How Phenom Is Using AI To Transform Hiring And Talent Intelligence

[00:00:03] What does it look like when HR stops just being a back office function and starts acting like a real control centre for talent, trust and business performance? Yeah, we're talking about value add here. And today's episode I think is quite a timely one, especially with the recent news that a company called Phenom has been on the move.

[00:00:25] Because they recently announced the acquisition of Be Applied, which is an AI driven cognitive assessment solution designed to validate candidates and employees capabilities and do so at scale. And this follows another acquisition of a company called Included, which is an AI native agentic people analytics platform.

[00:00:49] Something that was built to surface actionable insights so leaders can make faster, smarter workforce decisions. Two acquisitions in quick succession that usually signal something bigger, a company that's placing a bet on where work is heading next. So today I've invited the CEO and co-founder of Phenom back onto the podcast. I was very fortunate to meet him in person at Web Summit. I think it was late 2024.

[00:01:18] But since that conversation, a lot has changed in the world. AI has moved from experimentation to deployment. Regulation is evolving in real time. And HR teams are now dealing with new threats, threats that feel closer to cybersecurity than recruitment. Because we have everything from fake candidates, automated job applications, app volume, and even real world cases of bad actors entering organizations through hiring pipelines.

[00:01:47] All of these things are forcing leaders to rethink what trust means inside talent operations. So today we'll break down how responsible AI, what that looks like in practice when you're working with global enterprises, and why policies are varying region by region, even state by state. And why companies need frameworks. Frameworks that reflect the very real world complexity of roles, industries, workflows, and local regulation.

[00:02:17] And yet, although it's a tech podcast and we'll talk about AI, at the heart of it all, we'll talk about bringing more humanity into work, while also driving productivity. And how HR can actually become the execution engine for this shift, rather than just another casualty on it. So if you find yourself trying to figure out how to use AI for hiring, development, and retention,

[00:02:44] without losing that human element, and without opening the door to fraud and risk, good news, you came to the right place. So that means it's time for me to introduce you to my guest now. So a massive warm welcome back to the show. For anyone that missed our last conversation, can you remind them with a little about who you are and what you do? Yep. I'm Ahibayad, the CEO, co-founder of Phenom.

[00:03:11] And if you really look at what we do as a company, our purpose is to help a billion people find the right world. That's the only reason why we exist. And we actually make the overall HR process much more intelligent, much more automated, but deliver a hyper-personality experience throughout the whole HR ecosystem by an omni-channel infrastructure. Means specific industry, if you want a front line, or if you are really looking for a knowledge worker,

[00:03:41] the kind of experience you get is different based on where you are, and what you're really operating and which device you're on. And whether it's in person or it's on online. So all that is what we really take into consideration, make the overall workflows much more extremely effective and productive. And we last spoke around 18 months ago, I think it'll be Web Summit 2024. And in technological terms, that's a lifetime ago with the speed of everything moving now.

[00:04:11] So when we look back at that last chat at Web Summit, what shifts in AI adoption and regulation have stood out most to you since then? Especially for HR leaders just trying to make sense of the new rules and the new tech landscape. What's changed since our last conversation? So it changed by different reasons. Like it actually been effective in different spots differently. What I mean by that.

[00:04:36] So in Europe, things have changed a lot more differently than what happened in US versus the rest of the Asia pack. Like each particular segment is how much automation can be done and how much automation cannot be done. We each particular layer is actually looking at a bit differently. But the coding is one area where there is so much of automation is done because of that. Like writing a code became cheaper comparatively.

[00:05:06] But not editing and evolving code is not cheaper. It became more expensive. Because all this generative infrastructure generates the code. But every time you ask for a revision, it regenerates the code. It won't take the existing code and improve it. Because of that, like improvisation has to be done by a human. The same thing is actually happening for everything. Like you can really rewrite everything. But like every time if you rewrite, like you cannot really progress. Evolution is not like you kill everything and restart.

[00:05:35] It's about evolutionary infrastructure which you have to bring in to make it precise. So that is something which is really changing. But because of that, like what is happening is policy by policy, it is differing. Even in US, if you really look at state by state, the policies are different. But in Europe, you basically right now have what we are seeing in terms of EU versus EU AI Act versus National Privacy Bill.

[00:06:01] All those things are really giving different viewpoints of how to be compliant with deploying AI in hiring, talent development, like people connecting to the overall companies and all that stuff is actually changing dramatically. And I'm glad you mentioned policy and regulations there because I think over the last three years,

[00:06:23] many companies got caught up in the excitement, rushed towards AI without fully understanding things like the National Privacy Bill or the EU AI Act over here. So how should leaders be thinking about compliance before they think about deploying any AI across hiring or talent development? I think it's more important than ever when we're talking this year, particularly about agents throughout the workplace. There's a lot of considerations to be had here, isn't there? There are two things, right?

[00:06:52] One is people having a knowledge about what are the implications of the law. Part of the law is like is very well written. Part of the law, they don't even know what is written even sometimes. Because the problem is it's not because the people are really writing the policies are really writing it wrong. Where these AI engines are going, nobody knows yet. So the policy is always written a year backward or a six months backward at the minimum.

[00:07:18] And every time a model releases and every model is getting released every three months and multiple revisions are really coming, So the target is constantly shifting. We are in this extreme transformation of what is really going on. So the policies really have to catch up over a period of time. And now the operators has to think how the policy versus the model improvement is actually catching up.

[00:07:44] If you look at last year, the most important thing, what worked was reasoning engines and RL. Those are the only two things which worked, like reinforcement learning and the reasoning engine, and sometimes which is an imitation engine, like imitation learning. Those are the three threads what worked. Because in general, other than Google TPUs, there is no hardware improvement really came into existence.

[00:08:10] So because of that, like the policies of, like if you take the policies in general, policies are written, hey, people don't get just the, what do you call us, chips into the hands. That's not enough. But there are so many policies which are changing on the fly, which is very interesting. But how you apply it to HR is very critical. But without changing the company's culture, without really disrupting the policy of each particular country,

[00:08:36] each particular region, based on the geo and the industry, people have to start thinking. So we built a framework called the Hypercell Framework. What it means is there are five dimensions. The first dimension is what industry a particular company belongs to. And in that, you have to think about which particular function, the business unit or the departmental unit, which you are really thinking about. Then you have to look at which role, then what geo. Based on geo, the policies are different, the rules are different.

[00:09:05] And then now which workflow you are really looking at, what piece of the workflow you can really automate. I'll give you a best example. What is the difference between an autonomous car like Waymo versus an autonomous car which is Mercedes-released, or like which is halfway through? And then you have like what you call as the traditional cars, which have the basic speed lane, like lane management and all that.

[00:09:32] So each particular thing is automating differently, but everything has a policy. But even if you take Waymo, which is Google's autonomous car, it's only operating in probably 10 to 15 cities in the world. So there is a limit where you can apply which particular policy to what extent. And that's happening on autonomous cars. That's happening in every particular field. There is no different right now.

[00:09:59] And you're in quite an exciting space here because FINA works with many global brands. I'll name a few here. DHL and United Airlines, two massive ones there. And I'm curious, when you're working with so many high-profile companies, what have they learned about responsible AI use that maybe smaller and mid-sized teams can, especially if they're listening now, can start applying to? What have they learned and how are you helping them?

[00:10:24] The primary learning is, where is the flaw of the AI can be? And how do you really use evals and the guardrails infrastructure and the policy infrastructure to make sure they operate in the right format? And you use a product like Phenom to make the whole equation work for you so that you can deliver the right experience. But in general, no big company is deploying on a global basis right now.

[00:10:52] They're taking each particular unit where they have the highest pain, where the policy is most relaxed, and that's where they're painting the first picture and really seeing, is it working? And they're also looking at the pain is based on is the company growing fast enough, in which particular spot, and do they have to hire a lot of people or retain a lot of people, and in which country the policy is relaxed a bit.

[00:11:19] And based on that, people are deploying AI responsibility. But what they're really learning is, how can they make the overall experience better, but at the same time, how can they bring productivity? But then they have to really rethink about where the people really go into the equation, where they can move into different spots within the company, so that there are so many things AI can do that can be rectified by using a human in the loop.

[00:11:46] So that is something everybody is learning constantly. And one of the things I've always loved about you is you often talk about balancing automation with human oversight, and a big theme of this podcast as well. So where do you see that balance working well today, and where are they still gaps that worry you? You must say a lot on both sides here. Yeah, so the most important thread what we're really seeing is if they really use the ontologies of understanding

[00:12:15] how their operational layer works, like which is where, how they collect the data, how they take actions, how their data is actually really made. If they can understand that particular data, they are really making the policy work more effectively, and they understand what the constraints are, and then they can turn that into policy, and then what the outcomes are. In general, this is where I'm seeing the maximum benefit.

[00:12:41] People are really looking at how to balance automation and oversight, because there are spots where, like let's say screening, or let's say assessments, you cannot do all of it automated, but you can make part of it intelligent, so that the end user who is taking the final decision has all the inputs to make the right decision, so that automation and human insight is working in parallel.

[00:13:09] But as the software cost is coming down, what is happening is you can write software for a cluster of orchestrations, means in a particular process corridor, in the full talent lifecycle, you can say which pieces you can really make it effective, and which pieces you can make it like in a much more unique format, and the human in the loop is critical.

[00:13:38] There, they are taking advantages of the overall equation. We have seen this multiple times, but our point of view is we as an industry has to bring humanity to work. Like we cannot really take like, okay, whatever is really being done, because the visibility into a full person and how we all live in business has to be correlated. Just deleting an instance through an automation is not the perfect point.

[00:14:08] The operating system is completely switching. So, HR can be the execution fuel to make this whole thing work. So, can we start really treating people as just human resources, as if like it's a resource, rather than people as people, and then really see like which particular pieces where they will be really helpful, but at the same time how the company can be profitable and productive, can also be interlinked for all of this,

[00:14:37] so that business is not at risk of losing humanity at the bottom level. And HR are facing a number of challenges right now from posting a job. There are millions of, not millions, I'm probably exaggerating, there are so many different sites out there now that allow any candidate to apply for hundreds of jobs within a few clicks. I know we've got the rise of fake candidates that surprised a lot of departments as well.

[00:15:02] So, how widespread is this problem away from our news feeds and those big headlines that we've seen? And why has HR become such an unexpected point of vulnerability? So, the problems are different by the industry. Yeah. It's different by the role and the function and the location. So, what I mean by that. So, fraud is different in financial industry in U.S.

[00:15:32] versus fraud is different from a technology industry in India or Vietnam. And that is different from a frontline industry, specifically in the retail worker who's showing up for a job. Or somebody who is working in, let's say, legal in Europe. So, fraud is right now very uniquely structured.

[00:15:58] So, the point is, how can you really bring people together into this five-dimensional cohort and understand where fraud has to be operated? There is no universal fraud. Because the supply demand, the value systems, fraud also really comes with part of how people really think about like haves and have-nots. All those things come into picture in fraud. So, how you validate it do differ in an extreme format.

[00:16:28] So, we build fraud agents right now which can really detect this. But at the same time, they're constantly looking at what is a fake candidate, where they are, where the problem is penetrated, which industry it's penetrated at what level, and what kind of tests you have to really do at every level. Fraud is not injected in the first place. It's injected throughout the workflow. But AI don't really say, like, somebody is fraud.

[00:16:53] AI just indicates to somebody who is in the loop, human in the loop, to say, like, it looks suspicious. Do you want to really oversee it? And the recent case involving North Korean workers inside major US companies was also a wake-up call for many people. So, why should HR leaders be rethinking risks when job applicants themselves can be even part of the cyber threat? Can't they? It's like the gains and the rules have completely changed. Yeah.

[00:17:22] See, but if you really look at, like, North Korean workers working inside US companies, it's not just automation. It's like work from home. It's working from, like, people really working from any particular part, like, not validating the data sets of where these people are in an effective format. All those things are really coming to picture.

[00:17:43] So, people have to really put the human, personal, in-person equation somewhere in the process. We cannot eliminate all of it completely. That is something which people have to start really thinking right now across every industry. And Phenom has built a fraud detection AI agent that supports interviewers in real time. So, that's incredibly cool. But how does this work in practice?

[00:18:12] And what has been the response from HR teams that have not only used it but come to rely on it recently? So, talent identity is one piece. Yeah. It's the same person, really, like, the emails or the names or whatever. They're all connected. Then, what they're really using as, once identity is done, who is doing assessments? Who is doing, like, what do you call as interviews? Then, who is doing screening? Is that really matching? If you're doing on voice, is it matching throughout?

[00:18:41] We are doing in person, are they matching, like, phone versus voice versus in person? And then, when they're really answering a question, how are they answering? Is it really answering through using the Gen AI infrastructure? If that is the answer they're giving, like, how do you really eliminate it? Or, how do you really resist it? Or ask a question which cannot be really answered by the Gen AI infrastructure. So, all that has to be injected into fraud detection.

[00:19:10] That's what I'm talking about. Fraud detection is hyper-personalized at a local level and an industry and a function level and a role level. But there is no universal fraud which will work for everyone. That don't exist anymore. And I always try and give everybody listening a valuable takeaway. So, for business leaders, in particular HR leaders as well, anyone listening, what should their company be prioritizing right now to protect their talent pipelines?

[00:19:38] Well, also using AI to improve hiring speed, development, and retention, and so much more. Any advice that you would offer there? So, we basically really call it five levels of what you call as automation and intelligence you can inject into your HR lifecycle. But you have to think where you really sit today. Are you in level 0, level 1, or level 1.5? And how far you want to move?

[00:20:04] And how far you want to move for which functionality in the company, which region in the company? And then really prioritize what you have to do for those areas. And then really make sure you're putting the right guardrails underneath. You have defined the policies right. You also put the evals right in the back end. So, that whatever you're really deploying there is really bringing you the right kind of an effort without really making humans suffer out of it. And then you have to look at like are they really bringing in outcomes?

[00:20:34] And your outcomes you need not really look for a very long time. You can do it in a week. You can do it in two weeks. You can take a small cohort of your company, try it, see it. If it works, you expand it. All those things are possible right now. Before, it's not possible. You have to really say like, hey, this is the only infrastructure I have to use globally. Because of that, you have to take advantage of what is happening in the market and how you can really build talent pipelines.

[00:20:58] But try in the areas where it will work and eventually really expand it to what is working and dial in and dial out and repurpose the whole human resource infrastructure in an effective way. I think that's a powerful moment to end on. But before I do let you go, for anyone listening wanting to dig a little bit deeper on anything that we discussed today and find out more information about Phenom, how you might be able to help or just keep a lookout for future announcements and contact you or your team.

[00:21:27] Where would you like to point everyone listening? So we have a conference coming up in Philadelphia. It's called as I Am Phenom Conference. That's in Philly from 10th, 11th and 12th of March. That's a good spot to really go after. And then all our social media infrastructure. We also have the same conference in Europe. We have the same conference in Asia. So we actually, but like we do different content pieces, very hyper-personalized for that particular industry and that particular region.

[00:21:55] Well, I'll include links to everything you mentioned there. I do urge people to check out the conference as well. I think it's so valuable getting down on a show floor and just chatting with like-minded people and people that you wouldn't normally bump into or have the privilege of a conversation with. So I'll add links to everything there. And I look forward to speaking with you again, maybe even later this year at Web Summit. I'm not sure if I can attend yet, but more than anything, just thank you for starting this conversation today. Awesome, Neil. Great meeting you again. Thank you very much.

[00:22:25] Keep in touch. So good to catch up with my guest again today, because I think it's a conversation that sat right at the intersection of people, policy and progress. And honestly, that intersection is where most organizations, I will think, are feeling the pressure most right now. And one of the key themes of that conversation was that there isn't a single universal playbook.

[00:22:50] The right level of automation depends on the industry, the role, the geography, the workflow and the risk profile to name but a few. And although, yeah, that sounds obvious when you say it out loud, it's easy to forget when most vendors out there are pitching promises as this, hey, one size fits all solution. And the reality is quite different. And the other theme that stayed with me today is how HR has become almost a frontline function of trust.

[00:23:19] And when candidate fraud is rising, when job applications can be generated at scale, and when hiring pipelines can be exploited by sophisticated actors, it changes the conversation. We need stronger validation, better guardrails and clearer understanding of where humans stay in the loop. We hear that phrase a lot, humans in the loop, but understanding where they fit is equally as important, if not more so.

[00:23:48] So if you want to learn more about Phenom and keep up with what they're building, I will include links in the show notes. And if you're attending the Phenom conference in Philadelphia in March, let me know any takeaways there. And now I'll hand the microphone over to you. What part of the HR lifecycle do you think is most ready for AI support? And where do you think humans need to stay closest in the decision making?

[00:24:15] I know this is a hot potato, but a lot of you will have different opinions on this, and I want to hear all sides. So please hop over to techtalksnetwork.com, leave me an audio message, a DM, whatever is easiest for you. And I look forward to hearing from you. But that's it now. I'm going to rest my vocal cords for 12 hours at least while I prepare for tomorrow's guest. So I'll speak with you all again tomorrow. Bye for now.