is rethinking how the work will get done going forward, how the talent acquisition, talent management processes will look like. What can we do to level the playing field for everyone? More opportunity for more people, higher consistency, higher quality.
[00:09:42] Speaker A: So, you know, I agree with you. I think in some ways the AI interviewer is a Trojan horse into, and I don't mean in a negative way, in a positive way into hundreds of applications in business, because presumably, and I've seen your interviewers spectacular, the AI interviewer could assess skills and potential and capabilities and desires and passions of people and in a way that a human interviewer may never get for recruiting, for internal mobility, for coaching, for development, for performance improvement and so forth. Is that possible? Now is, is, is that sort of where you see the AI interviewer going? Or is it just matching capabilities to a job description?
[00:10:25] Speaker B: We see that this is a platform which is going to transform every HR process. You talked about mentorship, you talked about coaching, you talked about performance management, things like employee good.
[00:10:35] Speaker A: So that's the same way I see it. So wouldn't the AI inter capability sort of be embedded into every chat bot that we have in business?
If the company wants to do that and the employees up for it.
[00:10:50] Speaker B: Correct. That is how it will be. Now, what's interesting is if you think of this problem of AI interviewing. So, Josh, my apologies in advance if I. It comes across negative, but I will make a case over here. Right, sure. There are many people who can do podcast interviews. They get the face of it. It is quite simple, right? You do a Microsoft Teams meeting or zoom meeting. You have a set of questions that you ask of that person and you get the response.
[00:11:15] Speaker A: It's not as easy as it sounds, but it does. There's an awful lot of people doing it, that's for sure.
[00:11:22] Speaker B: But Exactly. Exactly, Josh. Right. That is at the heart of what you just said. That mechanic is actually easy. But that is not what podcast is.
Podcast is about. Or interviewing is about asking the right questions in the right way, engaging the person, getting the right essence out of them.
[00:11:41] Speaker A: You know, I Ashutosh, I've almost always agreed with you on almost everything. I mean, I think the problem with the word interviewer is it sounds very mechanical, like I'm going to give it a script and that's all it's going to do. It's almost an interviewer, an assessor, a listener, coach, a psychologist. I mean, it's kind of all those things. Is that fair?
[00:11:59] Speaker B: Yes, that is correct. And that is why the way we are envisioning our product, yes, you can embed it anywhere. But the real technology, real innovation comes in is how it manages everything, being consistent, being unbiased, bringing the fairness over there. Right, Right.
[00:12:16] Speaker A: Well, and the point you brought up earlier about the fact that the problem with interviewing is you can only interview 10 people when there's 10,000 people applying, whereas the AI interviewer can interview everybody at the same time, you know, and suddenly identify the fact that of these 10 of these thousand people that applied for a job, there's 20 that are highly qualified, but there's also 10 people that are really exceptional for another position that we had open that we didn't even know we were looking for, et cetera, 200%.
[00:12:45] Speaker B: And the third perspective is, now, let's take from the candidate's perspective today, if I want to interview someone, I have to do it. On a weekday during office hours, take time off from my work. Now I can do it from the comfort of my home at 8pm on Friday, if that is what I want first and foremost. Then the second thing is let's say there are two companies. One says that you know what you can interview with AI and you will know things within a couple of hours. Other is like let's connect next week to schedule an interview that might happen three weeks from now because the product is busy in one case you have an offer in hand in 48 hours in the third case or 24 hours in the other case, you're waiting for two weeks to even first call to get set up. Right. Where are you going to go?
[00:13:28] Speaker A: Well, plus the inter. Or I mean, I know you guys have probably done this. The interviewer could even come back to the candidate and say, you know, you're a great fit for something a little bit different from what you applied for. Can I talk to you about that before we proceed, et cetera. Well, I agree with you. So that's a massive opportunity. So the point you're basically making for the people listening to the podcast is that all of the AI that Eightfold has built over the last decade can be applied in a much more in some sense scalable form through the interviewer to do some of the same amazing things it did before.
[00:13:59] Speaker B: And other part is the rest of the HR processes become less relevant. Right. You don't have to worry about scheduling interviews, which is a massive pain for enterprises. You don't have to chase hiring managers for interview feedback.
[00:14:12] Speaker A: Do you think your AI interviewer will be used for performance management by any chance? Have you guys thought kick that about it will be.
[00:14:19] Speaker B: It will, yeah.
[00:14:20] Speaker A: So let me ask you a question that's on a fairly big topic that you know a lot about.
I don't want to talk about the lawsuits, but there is a lot of fear about being interviewed by an AI being rejected for unknown reasons, where's the data going, et cetera. What do you think? And you've, you know, I think you have many, many published papers on the issues of bias and diversity, which continue to be a huge issue.
What do you think AI can do to counteract those fears? Can it explain what it's why it makes decisions? Can it be transparent with candidates as to why they didn't get a job? What is your thinking about all that?
[00:15:03] Speaker B: That's an excellent question, Josh. Right. And there are two parts to it today. The way we are working with enterprises, the ultimate decision decision is being made by the enterprise and humans and not AI. AI is a facilitator to get the right information.
But the decision whether to hire or not hire or do a next round or not is done by the human being.
What it has done is that now as a rector, Instead of spending 10 hours with the process, you can in 15 minutes you have all the information to make the call. So that is one part.
[00:15:34] Speaker A: So the prior, so it's not a completely blocking process.
[00:15:38] Speaker B: So it's like you do a take a SAT exam, right? You get this is your score, right? This is what you answered. Now college can decide whether to admit you or not on the human being side, right? So that is how one should think about it. Now what AI is able to do is bring consistency. It is able to take the bias out of the human out from the assistance we have always talked about, right?
Is as a candidate, right? You are human. That moment you're interviewing, what is the mood you are in and all kind of things, right? The flip side is on the interviewer, right? Interviewer walks in, they crack a joke with you, you like it, and now they're talking to you for an hour. They come in with a bad mood and they want to wrap it up in 10 minutes. And on the other side, as a candidate, you are struggling with all this. Now AI can bring consistency over there. AI can make sure that the right questions are being asked to everyone.
Right now what we are working on is making sure candidate can see what answers they give.
Okay?
[00:16:30] Speaker A: So yeah, working on explainability, transparency. I mean that would be a, that would be groundbreaking if that was possible. Slightly different kind of question.
You started out As a academic, PhD, worked in R and D, worked in some of the, you know, most renowned research labs in the world, started a company, grew it, started another company.
What have you learned about entrepreneurship? And by the way, you started another one we're going to talk about in a minute. I know you what. What should high tech, science, researchy people know about entrepreneurship and business? What would you like to share?
[00:17:07] Speaker B: Entrepreneurship is the best way, the most effective way to take your ideas, make them reality. It is the best way to. Instead of talking about research and hypothesis, you can go and focus on solving the real problems. Third is. It is also actually, I'll give you one perspective, right, Josh. Many times people ask me what is it that a startup make? Startup is able to succeed while big companies don't. And the key thing is startups see an opportunity that is about to unfold. The market is not there yet, but will happen like in our case in 2016, to your point, no one was thinking about AI. All the incumbents were like, yeah, this is whatever. Who cares that it's workday, Oracle, SAP, right? And we saw the writing on the wall. We are like, this is happening. This is about to happen. So let's just focus over here, right?
So entrepreneurship is the best way to make that happen. A big company will never make that. As a small company of five people, you go all in and you make it happen.
It's fun, it's hard, it's painful, it's hard.
[00:18:07] Speaker A: But it is definitely stimulating and enjoyable.
So along the lines of entrepreneurship, I guess Eightfold isn't enough for you. You decided to do another one at the same time. Vivint, which who we use, we are a customer of Vivint. We're very excited about it. We get a lot of value out of it. But tell everybody what this new thing is and why it's so exciting.
[00:18:29] Speaker B: Actually, as I started Women, I realized that my entire entrepreneurship and research journey has been only about two things. One is machine learning AI. And the second thing is about understanding people.
Whether it was for disambiguation, two big things, whether Google, whether Bloom reached now eight foot, right, Whether it was a skills, understanding talent, understanding kms. Anyway, so Vivian is really nothing else. But can we help each person have their digital counterpart? Can we help AI understand what I have done, why I have done it, knows everything about me, knows everything that I know and can dramatically increase my efficiency and effectiveness. That is what it is about.
Now I will give you one example of where we are headed as part of this. So we are constantly making it better and advancing more. Right?
And this is more of a consumer example, but still it's relevant, right? Is that let's say there's a digital twin of a college counselor and they have these students who have signed up with this college counselor. They get to meet the college counselor one hour a month. If they can afford it more, maybe one hour twice a month, that's it. But at 10pm on Wednesday, as they are panicking, how should they apply? What should they do? Whether they should go on certain path or not, they have no idea who to talk to. If that college counselor twin is available at that moment, the value addition, the comfort is phenomenal for that student. And the flip side is the value now college counselor can provide to that student has increased multifold.
[00:20:05] Speaker A: I think this digital twin idea is huge. I think that's a great example. The way we use it is more pragmatic we have, we have, we've set up vivint in our company. Thank you very much, Ashu, for helping us.
And we can find information or ask a person a question when that person is either asleep or on a plane and get the exact answer as if they were right there in the room. So from a business standpoint, it's very easy to understand the value of a digital twin for a customer service agent or a technical specialist or a salesperson or anything else.
Where do you see that technology going into the future?
[00:20:44] Speaker B: So where we are getting massive interest and working on is companies have a workforce that is about to retire.
They have these experts which are the bottleneck and they can't scale. They have these remote offices and the communication between them is breaking down. Someone leaves. As people leave the organization or they move to a different role within the organization, all the knowledge and expertise walks away with them. So by building a digital twin of each and every employee, each and every expert, we are able to fundamentally solve these things. I'm talking to a hundred thousand person company and they are freaking out because one third of their workforce is about to retire. So they're like, if you can build a digital twin of those people, it will be just transformative for them.
[00:21:26] Speaker A: It's massive.
[00:21:27] Speaker B: There is a company, they are struggling because their actuaries are constantly moving in and out of the company.
But the risk appetite, how they have evaluated all those things, that knowledge, expertise walks away every time and actually moves out.
[00:21:40] Speaker A: So you really never retire. You live forever.
What are companies, by the way? One of the questions that comes up when you start thinking about this is if an employee quits and goes to work somewhere else and their digital twin stays, does the employee have the right to come back and say, hey, I don't want to leave all my IP with you.
Have you had that conversation with any companies yet?
[00:22:01] Speaker B: So the way we have designed it, what IT characters today, is the it that is owned by the company.
And as an employee, you have the right to delete the things that you feel are very personal.
[00:22:12] Speaker A: So you could decide what's personal and what's company owned. Yeah, well, I think, I think, I mean, I think that business is as big or bigger than Eightfold, but we'll see how that goes.
[00:22:22] Speaker B: I think that's what I'm excited about.
[00:22:24] Speaker A: Well, one more quick thing. I know we've been going for almost half an hour, so most of the people listen to this podcast are either entrepreneurs or HR people, and some of them are tech executives. What's your advice to Someone who's a creative entrepreneur. Scientists, software people who really want to go after this space in some unique way. What. What would you tell them to do to get their ideas off the ground and into the market?
[00:22:50] Speaker B: First and foremost, this is a phenomenal space.
That impact you can have by helping someone with a better career is so transformative for the society. It's just unimaginable. So it's a phenomenal space to be in. And Josh, even you, the kind of impact you have had in this, on this community is massive. Massive. The second one is stay close to the crowd, talk to the customers, talk to the prospects, spend time with them, try to really hone down in understanding what the problems they are facing.
But the third and most important, I would say, is also understand the technology. The world we are in. Technology is evolving so fast. You really need to be able to understand where this technology is headed. Otherwise, whatever you build, by the time you build it, it will be updated.
[00:23:37] Speaker A: That's true. Well, I would agree with you, and I think you guys have done this exceptionally well. Is staying very, very close to the customer needs. And at least in hr, it sounds like a big market, but it's really a whole bunch of teeny, tiny markets that look big, but they're all different. Ashutosh, thank you for all your partnership over the years and for leading the industry into AI, into the world of AI. Frankly, I. And on behalf of all of the other people in this space, I want to thank you for that and thank you for giving us a little bit of insights into what you've been up to and your experience in Eightfold and everything you're doing.
[00:24:11] Speaker B: Thank you, Josh. Thank you. Thank you for the support, thank you for the help. And frankly, you've been a phenomenal mentor. So I just can't thank you enough.
[00:24:18] Speaker A: Bet. Thank you.
