Compliance Is the Quiet Feature That Sets AI Hiring Tools Apart

Somen Mondal:

So imagine a candidate came to your company and they’re like, “Delete all my data. I don’t want you in my system.” A typical company, now, there are some companies that are more kind of advanced, typical company will have no idea what to do. You know what they’re gonna do? They’re gonna go to legal. They’re gonna be like, “Hey, I got this email. What do I do? How do I delete this?”

That’s actually one of the workflows that we’ve taken care of. That is actually a great example of a compliance angle that many people don’t think about, but you need to have.

Steve Smith:

Hey everyone, welcome back to another episode of Work Tech Weekly. I'm Steve Smith, Managing Director of Growth at Rep Cap.

Ten years ago, there were no rules for an algorithm that screened job candidates. You could build whatever you wanted. That world is gone. AI in hiring now runs into regulations at the state and country level. Plenty of vendors are still building like nobody is checking.

My guest is Somen Mondal, Co-Founder and CEO of Amp. Amp uses AI agents to get frontline hires from an open role to their first shift, working across the systems a company already has.

Somen has been here before. This is his fourth startup with the same co-founder. He was running bias tests on his own hiring algorithms back in 2014, years before lawsuits made the topic hard to ignore.

In this conversation, we talk about why compliance is becoming a competitive advantage for AI vendors who can prove it. That includes what happens when a candidate asks a company to delete their data and no one knows who handles it. We also get into the idea that the future of talent acquisition is about capacity instead of headcount, and why domain expertise matters more as software gets easier to build.

If AI is part of how you hire, this is a good look at what compliance has to cover now.

Let's get into it.

Somen, welcome to the podcast.

Somen Mondal:

Steve, thanks so much for having me. Really appreciate it.

Steve Smith:

It is great to have you here. You know, we were talking a little bit earlier before we started recording and I was telling you, it’s just like I really love talking to innovative early stage startups. And you’re someone, you’ve got a lot of experience in the startup space.

This is now what, your fourth startup? You’ve had three exits behind you. You know, you’ve done this a few times now with the same co-founder. You know, what do you think that is your top takeaway over all of those exits that you bring into this new one that is sort of like guiding you and kind of front and center in your mind as you’re thinking about, okay, here’s what I need to make sure I’m doing right this time?

Somen Mondal:

You know, we started our first business when SaaS was just beginning, to be honest. And you know, with the new technology, there’s kind of a driving force or a theme throughout what I’m saying, is the new world of AI truly is a differentiator. And so I would say if there’s one takeaway, you know, not just for me, but for anyone starting a business now, it’s really having domain expertise.

As technology becomes easier to develop and quicker, having actual domain expertise is super important. So, you know, if that answers question, that’s what I think is super important right now. And you know, what I think is gonna help...

Steve Smith:

One thing I’m wondering is, you know, you’ve been through, you know, kind of the SaaS market cycle and now we’re really kind of entering this AI market cycle, you know, and certainly there’s a lot of market chatter about SaaS apocalypse and all of this. You know, whether that’s overblown or not, you know, something is different this time.

And as you’re thinking about how you are adapting your new company to kind of this current moment in the market, what do you think is applicable from, you know, your time building SaaS companies, and how are you maybe kind of shifting some of your thinking as you’re looking to lead a brand new startup into kind of this moment in the market?

Somen Mondal:

I think what’s applicable for any business, you know, whether it’s software or, you know, whatever type of business, is really to show ROI in what you’re doing. You know, ultimately if you aren’t able to show, like, quantifiable value in what you’re doing, are we able to, you know, save you time, make you more fair, make you more compliant, make you more accurate, more efficient? There’s a whole bunch of different factors. You know, it doesn’t matter what business.

So, you know, the thing I’ve learned across three different businesses as well as working at Dayforce for almost five years is, that’s something that never goes outta style. How can we show ROI for the customer that is undeniable?

So that is something that has stuck with me. And you know, like 20 years ago, maybe that’s something I wouldn’t have thought of right away. It’d be like, oh, I wanna sell them features. But being able to show value is something that I’ve taken through all of the companies, and it’s literally the most important thing that we focus on above anything else with our customers, is how do we show them quantifiable value?

And so that’s something that I think is a lesson that we’ve learned and throughout all the businesses.

Steve Smith:

So, you know, FieldID, Ideal, FairNow, you know, three very different products, but kind of going after that, you know, HR and ops buyer that’s trying to navigate compliance and risk. Is there any other through line that you kinda keep seeing as you’re navigating this market?

Somen Mondal:

You know, we’ve always been fortunate to kind of land on a new technology phase, and we’ve seen it throughout. You know, in the beginning it was SaaS and mobile, so our FieldID was SaaS and mobile. Nothing to do with HR tech, but our biggest problem at that company was hiring salespeople. That’s actually why we got into HR tech, is we just naively thought, well, we’re just gonna build software to help us hire better.

And we didn’t even know what an ATS was back then. And so that’s how we started Ideal. So we went from mobile SaaS, that was the big thing back then, and you’ll see like an adoption curve from there. And then we went to Ideal, which is what I call like AI 1.0, that’s like the recommendation engine, screening engines.

And so that, you know, grew its adoption phase and you saw a bunch of companies grow and be acquired, and then all of those functions are being integrated into the larger platforms. And now the trend is, you know, what I like to think is like AI 2.0. It’s ChatGPT, it’s Claude, it is intelligence at a scale that just wasn’t possible before.

And so, you know, these are the trends that I’m seeing and why I think this time is a little bit different. But again, you know, to bring it back to my first answer, regardless of the technology, the underlying business driver is, you know, how are we doing things better? Are we doing things more accurate? Are we showing ROI? Can we quantify that ROI? Gonna keep driving on that, ’cause I think that’s really important. A lot of times companies don’t focus on that, but we really do.

So, I really think this time’s a little bit different because I’ve never seen technology move this fast before. So, although there’s been fundamental changes in technology, this time it’s not only very different because you’re able to do intelligent things, and it’s also moving extremely quickly. So those two things combined make this a little bit different than before.

Steve Smith:

And I’m glad you brought up those two points because, you know, and first, thank you for validating that things are moving super fast, because it certainly feels that way to me. And I don’t know, just because I’m getting older and this is an unprecedented market cycle. But I think that, you know, the other thing is kinda like being able to like work with intelligence in the market.

And I think that, you know, you’ve spent a lot of time in and around compliance and compliance-driven industries, mitigating risk in organizations. When you’re looking at kind of the market broadly in tech, I mean, even beyond work tech, do you feel like perhaps some of these AI companies are being a little bit too cavalier with compliance needs and underestimating what’s really required to build a good compliance-driven product?

Somen Mondal:

Absolutely, and we’re actually leaning in on compliance as a differentiator. Ultimately, you know, the way that adoption curves work is that there’s gonna be a bunch of early adopters and there’s gonna be a lot of competition and a lot of startups, and we’ve seen this. If you think about eight years ago in screening and matching, there was a bunch of companies that came in. And then a bunch of companies left, and a few stuck around.

And I truly think where we are now, compliance is a major differentiator for companies that can do it properly. And you know, when you think about AI, you know, you have the traditional, you know, payroll compliance, that’s actually quite comparable. Payroll is done by state, done by country. It’s a very complicated... AI is becoming just like that. There are, you know, state level, continent level, country level AI regulations, AI governance.

So for us, you know, it isn’t the Wild West anymore. 10 years ago it was. When you’re building AI algorithms to, like, say, for example, screen a candidate, there weren’t any rules, you know, eight to 10 years ago. You can kind of do whatever you wanted. That world doesn’t exist anymore.

There are new regulations, new governance, and just like, think of it as a, you know, SOC 2 compliance, ISO 27001, there’s compliance that you gotta do just as a software business. You gotta do those things. I do think in terms of AI, you need to show that you’re compliant. You need to show what you’re doing can be auditable.

And so, I do think not as many companies lean on that because it’s an uncomfortable area that introduces risk to the company. But I do think it’s a competitive differentiator for the companies that can do it right.

Steve Smith:

With all that said, we talked a little bit about your background. Tell us about what you’re doing now at Amp. You know, you’re a very new startup and there’s probably a lot of folks out there who haven’t heard of you, dunno what you do. Tell us, you know, what are you doing and what’s different about your approach?

Somen Mondal:

For sure. You know, Amp is truly the culmination of the three businesses that we did in the past. So if you think about what Amp is, simply, the job of Amp is to get someone to their first shift, specifically in frontline hiring. If you think of Amp, it is the system that brings someone to their first shift in the most efficient, cost-effective, accurate, so that’s where the intelligence comes in, and fair way to do that.

So that is kind of our North Star. How do we get someone to their first shift, crossing different systems? You have applicant tracking, payroll, workforce management. It’s not just one system, so it’s cross-system. So when you think about Amp, think of it as getting someone to their first shift and all the steps that are involved.

And if you think about all of the steps, there are literally a million steps in between that. Even if you use software, there’s a bunch of clicking around, usually two to three different systems, to get to that final outcome. And then there typically isn’t learning across that. So that’s, you know, that’s what we do for frontline hiring, get people to the first shift using the systems that they already have.

And so if you actually combine all of those things, Ideal, FairNow, AI governance, and then you layer on the latest and greatest technology, you now are at a point where you can get someone, literally by talking to someone in Teams or Slack, talking to Amp and saying, “Hey, we just lost our warehouse worker or line chef. We need to find someone.”

Amp will create a plan to get that person to their first shift, keeping a human in the loop. It’s not fully automated, ’cause we don’t believe in that, but there is a human in the loop in that process. So again, all that experience is to get someone from “I need to hire” to their first shift using agents. So that’s what we do.

Steve Smith:

So another question I wanted to ask you is, you know, I think one of the advantages that you bring is your perspective, the fact that you have done this a few times before and you found yourself at the forefront of some of these industry changes before. Like, you were doing bias testing on your own algorithms in 2014.

That’s suddenly become a very top-of-mind thing in talent acquisition because of lawsuits and market concerns. What did it actually look like to you at the time, and what were you maybe doing wrong?

Somen Mondal:

So back then, you know, to measure bias, it is the same thing you would have to do to a human, which people weren’t doing as well. Like, you know, typically people were not doing it. But to measure an algorithm’s bias versus a human’s bias, it’s kind of the same thing. So we would use the four-fifths rule back then to quantify.

And we would use technology to infer certain demographics, this is in 2014, to quantify if we were introducing adverse impact into that. And we would do a bunch of things to reduce bias. So, you know, getting rid of age, name, and sex in any kind of decision-making process or any kind of recommendation.

So we were doing this back in 2014, and we were actually going down the path almost of academia, where we were starting to work with universities and colleges around how to measure bias. And then we got acquired, and then, you know, we basically wouldn’t say we stopped doing it, but it wasn’t, you know, the most important thing that we were doing at that time.

And what’s funny now is we picked that back up. We are again, hopefully we’re gonna announce this soon, we’re working with a professor who specializes in this field, so we’re picking it up. The technology has changed. LLMs do different things. You have to measure things differently, but you’re absolutely right. We started this in 2014. It has now become more common to do this.

You are noticing there are lawsuits and whatnot. So what’s gonna happen is people are going to be afraid of using this technology, and until people have systems that are so transparent and back them up with audit trails and reports that people are starting to be more comfortable. You know, I think people need to be, if they’re using a system where it’s like the Wild West and there’s no controls and there’s no mention of compliance around AI, then yeah, you should be worried about it.

But if you lean on it and are confident in trusting the vendor and seeing what they do, I do think that’s an advantage. And you’re right. We have learned from that from the past, and we do think it’s something that we are heavily leaning on, I would say.

Steve Smith:

Well, and you know, it’s interesting because, you know, FairNow was all about kind of an early AI governance, you know, and so it’s interesting when you are looking at bias and governance. I mean, you were working on, you know, current market themes and topics that are really blowing up now, I mean, years before they were on anybody’s radar.

And when, you know, you look at kind of what you’re building with Amp, it’s just like you’ve got this kind of governance and observability first, you know, planner and connector second. You know, it’s just, you really have, I think, a thorough approach to, you know, the governance and the compliance side.

It seems like we’re in a phase of the market where there’s a lot of ready, fire, aim on the part of buyers and even, I think, some vendors. You know, what are the things you point to say, “Hey, no, we are taking a different approach right now”? When you are pointing to the product and having those conversations to prospects, what are the things that you’re bringing up?

Somen Mondal:

So there’s two types of compliance that, and again, we would not have done this 20 years ago, but again, we do have experience now, and we’re like, okay, first thing we’ve gotta do is take care of IT compliance. So that’s any software company has to do, like, that’s SOC 2, 20000, ISO 27001.

There’s a bunch of compliance that you need to do just from an IT infrastructure. Where do you host your data? How do you manage data residency? So we started day one doing that. But the other part that most companies probably don’t focus on is how do we show compliance to our customers? And so we started hiring people immediately that had expertise in AI compliance.

So, obviously we started FairNow, so we have advisors and investors. Actually, one of the founders of FairNow is now an investor in this business. So we have their expertise, one, but we hired people full-time that, at a small company, we’re currently under 10 people, that just focus on compliance. So I don’t think that’s regular.

Two, we gladly point people to what we call the compliance center in our system, which constantly runs bias tests, which constantly looks at new AI rules, which will treat our customers... Say you’re in different states, well, there might be different rules, so we show them what the different rules are, and the system can change itself, act differently based on where it’s located or where, you know, job postings are, or where they’re attracting candidates or where the candidates are gonna... And there’s a whole bunch of rules of where they live, where they work. And so we take care of all of that.

I think it’s the first time. And so we had a couple of, like, amazing practitioners, CHROs, who looked at our system. The compliance part, we’re most proud of this because, you know, demos, you can demo some amazing stuff. Like, you know, you can demo, especially now, right? So that’s amazing. You know, someone can recreate a lot of stuff and it can look good in a demo, but the part that typically they will not show is how do you maintain compliance?

And that’s the first area we’re, like, proud to show people, ’cause we started out with the compliance framework. And, like, try to poke holes at our compliance. Like, I don’t know who would be able to do that. Like, I’m sure someone could, but that’s something that we’ve led with.

And so if that answers your question. Yeah. Like, I think we’re the only company with a compliance center that tells you everything, keeps track of everything, so make suggestions as well, that I don’t think other companies, even the large platforms, don’t do yet around talent. Obviously, in payroll side it’s a different story. Workforce management, different story. But on the talent acquisition side using AI, there’s not as many companies doing that.

Steve Smith:

You know, when you think about kind of where we’re at in this moment, and you’re starting to see these stories begin to emerge. A candidate gets screened out, they believe that it’s AI, and they don’t know why, and maybe they want to know why. You know, who is accountable? I mean, is it the vendor? Is it the customer, is a little bit of both?

I mean, what’s sort of your perspective on that? Do you need to build this into the product to make it customer-proof? Or can customers still find a way to break it and make bad choices?

Somen Mondal:

That is a great question, and I’m glad you brought it up. So imagine a candidate came to your company and they’re like, “Delete all my data. I don’t want you in my system.” A typical company, now, there are some companies that are more kind of advanced, typical company will have no idea what to do.

You know what they’re gonna do? They’re gonna go to legal. They’re gonna be like, “Hey, I got this email. What do I do? How do I delete this?” That’s actually one of the workflows that we’ve taken care of. So you can just tell Amp, “Hey, I got this request,” or you can set up your email address so it’s, you know, whatever, privacy@yourcompany.com or talent@yourcompany.com.

It will actually process that request and delete all of the data, provide a response to the actual employee, or canned response that you can send them. It’ll give you an audit trail. So I think when we showed you the demo, it shows you what the agents are actually doing, and it actually records what it’s doing as well from a compliance standpoint.

So this is a common thing. People don’t know how to answer that question. And then it, you know, either goes to outside counsel, inside counsel, depending on the size of the business. That specific use case we knew was happening with customers, and they were, you know, apprehensive. Oh, they didn’t know how to deal with it.

And so we take care of that, and we delete the records, we provide the audit trail, provide a candidate with the right response. That is actually a great example of a compliance angle that many people don’t think about, but you need to have.

Steve Smith:

And I remember seeing that in the demo, and that really was a very much a wow moment, because it does seem like that is completely being overlooked in the broader market today.

Somen Mondal:

And it’s very complicated. Like, the answer might be different. The retention rules might be different based on where you live, where the company is, what country you’re in. It is not trivial. It is not trivial.

Steve Smith:

Well, I want to ask you a question about, you know, a line out of your positioning that was, you know, as a marketer, it was very bold and provocative to me. The future of work is capacity, not headcount. Where did you come up with that, and how did you come to land on that as a copy point?

Somen Mondal:

You know, it is really from our own experience. You know, I would say FairNow, although we didn’t run that company, we saw firsthand. And anyone who uses LLMs, you mentioned that you used agents to help some of the research that you do. And so, for example, like, you can increase your capacity of research without hiring more people. That is relevant.

You know, I’m able to do a bunch of things in my personal life. A couple years ago when I started using LLMs, you know, I love the market, the stock markets. I love to do research. What I can do now as a single person, my capacity has increased 10x of what is possible I can do by myself on my own portfolio.

And so we’re just taking that very same concept and applying it to frontline work in talent acquisition. So when we talked to CHROs, and you know, the difference, you know, when we started this company, you know, the first people that we started speaking to were CEOs and CHROs. You know, and of course it’s the same story.

“Hey, we need to, you know, especially in frontline, we need to hire a lot of people, but I gotta maintain my budget. I wanna maintain my headcount.” That’s a dream scenario, right? I wanna hire 10 times as many people, but I don’t wanna hire more people in talent acquisition. So this is actually the only time when you can actually accomplish that.

Like, you can increase the capacity of each one of your people in talent acquisition. You can get them to do things that they want to do. You know, what talent acquisition person’s gonna put up their hand and be like, “Yeah, I like clicking around on different interfaces for 45% of my time”? No one. Well, they shouldn’t do that. Well, most people wouldn’t put their hand up.

So at the same time, you’re increasing someone’s capacity, improving, hopefully, their satisfaction of what they’re doing. And this is the only time that you can do that. That’s where that kind of bold statement came from. It is using real-world experience, and I think anyone can appreciate, anyone that uses a large language model, one of them, can understand, yeah, this can really improve my capacity.

You know, the simple example is, you know, 10 years ago, I’d have to write job descriptions. That is difficult to do. You can do that a little bit easier now, right? So, you know, that one example, let’s say you used to take 10 minutes before, and, I would say, that’s being generous. You know, now takes, with a properly trained model, you know, a couple seconds. That’s more than a 10x increase. So, that’s where that came...

Steve Smith:

Well, you know, another thing kind of jumps out that I wanted to ask you about is, you know, your relationship with Shaun. You’ve been together now in four companies. You know, being in a startup environment, having a co-founder, I mean, that is a pressure cooker relationship that really, I mean, puts even the best of relationships to the test.

What is it about your relationship with Shaun and how you work together that you think has, you know, not only allowed you to thrive and succeed with that relationship, but want to go back to it again and again?

Somen Mondal:

You know, people always ask me that question, and it’s like, I always tell people, “Hey, it’s like getting married.” You know, you’re kind of, it’s a bit of a luck, like you’re gonna find the right partner and it works out, and there’s no formula. I would say the real thing that sticks out is we have complementary skills.

And so, you know, where Shaun is very good at the product, very good at customer success, I would say I lean more on sales and marketing on the opposite side. So, you know, the kind of the ongoing joke is, you know, I love getting the big customers and closing them and then passing them on to Shaun. And Shaun makes sure that they renew and are very happy and that the product is incredible.

And so, even if you look at our tenure at Dayforce, five years, 10,000-person company, you know, growing very well. You know, Shaun eventually reported to the CTO and then I eventually reported to the CRO. So, you know, we have those complementary skills, and we’re just aligned and we want to have a good time.

And so again, people are like, “Oh, why are you doing a fourth company?” We are truly having a good time, and we would not want to do anything else. And I think ultimately that’s the most important thing for us. At this point, the new technology’s incredible, and we’re having such a great time. That’s super important to us.

Steve Smith:

And that’s a unique balance because, you know, it seems like where a lot of startups, you know, miss the mark early on is they over-rotate on the product. They don’t think about, okay, how are we gonna take it to market and sell it?

And just like, if you’ve got, you know, a founder who is great with product, combined with a founder who is, you know, not necessarily non-technical, but understands, okay, here’s how we’re actually gonna sell this product. That really does seem like lightning in a bottle.

Somen Mondal:

Yeah, it really, and we both enjoyed those two different aspects of it. Now I am leaning a little bit more on the technical side more these days, ’cause it’s just so interesting. Like if you can do so much with it. And so it is kind of interesting from a go-to-market standpoint, like the stuff that we’re doing in the backend around our go-to-market system.

But yeah, I’m sure they don’t want me to get into the code, but it is, with anyone with a little bit of technical background, I am a computer engineer, like, what you can do now is so shocking that it is hard not to, like, dabble a little bit in it.

Steve Smith:

Well, and I guess with that in mind, and, you know, kind of going back to your comment about, like, you know, the use case around, like, how you are approaching using AI to make you more productive around, you know, kind of managing your investment portfolio, has vibe coding and that whole movement changed kind of the way that you approach technology and maybe, you know, prompted you to approach some things differently or think about things differently?

Somen Mondal:

And it’s, you know, sometimes people think, “Oh, I’m just gonna vibe code my payroll system.” And I’m like, that’s not gonna happen. And you know, an example I heard about, I forget where, but it was, you know, when people started inventing, like, woodworking tools like the lathe. You know, they’re like, “Oh, all of the people that make furniture are gonna go outta business because everyone’s gonna make their own furniture.” Well, no, you still need that skill to do that.

So the key takeaway in building this company is, again, domain expertise is, to me, very important. Like, understanding what a job req is or what a stage change is. And then having experience developing fast-forwards you a hundred times with the technical.

So that’s something that we’re doing a little bit differently. We are only looking at people that have domain expertise in almost all roles. Not every single one, but all technical roles. Almost every single person has domain expertise, so that just allows us to move, yeah, you know, I’ll say 10 times faster. I would like to think faster than that, and then Shaun will correct me and say, “Oh, it’s not that fast,” but it’s somewhere in between. So I think that is super important, you know, having that domain expertise.

Steve Smith:

Well, Somen, this has been a fantastic conversation. Before we wrap up, I wanted to give you a chance. Is there a question that I didn’t ask and I should have, something that you would like to add that you think would add additional color or context to the conversation we just had for our audience? Did I miss anything?

Somen Mondal:

No, I think the takeaway is, you know, it truly is a magical moment in technology, and I do think this is a fundamental shift in the way that people will work, at least, you know, in our area, which is talent acquisition. And it’s something that companies should try to embrace, especially if all the compliance pieces are there.

So that is, you know, the one takeaway. You know, whether it’s our system you’re experimenting with or someone else’s, or in a totally different field, it is magical. And, you know, the people that are using it at the level that you can, it truly provides value, in my opinion. And so, don’t, you know, don’t shy away from it. Try to use the system and the technology, ’cause it will change the way you do things.

Steve Smith:

Somen, thanks very much for your time. I’m excited to see where you go, and keep us updated with how your progress is going. Looking forward to continuing the conversation.

Somen Mondal:

Awesome. Thanks, Steve. Thank you so much for your time.

Steve Smith:

AI can now write a job description in seconds. That used to take a recruiter ten minutes or more. The speed part is settled, but proving the work holds up when someone asks how a decision got made is much harder.

Hiring is heading where payroll went years ago. The rules differ by state and by country, and so do the rules on how long you keep a candidate's data. A request to delete someone's records can land in an inbox with no clear owner. None of that shows up in a pitch about productivity.

That changes what a good AI hiring vendor looks like. Speed is easy to sell, but an audit trail is much harder to fake. The vendors who build compliance into the product will earn trust. The ones who plan to deal with it later are running on borrowed time.

The same logic applies to who builds these tools. Anyone can spin up software now. Knowing what a job requisition or a stage change means still takes years in the field. The tools got faster, but the expertise didn't get any cheaper.

So if you're evaluating AI for hiring right now, ask how the system shows its work before you ask how fast it runs.

If you enjoyed this episode, make sure to subscribe to Work Tech Weekly on Apple Podcasts, Spotify, or YouTube. And I'll see you next time.