A Global Tech Podcast: Sachit Kamat on Agentic Recruiting and AI at Scale

So, let us bring in our next guest today, Sachit Kamat, the chief product officer at 8.

How are you doing today? I am doing super duper. You can hear me?

Okay.

You are perfectly clear. Thank you again for coming on and doing this. Are you standing by the way?

Not. No, I’m I’m sitting down. I know it may look like it the thing is a little high, but No, that’s okay. It’s okay. I’m just curious. I I don’t mind standing desks

at all. So, look, we thank you for coming and doing this. We really appreciate you spending the time with us. We wanted to talk to you about how responsible AI powered talent

intelligence are just going to help companies hire better, retain stronger employees, and build more adaptive workforces, which is becoming more and

more important. Before we get into the details, why don’t you tell our audience a little bit about the work that 8-fold does?

Yeah, 8fold is a global uh company that uh operates in the uh AI for HR tech space, right? And we support some of the

world’s largest companies in how they hire, retain uh and effectively utilize the talent across all the strategic

initiatives that these companies are undertaking. And these, you know, our customers represent tens of trillions of dollars of market value. Uh and we are

over operating in over 150 countries today.

Oh my god. Where was the company founded?

In Silicon Valley. So we’re a Silicon Valley company headquartered uh here.

But uh again we operate you know across the globe.

I love it. 150 countries is basically every country where it’s necessary to operate.

Yeah. Yeah. And some of our uh you know customers operate in almost every one of those geographies.

Sure. Makes perfect sense. How is AI changing the way companies approach recruiting?

Um it’s kind of a broad question but especially when you know these hiring decisions just are so important not just to the companies themselves but to the

employees that are actually getting hired what what role is AI playing here?

Yeah. So it’s important to just understand like how AI has evolved right in recruiting. So in the I would say early days which you know basically

takes us back uh maybe a little over five to 10 years ago. uh the best way that AI was being used is to go through

lots of data on candidates and then effectively identify you know the right uh sort of folks for any given role at

the company right so going from tens of millions of people with some of these companies have the data sets that you know that scale to that that number down

to the 50 or 100 people that matter for any given role based on their backgrounds and their unique qualifications right and so it was more

of a way for humans to do something that was not human scalable before, right? So identify the right people and then

effectively pursue them using the right tools uh that are available to them.

What has happened now in the last couple of years is that that transformed from just the intelligence that powers the

human workflows towards actually orchestrating some of these previously human workflows itself. And so now we have agents that can go much much deeper

in the recruiting uh workflows. They can actually interact with uh people like you and me. They can have conversations

in real time. And you are in a world now where things that were not human scalable two years ago are completely human scalable now. Right. Where

a recruiter can potentially consider hundreds, thousands of people for any given position. Uh versus Yeah.

How does it change the way or the perception of a company when the employees first interaction is

with an agent? meaning it meaning an AI agentic AI if that makes sense.

Yeah, it’s a that’s a great question. So what I would say is that it depends on the industry that you operate in. Right?

So if you are a future forward company, if you are someone that operates in technology for example, it’s not an odd interaction at all. Right? Because you

know many technologists are used to interacting with AI on an everyday basis. Sure.

Uh that being said, yes, there is a little bit of human change management that comes into play depending on the types of personas that you’re trying to

hire. Right. So we have one example of a company that is uh effectively operating uh you know effectively in the in the transportation

vertical and they’re hiring a bunch of folks that are doing uh you know physical labor tasks and there is

definitely a little bit of uh training associated with what do you expect in an AI interview that they’re you know customer that and candidate is going to

have to go through in that experience and the product is designed in that fashion right so 100% it depends on the type of person you’re trying to hire and

that interaction will determine how they see the recruiting process.

And how does it change the way people use their soft skills? I mean, maybe that’s the wrong question, but you know, humans make I think I always say like

humans make visceral connections with other humans. When I meet somebody, I can feel like am I going to get along with this person? And by by almost by

proxy, am I going to get along with the culture that sits inside that company?

Is there a way to make the initial agentic um interaction still feel like it’s

human in a way so that I can see if the culture is appropriate for me as a potential hireer for that for that firm?

What I would say is that culture is always a double-edged sword, right?

Because you know hidden within that is also human bias which comes into play.

Yep. And so the agents that uh you know I would say many companies like us have built are more focused on assessing the

hard skills versus the soft skills right so the the things that you can actually measure in a quantifiable way and then the idea here is that we’re

simply protecting the human on the other side in the you know basically the the hiring team from having to go through and talk to people that effectively

don’t meet the hard skills right I think in the future we will continue to evolve how these agents are able to assess soft skills as well and uh you know things

will get better on that dimension but I would say primarily right now it’s more of a mechanism where uh you are able to

build more deeper human connections by talking to the right people that have the hard skills in the first place and are we suggesting that these

interactions these agentic interactions are driven by avatar-l like experiences or is it more just an audio thing rather than an audiovisisual thing as well

it’s actually a video interaction and uh there are two alternatives that are available within our platform in some cases it is a human persona with a human

face with expressions uh you know that where the mouth is moving obviously as as the as the AI is actually speaking or

listening etc right so uh that is available but it’s also uh we do have what we call the bubble right which is

it takes away the human aspect of it and it’s more of a situation where it’s quite clearly AI that is speaking to the human on the other side

so it’s a It’s a video conversation, but at the same time, the client gets to pick what is the right thing for their business because yeah, in some cases,

you want that human type interaction. In other case, you might not.

What type of and you you’re just triggering stuff in my head like I just can’t stop thinking about the way this is going to work because I haven’t been through the experience myself. So,

forgive me for that, but what c can I give you an example of something if you don’t mind? Absolutely.

When I was in college, and this is going to date myself, but I don’t really care.

I did I I was going to school in Connecticut and I did this job that was um you know if you wanted to know somebody’s phone number I think you

would dial 411 and you it was called information. So you’d give somebody’s last name. I would type it into a computer and I would interact with that

person and give them stuff. But my manager would monitor the call sometimes, you know, to see if I was mucking around

or to see if I was actually giving out good information or if I was just good at my job interaction.

Is there monitoring and governance of these agents that are doing some of these initial interactions with potential hires? And are those is that

monitoring governance being done by other agents or is it being done by humans or is it some hybrid version of that?

Yeah, short answer is it’s a hybrid version of that. Right. So what you’re referring to is the guardrails that the agent is able to operate within. Right.

Which is what’s the context of information that it has to answer questions that might be asked by the candidate. What are the uh situations where it’ll decline to

answer? Right? So, for example, if a candidate asks the question, am I going to be hired based on my responses? Right?

Clearly, that’s an example of a situation where the guardrails need to come into effect.

So, the the interesting thing about the agents is that you can scale the uh the guardrails, right? Which means that every interaction uh you only have you

only have to put the guardrail in once and it is effectively across the entire set of conversations that the agents can simultaneously have. And these agents

are obviously infinitely scalable, right? So tens of thousands of interactions can happen simultaneously.

What is the output? God, I have so many questions. This is so interesting to me.

I hope you don’t mind. But what is the output of that interaction? So I’m I’m not the agent, but I’m a potential hireer. I’m interacting with the agent. I presume the whole thing is recorded.

It’s transcripted and stuff like that.

But at the end of the day, the hiring manager is going to want to see some kind of report about some kind of information about that person when it gets to them, right? Or whoever the

first human interaction is. what is the output that gets created? How is it curated? Do you know what I mean? All that kind of stuff. What does that look like?

So, the beauty of AI is that it can take a large volume of data and summarize it in a very uh easy to consume way. Right?

So, there’s a summarization of the conversation.

Uh there is the uh identification of where uh skills based evidence was deter was demonstrated within the

conversation. the ability to hyperlink into a specific part of the conversation and actually see it as a human and use human judgment as well.

And you know absolutely there are uh the criteria that the that the corporation has laid out in terms of what they care about demonstration of those specific

areas is something that the report makes very very easy right in terms of what is collected by the agent. Of course, the entire conversation is recorded. The

entire conversation is transcribed and that is obviously available to every every recruiter as they’re making a determination on who to take for. Right?

So, again, I’m just going to do a little bit of math in my head and I’m going to get some of this wrong, but just bear with me. If a recruiter can do seven interviews a day, that’s 35 a week.

Yep.

But the agents can literally do thousands of these things a week. And for large companies, the level of efficiency that an agent that an agentic AI can actually have is just like it’s

unfathomable at some level at some level.

But in that context, you may have like a hundred potential hires that are really good. I presume

the AI before it gets to like the next step also be after the transcripting all you know after watching the videos and stuff like that also does some sort of

comparison based on some criteria that this specific role has which makes the JD I would presume much more relevant in

this case and because it’s an agent interacting there’s probably less bias although some bias I I would argue but less bias in

there but what level of comparison is actually taking place and then what output is created from that if you know what I Yeah. So it depends on the criteria that

the uh you know that the hiring manager and the recruiter actually care about because that is configurable within the platform.

But again keep in mind that this criteria is all objective and skills based right in our in our solution. Yep.

So again I going back to your initial question right it’s not using any kind of human traits. Uh it’s not it’s not it’s not determining confidence and the

other things that we might assess in an interview. It is looking at the hard skill sets and whether those skill sets have been demonstrated at the right level. And then yes, it is comparing

various candidates in terms of who stood out the best in terms of uh the experiences that they were able to demonstrate the the stories that they

were able to tell related to whether or not they have a particular skill, right?

And that comparison can be made on the entire interview. It can be made on a specific skill. That’s how you would then see determine which ones you want

to dig into further and then uh validate the evidence and then decide who to take forward in the process.

What what kind of training is necessary for HR teams inside of these organizations that haven’t had access to

these tools prior? Like what do they need to know? What are they looking for when a lot of the sort of pre- stuff

that they might have done themselves before or or some of their junior staff had done before is now being automated.

So now they get all these reports that didn’t exist prior like what do they have to learn? What do they have to be taught to the so that they can get their minds around like okay this is new

information maybe even better information and now I can make better hiring decisions but I have to know systemically how to actually use this stuff. What we find in almost every

situation is that the uh difficult part is not the actual technology. It’s actually the change management associated with uh changing the way

people operate. Right. Which is anytime a new technology comes about, you have to change your own behavior to actually benefit from it. Exactly.

And that change management can take a bit of time, right? Uh we’ve designed these systems to be like fairly easy to use um and fairly easy to, you know,

basically orchestrate, right? These agents, etc. That’s not the challenge.

It’s more sort of how do you make sure that uh in in any situation where you’re changing the entire flow of the process,

how do you how do you identify the humanpowered bottlenecks that exist in that flow and how do you address those?

Right? Like the fact that the agent is able to come back with screenings for 100 candidates means that the recruiter is now going to have to be very very

good at assessing uh which of those 100 to move forward based on what the agent is saying. There’s still a bottleneck in that process that needs to be addressed.

So th those are the kinds of training that is necessary to benefit from the technology and to drive towards a world where hiring can happen faster with higher quality.

So I hadn’t thought about this before but the more I read about FDEES the more I think that it’s a really great business model. I’m curious if you as a

firm employ this in the sense that like you said the change management is almost important is more important than like the technology. I agree with you. If the

technology is built in the right way, it should be sort of seamless for people to use, but it’s like how do you use it in the in a way that’s most effective? What

what I found with the FDE stuff again in the context of I think about this a lot in the context of anthropic and open and you’ve seen the news, so I don’t need to

explain that to you. But the point is, is there some sense of you guys doing that as well? One, so you can train the people properly internally how to use

all these systems and to understand what the output means. But second, so you can get feedback back into your own company so you can understand, oh, we should

either redesign this or we can actually add some kind of functionality because that’s what the customers that we’re working with actually want.

So what I would say is that the entire FD uh you know sort of how software companies are evolving is changing dramatically. Right? our world what

we’re seeing is given the concentration of customers we have is heavily fortune 500 and I would say even high fortune 50

right it’s a situation where every single company has a use case that is very very unique to them

and what we are now seeing is is this world where software is entering the mass customization moment right where every one of these companies

can effectively build solutions on part on on top of software platforms like us in a way that was not possible before,

right? So, the ability to use uh wipe coding agents, the ability to describe problems in a unique way and then benefit from the software in

infrastructure that we provide, that’s a reality right now and that’s a solution that we recently announced which we call talent forge which allows customers to build on top of our platform.

That’s what I I’m so glad you mentioned that cuz I was going to ask you about the custom customization or is it customizable as well? And the example I was going to give inside of an insurance

company like hiring an actuary is not the same as hiring an insurance salesman even though they’re both working in the insurance industry. A and and and to be

fair then hiring somebody to work at some other firm that has nothing to do with insurance is very different than it is for an insurance company. So it needs to be customizable at some level. Are

you suggesting as well if people can build on top of it that all of the functionality that sits inside your core systems are API callable as well?

Yes, we’ve had APIs for a long period of time. We’ve had an app marketplace for a long long period of time. But I think what the difference is right now is that

we’ve re-engineered our API so that agents can talk to them. That’s what I want to know. Yeah.

Yeah. And so that’s a direction that we have gone in very very recently and also built out protocols like MCP which allow agents to communicate with other agents.

Yep.

And it’s a world where obviously you know we are building for a different user right which is the agent versus the human.

How has it changed the thought process inside of your own company?

This because look again one of the things we talk about is Salesforce massive company big you know big SAS company but what they did was they completely rewrote all their software

Salesforce 360 all of the stuff is API callable all of it’s MCP compliant but now it changes the way that the interaction takes place with their

software as well. in the same way that you’re suggesting in the sense that a lot of the new interactivity that’s going to happen with Salesforce 360 is going to be agentic as opposed to human

that’s going to grow rapidly. What type of mindset change inside of 8-fold had to happen so that that was actually possible when you go to people and say

we need to make all this stuff again MCP compliant but also agentic callable. How does that change the way people think about developing the platform? In a lot

of ways as a software company you have to be prepared to disrupt yourself right and that is the strategic conversations that were

being had within our company right in terms of does the UI is should we be uh set you know should the UI be set in

stone or should we make it completely malleable to a new use case how do we how do we reimagine where the value comes from right is it the

workflows that our customers have configured or is it a world where uh effectively they can build their own capabilities using the platform platform

uh that we have built, right? So those are the strategic conversations that need to play out and then it’s also a question of how do we take it to market?

How do we monetize? All of these things have to change in the world that we’re in now.

It’s never been a more exciting time, I think, to be building these software companies. Um can I ask some more stuff if you don’t mind?

Oh yeah, absolutely. How should companies measure whether AIdriven recruiting is actually giving them better candidates as opposed to just

making like going through that funnel faster like how do they do that sort of measurement?

So you know we as an AI company that’s been in the space had built a measure which we call our MAT score right which is a measure of quality of the

individual that folks are hiring right which is the fit of the individual relative to the the role that they’re being hired for. That’s a measure that

we use in practice, right? To measure quality. But what we have been able to demonstrate is people that have higher math scores also stay longer within the

role, right? There’s the retention of those people tends to be higher, which is a proxy for uh you know, effectively did were they a good fit for the role or

not, right? Um those are the kinds of measures that we bring to the table and I I would say like at the end of the day, recruiting at at you know can be

measured in terms of time to hire as well as you know effectively the quality of hire. quality higher to us is that quantitative measurement that I just talked about.

Yeah, the retention thing is actually a really big deal. Does are we seeing changes in the numbers related to

retention based on the usage of Agentic AI in the hiring space in the sense that you should have way more information the

match score should be higher because again hiring someone that quits in two days after they get hired is just a complete waste of time and money and a

drag on the economy. Are we seeing that those things those metrics are moving in the right direction or are they kind of static?

Where we are seeing uh immediate benefits is that the time that it takes for the first interview and then obviously as a result the time to hire

has gone down dramatically right as a result of some of the technology. I’d say the quality metric of hire is also improving but like the retention is

still something that’s going to take a while for it to mature and we are absolutely tracking that right to determine whether the the core uh you

know matching product that we have built versus the agentic product what is the difference between them we’re still running that experiment as we speak

and as this progresses are there things that institutions or enterprises can

learn from this process and all the data that gathering that over time is just going to make their hiring um

capabilities more efficient but also more effective and m make their retention numbers go down because of the way that they can

gather data. Now, I’m just going to say like in the old days, you know, you’d have a recruiter with a pen and a pad of paper just like writing stuff down or

checking stuff off. At some point that was digitized, but like that type of digital transformation frankly was not that effective. This is real transformation in the sense that you’re

getting all this sort of granular data about individual candidates and you can have way more candidates as well. So way more data. Is there a way then to use

that data later to say to make the hiring process like as it’s looping around to make the hiring process just way more effective because now you know

the types of people and candidates that you want to hire in a way that was better than you could know before. So let me give you some real examples.

Right. Once you start to record and transcribe an interview conversation.

Yep. that is gold when it comes to uh you know solving some real life use cases. So think about like providing feedback to the interviewer themselves

right in terms of uh the coverage of the interview. Give them feedback on the quality of the interview, the quality of the answers that they were able to extract. Give them feedback on what

things they might change and then to be able to bubble that back up into uh effectively redesigning the process itself. Right? Are

are you covering the right things in the interview and should you redesign that?

Those are things you can answer once you start to digitize and use the insights that come out of the AI.

Okay. The last thing and I mean I have so many other things I want to talk to you about but I’ll let you go.

What are there organizational changes that need to get made inside of HR departments at scale because of this new

way that they’re using your platform to do their recruiting?

Yeah. So I mean it it comes down to this concept where every one of us is going to be have to be trained on how to orchestrate and manage agents in our lives, right? We’re doing it every day.

I’m doing it as the as a product leader at the moment.

Recruiters are no different, right? In this in the sense that they’re also going to have to change the way they operate because the operational tasks are no longer the

core of the job. It’s more the human intelligence building uh connections with the candidates you know convincing someone to leave a partic leave their

existing job and come over to the company agents are not better than humans at doing that right we as humans are better at selling

humans on you know following through on on a particular path right and I would say like we are going to get have to get a lot more comfortable and have a lot

more trust in you know orchestrating these agents running things at scale evaluating the work they do and constantly managing them, right? Like

that is a world that we’re going to head into and recruiters are going to have to learn to do that as well.

Perfect. Okay. Thank you so much for coming and doing this session. I really appreciate your time. Hopefully you enjoyed this as much as as we did and

frankly I learned a lot today as well and that’s actually more important than anything else. Thank you again for coming to do this. I really appreciate it.

Thanks for having me. Awesome. Thank you.

Bye for now. Yeah. Very nice to meet you as well. Okay, that is it today for AGTP Daily.