Hiring the People Behind the AI Boom: Marcus Pask on Scaling Nebius from 300 to 3,000

Show notes

In this episode of the Leaders in Talent podcast, host Adriaan Kolff interviews Marcus Pask, a talent acquisition leader with 20 years in the field who is helping AI infrastructure company Nebius scale from 300 to 3,000 people in 18 months. Marcus traces his path from an agency desk at 18 through Hays and Expedia to Miro, where he helped grow the company from around 200 to 2,500 people, and explains why Nebius's growth is different: it is tied to contracts already signed with the likes of Meta and Microsoft, not projected revenue.

Marcus and Adriaan dig into the hiring problem nobody talks about, the builders and specialists who put data centers up in remote locations and are not on LinkedIn, how Nebius brings talent acquisition into the conversation early, and why neither of them is worried about AI replacing recruiters any time soon. They also compare how Nebius and Matchr are adopting AI with the right guardrails and context.

Timecodes

01:07 Welcome and Marcus's 20 years in talent acquisition

03:40 Leaving Miro, career coaching, and joining Nebius

07:30 Lessons from starting a podcast

13:30 MIT, robots, and exponential growth

15:43 What Nebius is: the engine room behind the AI boom

17:24 Why growth is tied to signed contracts, not projections

19:13 Hiring for data centers in the middle of nowhere

23:30 Will AI take recruiters' jobs?

29:26 From coordinators to TA operators

34:53 Using AI at Nebius and Matchr with the right context

48:21 How the Matchr and Nebius partnership came about

51:40 Where to find Marcus

___________________________

Connect with us on LinkedIn: https://www.linkedin.com/company/matchr/

Get in touch with us: https://www.matchr.io/who-we-are/contact/

___________________________

Connect with Marcus Pask: https://www.linkedin.com/in/marcuspask/
Marcus's podcast, Talent Unplugged Across the Pond (on Spotify and YouTube)
Marcus's career coaching, Shaping Your Path: https://shapingyourpath.com

Connect with Adriaan Kolff: https://www.linkedin.com/in/adriaankolff/

___________________________

RSS feed: https://media.rss.com/leaders-in-talent/feed.xml

Transcription

[00:01:07] Adriaan: All right, ladies and gentlemen, and welcome to another episode of the Leaders in Talent podcast. Today, an old and longtime friend, Marcus Pask, is on the podcast. Marcus, welcome.

[00:01:20] Marcus: Thank you. It's a pleasure to be here. I'm super excited about joining your podcast and sharing everything about my background in talent acquisition.

[00:01:28] Adriaan: And you're only the second person I've invited to this podcast who has his own podcast, so I'm probably going to be learning from you as well. I'm excited for this conversation.

[00:01:38] Marcus: Me too. This is the first time in a while that I've been a guest on someone else's, so I'm also going to pick up some tips I can take across with me. It's nice to be here.

[00:01:47] Adriaan: Marcus, for the people who don't know you, can you give us a little background about your career and what you're up to these days?

[00:01:54] Marcus: Yeah, for sure. I'm now in my 20th year in talent acquisition, which is crazy. I didn't go to university, so I started when I was 18 years old, straight into a recruitment agency. I knocked on their door and said, "Have you got a job for me?" And they said, "No, we don't have a job for you, but there's a company that needs an on-site recruiter." This was 20 years ago. No one knew what on-site recruiters were back then. It was a brand new thing. So I went and met this lady, and she said, "Yeah, can you start on Monday?" And I said, "Sure, what do I do?" That was my first venture into recruitment. Those first few years, I was figuring out what on earth I was supposed to be doing, on-site, hiring and interviewing people with no training.

[00:02:35] Marcus: It went from there. I carried on agency-side in the UK, working for a large agency called Hays, primarily with investment banks like Bank of America, Goldman Sachs, and UBS, all in Canary Wharf in London. A very different type of customer. I learned a lot of good things, and a lot of ways of not working, at the same time. Then I moved in-house to Expedia, the online travel agency, the equivalent of Booking.com. I spent five years there doing go-to-market, or sales and account management, hiring, and managing a small team across Europe. Then I moved to Amsterdam, where I was lucky enough to join Miro, an online collaboration tool, in theory. People call it something else, but it's still an online collaboration tool. I was very lucky to join them when they were just 200 people, and we scaled it to 2,500 people and 700 million in revenue. I had a team of 25 towards the end, leading all of global tech recruitment across the business, so engineering, product, design, all of those types of roles.

[00:03:40] Marcus: Then I decided to leave at the beginning of last year. I took voluntary redundancy and started my own thing. I saw what I thought was a gap in the market, which was job search coaching, helping people who need to find a job, and doing some career coaching. So I started my own company. While doing that, I also started my own podcast with a dear friend and ex-colleague of mine, Megan, who I worked with at Miro. She was the head of go-to-market recruiting there, so it's nice to have her on board. Then I got approached by a company called Nebius, who I had never heard of. It was an AI company that said they needed someone on a three-month contract to come and help them build the talent acquisition team, the projects, and the tools. A 300-person company, and this was in June, and now by the end of this year we're going to be 3,000. I don't know how, but I've managed to attract myself to companies that go through incredible growth in a small amount of time, and I'm in the right place at the right time. I'm still working for them, and I'll be for the rest of the year, while doing my coaching and my podcast at the same time.

[00:04:48] Marcus: So I'm busy right now, but it's good busy.

[00:04:50] Adriaan: There's a lot to unpack there, in terms of that transition from Miro to starting something for yourself, and now being back at Nebius. Tell me a little bit more about that journey and the gap in the market that you saw. Is the gap still there, and what have the learnings been for you?

[00:05:08] Marcus: Towards the end of my time at Miro, and all the way through my career, I've noticed a few things. One, candidates really don't know much about what happens in talent acquisition and recruitment. They feel they do, but it feels a bit... I don't like to use this word because it gets thrown around a lot, broken. I don't think it's broken, but people say it's broken. They were struggling with getting interviews, with their CV, with having someone coach them through a process, and these could be very senior people or even junior. So I saw a gap. I thought maybe I could use my experience to coach them on what a good CV looks like and how to interview. We do mock interviews and prep, and I use my network to help get them referrals and interviews. I thought that was a gap, but the reason it isn't that successful for a lot of people is that you're feeding into a market where people need the help, but they're either out of a job and don't have the money to pay for a coach. That's one challenge.

[00:06:07] Adriaan: Yeah.

[00:06:07] Marcus: Two, the market is incredibly difficult right now. As much as I can coach someone through the whole process, at the end of the day I'm not the decision maker, so I can't give them a guarantee. So you build a bit of a blocker in terms of what's going on at the company. I've shifted my approach to be more of a career coach for senior-level execs right now, purely because it's coaching on a different level, which I still enjoy and can add value to. I decided to do that because I just wanted to work for myself and do things on my own. And the shift to Nebius was because I wanted to keep a foot in the market. I wanted to understand what was going on in recruitment in today's world, and I felt AI was a huge space over the last 12 months. I thought if I could join an AI company on the side and stay up to date with what they're doing, I could help my customers with my own coaching, as well as my own podcast.

[00:07:05] Marcus: So I was being a little bit selfish at the same time, trying to help my own two products, but also knowing I could plug my skills into a company as a fractional TA consultant. Fractional is another word that gets thrown around a lot these last 12 months. But interim consulting, whatever you want to call it, that's the reason for doing both parts at the same time.

[00:07:30] Adriaan: Tell me a little bit more about the podcast and your experience of starting something like that. I know quite a number of people are thinking about starting a podcast, but they're not sure how, or whether it's the right fit for them. What's your experience so far?

[00:07:44] Marcus: I'm the least techy person, and I didn't understand how the process of running a podcast should work, so I had my wife help me in the background. She hadn't run one herself, but she can get very deep into what needs to get done, and she's very good with brand and awareness and what that needs to look like, and with understanding what the gap is and what you're going to talk about. For me, a podcast was always something I wanted to do, similar to my coaching. I wanted to help people. So the idea of my podcast with my ex-colleague is to give free advice to candidates in the market who don't understand what's happening in recruitment. We always do myth-busting. It could be around whether a CV needs to be two pages long instead of one, whether reference checks are really important, whether you should have referrals in your CV, or whether AI is looking at your CV first and not recruiters. There are a lot of myths out there that we wanted to talk about and give advice on.

[00:08:41] Marcus: The reason for starting the podcast, and for anyone looking to do it, is that it's probably the easiest thing I've done, because for me I'm just talking to other professionals in the market, hopefully about a topic you're passionate about. That's one thing: if you're going to do a podcast, you need to be comfortable with the topic and what you're trying to give, because that's critical. The next part is understanding how you're going to market it. For me, my podcast isn't there to be listened to by a million people. If it is, amazing. But really it's there to help those one, five, or ten people who are really struggling to find a job and need expert help but can't afford to pay for it.

[00:09:27] Marcus: So that's why we do the podcast. Ours is there to fill that niche more than anything else. The one thing we've struggled with most is finding the time to do it and being consistent, setting it in your diary, in your calendar. We do ours every Thursday at 5:00 p.m. European time, because my partner is based in the US. And life happens. She's got children, one of them might be sick, I might have something else going on. We have to move it, then we miss one. So the hardest part is continuing to do it in a seamless manner, every two weeks, once a month, whatever your rotation is.

[00:10:06] Marcus: But also doing topics that are important to you and that you feel will be picked up by an audience, and that takes time too. I'm sure for you, when you're speaking to people, you don't want to repeat everything every single week, or else people aren't going to listen. You have to think about new and fresh things. That can be quite hard, so you have to be up to date, asking people in your industry what they think could be useful, or asking your audience. Have you felt the same?

[00:10:34] Adriaan: Yeah. 90% of podcasts don't make it past the 10th episode.

[00:10:38] Marcus: Oh yeah. We're on episode 15. You're on 25, I think.

[00:10:41] Adriaan: Yeah, something like that. What I really recognize is the consistency. In the beginning we got right out of the gate, we had multiple guests lined up, and then it slowed down a little bit, and life got in the way. It's "oh shit, we didn't have a recording," and then a guest cancels last minute. We had one guest who didn't get approval from his company all of a sudden. So these kinds of things happen. But I find it a very powerful way to have a conversation with people I want to talk to anyway, and to structure it in a way that could be valuable for the wider talent community.

[00:11:18] Marcus: Yeah, exactly. You also found a gap. I looked at all the guests you've had, and they're amazing guests with different perspectives on our industry, whether it's banking or AI or finance. There's a real mix of people, and we've tried to do the same with ours, bringing in HR business partners, someone from compliance, a chief people officer, trying to blend lots of different guests and experiences. I think that's been quite successful, and I see that in yours too.

[00:11:50] Adriaan: I don't know how you see it, but for me the conversations that work best are when the topic is really front and center for me personally and I'm genuinely interested. Sometimes I interview someone about a topic I know our audience would find interesting, or that the person is an expert on, but maybe I don't know that much about it, or I'm just generally less interested in it than other things I talk about. There's a direct correlation in the quality of the conversation. I find the best podcasts are the ones that are relatively unscripted, where it's much more of a conversation.

[00:12:32] Marcus: Yeah, 100%. Every time we finish one of ours, at the end we'll go, "Wow, that was easy." Or you've done 40 minutes and a guest has been on, and you go, "Wow, we really had to pull at that one to get what we needed out of it," and it wasn't really keeping to topic. I feel the same as you: just having general conversations, but with an idea of what you want to talk about, is key. For us, with the guests we bring on board, we make sure each of us does our research, because it may be that my co-host knows the person bringing them onto the podcast and I've not met them. So I have to do my research to really understand who they are and what their background is. Otherwise the conversation becomes very much between my co-host and them, and I'm only coming in a little bit. You can notice it when podcasts happen that way, that it doesn't feel like a three-way conversation. Instead it feels very much like just two people at the end of the day.

[00:13:30] Adriaan: Yeah. Shifting gears, because here's a part I'd also like to dive into. Yesterday I was at an event at MIT in Boston led by Peter Diamandis. He's the host of Moonshots, a very successful entrepreneur, and he has guests like Elon Musk and Sam Altman on his podcast. He's very future-forward and positive about the future and what it has in store for us. It was super inspirational to take a step back, look at the overall landscape, and hear him do a live podcast with three of his guests and think about what the future has in store. One of the things that stuck with me is that they're very bullish around AI and how it's going to revolutionize work and almost eliminate every white collar job. That was the conversation they had. That universal basic income is a necessity because we don't really have to work anymore, since AI will be doing the work. And how close we are to robotics taking over many of the tasks that we deemed something a human always needs to do.

[00:14:52] Adriaan: One statistic that really stood out for me is that the expectation is that this year, Tesla will produce 20,000 Optimus robots, and by the end of 2030 they should be close to a million, if I remember the statistic correctly. To put things in perspective, and this was the part that really stuck with me, when the iPhone came out, the first year they sold and produced a million iPhones. And it was 2007, if I'm not blanking on the year. I think 2007. Fast-forward to today, and that's currently 250 million. So in terms of exponential growth, that was an interesting frame of reference for how exponential these things can go. And of course, with your work at Nebius, when you started at 300 people, maybe give our listeners a little bit of context. What is Nebius, why is it growing so rapidly, and what do you see working in that type of environment when I share these things?

[00:16:02] Marcus: Yeah. So Nebius, and I had to look this up before I joined them because I was also like, "Who is Nebius?" In essence, it's an AI infrastructure company. What that means is we're the engine room behind the AI boom right now. You have all of these AI companies and models that everyone's talking about. They need a huge amount of computing power to actually run, and that's where we come in. So everything around data centers, and the software around it, we are basically the owners of that. Just this year alone, we're going to create, I think, 20-odd data centers across the globe. The data centers are everything that helps the apps, the stuff at the front, the chatbots, the models. They store all of that and make sure the infrastructure is running behind it.

[00:16:51] Marcus: So the big companies like Meta, NVIDIA, and Microsoft can go off and do all the things they want to do with AI, and they use us as the infrastructure behind it. That's the easiest way to describe what's going on. I had another analogy: if AI is the rocket, we're the fuel and the launchpad behind it. That's the easiest way to think about us as a business. And it's growing significantly. Like I said, I joined when it was 300 people. We're going to be 3,000 by the end of this year. I think the difference is when you look at the...

[00:17:24] Adriaan: So that's an 18-month runway. What is that, 24 months? 18?

[00:17:28] Marcus: 18 months. Yeah. When I joined, it was, "Okay, we're going to grow to maybe seven or eight hundred by the end of last year, and we'll look to be a thousand, maybe fifteen hundred, by the end of this year." So within 18 months they were thinking of doubling, which is still a lot for a company that's only been around five or six years. But it's growing way more than that, and the reason is that it's actually correlated to the contracts that are signed. This is all public knowledge because it's a company on the Nasdaq, and in essence it's linked to the contracts they signed with Meta and Microsoft, worth twenty billion, twenty-five billion. So all of the hires we need to make are so that we can go and build the data centers to support the contracts we've already signed. A lot of companies, when they say, "We're going to go from five hundred to two thousand by the end of next year," that's based on predictive revenue they hope will come in and what they think about the growth of the market.

[00:18:24] Marcus: The difference with Nebius right now is that the contracts are already signed. This growth is just to make sure we can fulfill those contracts. So if you start adding more contracts onto that, the size gets bigger and bigger. Right now within Nebius, and I'm sure we made this public not long ago, we're saying no to contracts at the moment. We are turning down huge business because we also don't have the space to store the GPUs and everything, the storage that's needed. So we're being very particular about the clients we sign up with, which is why we also need to go and build even more data centers, and that's the biggest part of how we're doing it.

[00:19:02] Marcus: So it's a huge market that I knew nothing about. I still only know a crumb of the loaf right now, but it's an interesting place to be.

[00:19:13] Adriaan: Super competitive space, right? A very niche talent pool. Data centers typically don't get opened in the most densely populated areas. You need huge plots of land, very specialized roles. Help me understand a little bit more about that talent landscape, the fight for talent, and what's happening there.

[00:19:34] Marcus: Yeah, this is where it gets interesting, because you're completely correct. To build data centers, they're usually in the middle of nowhere, on a huge amount of land, not close to a major city. Or if it's close to a major city, it's a city that usually won't have the type of candidates you're looking for. What was happening at the beginning of Nebius was that we would go and open these data centers, and it would be like that. We'd sign up, get the data center, and go, "Hey, recruitment, now we need you to plan how we're going to fill these data centers and find the talent." Now it's switching, where we get involved in these conversations beforehand and come in early and say to the business, "You want to open up in this town in North Carolina? Okay, here's a talent map of that area. This is where the talent is. This is how difficult it's going to be to hire these people. This is the compensation you need," and so on. So now we're coming earlier to the table. To give you an example, we opened a data center in Iceland. Iceland? Finland? Iceland, I'm pretty sure. The nearest town is an hour's drive away, and nobody wants to go to this town where we've built the data center. So to incentivize, we pay for possibly a car, possibly a travel allowance. We don't expect them to be in five days a week, because guess what?

[00:20:59] Marcus: In Iceland it gets pretty cold and pretty icy, and sometimes the roads are blocked, so it's not always possible for them to go in. So we also put them up in hotels. We sort out lodging. There are ways we're trying to attract talent and give them the right benefits to be interested in joining us. But we also talk about the story of how big we're going to be and what this is going to look like, and the fact that we can say we're going to open up 20 more data centers. We're trying to think of a model where you don't just hire the team locally. Instead, you hire a team that goes from one data center to the next, and they basically move with the job.

[00:21:37] Marcus: So you rotate them around rather than them being fixed to one state like Carolina. They'll do two months building in Carolina, then they'll go to Arizona, then New York, and so on. They'll do those bits all around, as a way for us to retain talent and to not feel like we constantly have to start finding people in each place. Instead, we can rotate them around to certain places. It doesn't work for everyone, but it's a good way for us to attract and understand the market that's out there. And the biggest part, as I said at the start, is bringing us, TA, into the conversations way earlier, so that we're not in a position where we open a data center in Iceland and then wonder how we're going to fill it.

[00:22:22] Adriaan: Yeah. What I find interesting, and what I also thought was lacking yesterday in the conversation, is that there's a huge constraint on compute right now, especially with the growth of the models. If you look at SpaceX recently buying Cursor, or at least having the intention to buy them, there's a clause that if they don't, they have to pay a 10 billion fine. The reason this was a very lucrative deal for Cursor is that Cursor was running into compute constraints. Tesla, or SpaceX, is hoping to get their Colossus up and running in the next 12 to 18 months, depending on how hard Musk wants to push, and they actually have compute. Compute is one thing, but I rarely hear people talk about the people who need to build these data centers in all these different locations. Yes, AI is going to take our jobs, but I don't see that being resolved anytime soon. What's your point of view, being on the ground and experiencing this?

[00:23:30] Marcus: Absolutely not. I completely agree. It's not going to happen with some of the people we need to find, because you need to find builders, and AI is not going to be able to build that type of infrastructure on the vast amount of ground that we need. Same with electricians and plumbers, and then you need people to look after the storage and everything else that goes with it. For us, the market right now is such that if you're able to build data centers, you're at a premium. You're paying good money. They're well in demand, if anything just as in demand as the engineers or anyone else we're trying to hire, which is great for an industry that usually doesn't experience that type of growth.

[00:24:14] Marcus: But the market right now is so competitive that to stand out is really hard. Would you go to something known like Amazon, where you can build one of their sites, with the land and the size of a data center that might not be used the way we use it? Or do you go for something called Nebius, where, again, our employer branding, our careers page, and our general branding are very, very small? Not many people know about us as a business. You'd have heard of NVIDIA, of Meta, of Tesla. People will go to them. To come to us is a different story. So usually you've then got to do something else to incentivize them. The market itself is extremely competitive and moving so quickly that I've also seen recruitment agencies suddenly turn into data center recruitment agencies as fast as I've seen in the last 12 months. They've made that pivot from being a traditional agency that covers all types of tech to saying, "We're just going to focus on this part," because the need is there, and companies like us are happy to pay for it, because we know we're going to struggle to find the talent. You also don't find that talent on LinkedIn.

[00:25:33] Marcus: You find that talent through local market referrals and so on. So how you attract them is a very difficult thing to do. But as for how Elon's feeling and how the whole AI community's feeling, I don't feel any of that. I don't think our jobs are at risk in the next five years. Automation is going to happen, and that's already happening, and I'm all for it in the right way. But am I worried about the risk to a recruiter right now? No. I think the human touch still needs to happen. Automation will happen with reviewing CVs, creating contracts, and scheduling interviews. But a fully AI-oriented process, where you're speaking to an AI bot all the way through, I don't think we're ready for it yet.

[00:26:19] Marcus: In five years, maybe I'm going to be proven wrong and I really don't have a job, but I just don't think it's going to...

[00:26:23] Adriaan: And then this will have aged well.

[00:26:25] Marcus: Yeah, exactly. What do you think? Do you feel like it's going to...

[00:26:31] Adriaan: I don't think so. While you were talking, I was thinking about when I was a student. I once did a business course at Booz & Company, a strategy consultancy firm. We had a three-day course where the goal was to design the city of the future. The time we were designing for was 2050. Everyone looked at different elements of the city, and we came up with business cases. We thought really out of the box: what would a city in 2050 look like? I remember we presented to one of the senior partners. We focused on housing. We had designed this render of huge skyscrapers where people would live in apartments, everything recycled, no emissions, all sustainably built.

[00:27:30] Marcus: And the partner looked at this and said, "Describe your dream home. What would your dream home look like?" And I said, "What do you mean, in 2050?" He said, "No, no, now. What would your dream home look like?" I said, "Well, I have a garden, hopefully on a lake or maybe on the beach."

[00:27:47] Adriaan: "You know, with two floors. That's definitely my dream house." He said, "Okay, next one." Same type of house. And the third one, also somewhere in the mountains. And then he said, "But you've designed 2050 with everyone stuck in these skyscrapers and boxes where everything is sustainable. Do you think the needs and desires of people will change that much?" It was such a good question, because that's not how we thought about it. We just thought we're going to have flying cars and we're all going to live in these towers. So this is a long-winded way of saying I don't think there will ever be a desire for people to apply at a company without interacting, at some point in the process, with an actual human.

[00:28:33] Marcus: Yeah, we're seeing that, actually. As candidates go through an interview process, they'll say, "Okay, what are my roles and responsibilities?" It's a good question. It's what we put in the job description. And then they'll say, "Well, that's just a part of what I do in my current job. It's only 40% of what I want to do. What about the other 60% that I still enjoy?" "Oh, no, we've automated that." Okay, well then where's the fun and the love in doing that 40%? As a recruiter, if you told me I wasn't going to be able to speak to candidates, or to my hiring managers or stakeholders, or to the business, and instead all I needed to do was put it into an AI bot and it would send it out, that's not fun. That's also not going to get people interested. So I understand where you're coming from, and that's also an unbelievable set of questions he came back to you with. That really gets you thinking.

[00:29:26] Adriaan: Yeah, it gets you thinking. There was another great example. I do see our profession changing significantly. Are we going to need fewer recruiters? Could be. Who knows? We are going to be able to do more with the tools and technology that's out there. I strongly believe our recruiters will be much more like TA operators, where the top of the funnel, and probably that first part of the screening, can be automated and done. So that the candidates you do interact with are the right fits and have already had a great candidate experience because of the automation that's been put in place. But you're there to build that human connection and then help them navigate the rest of the organization. In, I think, the 40s or 50s, don't pin me down on the year, telephone systems still worked through a switchboard, where someone had to plug in a line to connect you with the other person, and there were 800,000 predominantly women working in that industry. That's a huge number of people, especially for that time.

[00:30:32] Adriaan: And then Bell invented the switchboard, so instead of a human having to plug in these lines, the switchboards did it automatically, and within a few years no one was working in those switchboards anymore. And guess what happened to those people?

[00:31:07] Marcus: They went and did some customer service.

[00:31:10] Adriaan: Exactly. Oh my goodness, Marcus, you're going to be a returning customer. Exactly. Because of this evolution, 1-800 numbers were made possible, because all of a sudden it became cheap to make phone calls. And because of that, all these businesses started to open up, and they needed customer service. So those 800,000 people who were doing the lines were then rehired to be customer service agents. With any technological change, this is what happens, and we're just seeing it at the forefront of what's going to happen to many of the jobs out there.

[00:31:49] Marcus: Yeah, and I see it with coordinators, talent acquisition and recruitment coordinators. What you're seeing now in our world are scheduling tools that help schedule. It means you don't need as many coordinators, because you've got a tool that can take a lot of that away. But what it also means is that a coordinator is being asked to make sure the tool is working correctly, that the automation is working correctly in the background. So the role is shifting from "I'm going to coordinate" to "I'm going to make sure the tool is doing what it needs," and then they can go and look at process improvement, or at some of the metrics of who's using it.

[00:32:08] Marcus: So they become, as you mentioned, a little bit more TA operations than coordination at its core. So roles evolve rather than necessarily getting replaced, in my eyes. Something I'm doing on my own is testing a lot of these on other companies. I applied for a job the other day, a freelance thing, and it said, "Your whole process will be AI-done." And I was like...

[00:32:54] Adriaan: Yeah, yeah.

[00:32:57] Marcus: I was qualified, so I thought, "Okay, I'm hopefully going to go through the process." The first part was just things I had to write out, making sure I put my hand in front of my face so I was still there, and that I also wasn't using AI. Then it said, "Okay, an AI bot is now going to ask you some questions." It would ask me a question, I'd answer it, and then it would say, "That's a really good answer. Thank you. Okay, you've talked about this, this, and this. Next question." As it went through all of this, it asked me five questions, and at not one point did it ask me a follow-up question. Instead, the next stage is that someone, a recruiter, has to go and watch my video and my answers, which will take them 30 minutes, because that's how long the interview was. And I think, what time have you saved there? The AI has done it, but now someone has to watch it to make sure I'm doing it, and they're not able to ask me a follow-up question about what I'm saying to really unpick the truth. That's where a human is able to go, "Oh, tell me more about this," or, "Actually, that's quite interesting. Could you explain that?" That bit was very interesting to me. Is this a time-saving piece, or an evolution we've not quite got right? I was in limbo, not quite understanding the intention. But I was also quite happy, because it made me realize that recruiters are very much still needed in that particular part of the process, at least.

[00:34:27] Adriaan: That's funny. Yeah, it doesn't make any sense to outsource that. Can you imagine being a recruiter and having to listen to the AI interviewing someone? That's even worse.

[00:34:38] Marcus: Yeah. And every time it said, "That was a great answer, thank you for that," I'm like, "Well, was it? Are you going to tell me it's a bad answer? I don't know. Is this just an automated answer?" That's where it becomes a little bit impersonal, which I struggle with.

[00:34:53] Adriaan: Yeah, I turned off, with all my AIs, the positive affirmation with every question I ask. Hey, switching gears a little bit, this is something I'm interested in. I know at Nebius you work with Claude and Cowork. It's rolled out across the company. Can you share a little bit of best practices, or how you operate it at scale? Are there specific guidelines? How is the work of Cowork being done? Tell me a little bit more about the setup.

[00:35:32] Marcus: Yeah, it's funny. We're an AI infrastructure company, but there was a little bit of hesitation to begin with about using Claude and other tools across our business, which is ironic to say the least. Each team is using it differently. However, we do have general company guidelines we need to follow in terms of the data we share, the files we upload, and the things we talk about, because it's going to be stored somewhere. If you're talking about an acquisition that's about to go through, and it needs to read a contract or pick up the terms, that's data security we're worried about. So there's a lot on the security side that we're still unpicking, and our privacy bar at Nebius is really high.

[00:36:19] Marcus: I've worked at a lot of companies where privacy is important. I've not worked at a company where privacy is as high up as it is here at Nebius, which is amazing. But it does mean we're using it in a very basic way, which surprises me, that we still want to do a lot of things through normal conversations and not put it into AI, just making sure the automation of things is done in a human way. But there are teams that are really pushing forward. We have AI teams looking at how we blend the people org with the talent acquisition org and other parts, and it uses Claude to help with that. As well as that, all of our engineers are welcome to use Claude and other AI tools for coding, because engineers at every other company are using AI for coding, which surprises a lot of people.

[00:37:17] Marcus: A lot of people think, "Oh, why should they be able to use that, because that's cheating?" It's like, no, it can help you with it, but you still need to know what the code is and double-check it, so it's fine. So to answer your question, it's in its infancy, I feel, with how we're adopting it, purely because of our privacy side. But across the business, we have rules and regulations we all need to follow, and we need to make sure that if we're using it, it's used to solve the right problems, not just everyday mundane tasks like replying to an email or doing X, Y, and Z. It needs to be done with purpose. That's the way we're seeing it right now. But we're still not as advanced as other companies at the moment, and that's intentional. Do you see that differently across other businesses and the people you speak to?

[00:38:03] Adriaan: It varies across all of our clients, including ourselves. We were building a lot with AI. We've built a lot of automations and done a lot on the backend in terms of our operations, that have helped. The difficulty about how we work as an embedded company is that we work in our client's system, so there's only so much we can do. It's predominantly the freedom and access to tools we receive from our clients that allow us to do anything meaningful or not. So that's a bit of a constraint in the majority of cases. What we're now implementing ourselves is rolling out AI across the organization, putting the right guardrails in. Similar to what you say, we need to be very careful about what data we can and can't use when we work with AI.

[00:38:59] Adriaan: And we are now creating the context and the environment so that everyone who uses AI understands the context and the environment we operate in as a business. What I mean is that, especially now with the power of Cowork and some of the more advanced AIs, you can really create context specifically for your use case, and that's definitely where we've seen the biggest benefits. The context we're providing now is, one, overall: who is Matchr, what do we stand for, what does the competitive landscape look like, what does great recruitment look like from an overall perspective, and then also how do we operate as a business? What are all the steps our recruiters need to do from a delivery perspective?

[00:39:47] Adriaan: That's a big master file that we control and update across the organization, and then whenever someone has a question about Matchr or how to operate, it only pulls data from that context. That's been a big game changer compared to how we previously used AI, where AI would understand who Matchr is but had no idea how we operate and work. Especially when you take time to really read through some of the answers, you see that 70% is good but kind of generic, and then that 20% to 30% is either wrong or just not really correct. So that's something we've found to be very helpful: giving it the right context, and also staying in control of that context.

[00:40:41] Marcus: Yeah. For us, we would like to be where you are in maybe six or 12 months' time. The challenge we have is that because we're growing so quickly, it's difficult to understand the landscape of where we want to be as a TA team, a people team, and an org, because everything is changing so quickly. But I do think there's soon to be this idea of what our purpose is as a talent acquisition function, what our goals are, what our values are, all of those pieces. For us, because we've grown so quickly, a lot of it still needs to be built, and it needs to be very intentional, which is also a difficult thing to get right, bringing people into a room, virtually or in person, to have their say on it.

[00:41:30] Marcus: That's a challenge we've had, going from 12 when I joined to 80 people in the TA team in the space of nine months. Everyone is understanding what their job is first. So to then bring them on the journey of who is Matchr, who is Nebius, what are our values, how do we work, how do we operate, we're also trying to figure that out, because the new leadership team is coming in to manage some of these parts of the org. So I think we'd like to be where you are, at least in the future. But for now, we're playing it safe, just because of the revolving doors and the expectations happening right now within our business.

[00:42:11] Marcus: But it will come, and it should come, similar to you.

[00:42:15] Adriaan: Yeah.

[00:42:16] Marcus: If you're going to use it and then you need to check 30% of it and realize that 30% is wrong, you're like, "Where is it giving me the saved time? Was it worth me just doing this myself, knowing I'm going to have to check it over?" What is the value of it sometimes, versus actually doing it yourself?

[00:42:36] Adriaan: I'll give you a good example. I spent probably six hours, if I add it all up, giving it the sales context of Matchr. Also because in the beginning I hadn't saved something correctly, which is part of the iteration. What I've seen is that AI is your best mentor and teacher. I just take screenshots: why is this not working, why is this going wrong? And eventually you get somewhere. What I did is I uploaded all of my sales conversations from the last year and had it interview me, to really understand what it means to do sales within a Matchr environment.

[00:43:14] Adriaan: I uploaded five sales presentations, and our head of marketing created a Matchr brand skill book, which I uploaded into the AI so that it would know our branding, our fonts, and how we want to structure our presentations. That probably took me five to six hours to really get dialed in, so that whenever I ask it a question, it really understands how sales works at Matchr. What was interesting is that I had a sales conversation with a well-known global strategy consultant, and I uploaded the 40-minute transcript into my sales context. I call my sales context Rob. I asked it two questions. I said, "Analyze what I did well in the sales call and analyze where I can improve."

[00:44:08] Adriaan: And two, "Can you create the outline of the presentation, selling into their needs?"

[00:44:15] Marcus: Yeah.

[00:44:15] Adriaan: One, it gave me really good feedback on some of my blind spots. I didn't properly lock in a follow-up conversation. I should have been harder on that. There's a manager the team had to get approval from, and I did a good job of asking what's going to be important for this manager, but I didn't do a good job on where he sits in the overall decision-making structure. So, very useful feedback. But what really blew me away is that it then created a presentation in our style, completely tailored to the needs of this particular customer. Every slide was tailor-made, and it did that based on the context and a previous presentation I gave it.

[00:44:59] Adriaan: I probably spent four hours redesigning it to a level where I was like, "Okay, now I can send it to the client." That was mainly because some of the context wasn't 100% there, or it hadn't used some of our previous graphs that are powerful, it didn't have the rate cards yet, and I didn't have use cases. The use cases took me quite some time to get to a level where, both visually and from a content perspective, we were at the right level. But normally I would have sent our standard deck with one or two custom slides, and the rest standard, and I'd have used a designer for an hour or two to make everything look nice. Now the entire deck is customized, and I didn't pay a single designer to help me with anything. And by the looks of it, they're actually going to become a client.

[00:45:54] Marcus: Oh, wow. It shows the investment you put into that early part, but the iterations are super important. I've done the same with my podcast setup. I uploaded all of the episodes I've done already, and it was like, "This is what's important, this is how we do our descriptions, these are the topics we've talked about." And then, "Okay, this is the branding. Can you now create a LinkedIn graphic of five tips on how to nail an interview," for example. And it does it in the branding, it does it from previous episodes, it relates it back, and I'm like, you've saved me a good hour on something that's quite simple for a LinkedIn post. I only had to spend probably 10 minutes editing it, whereas building it from scratch would have taken me quite a bit of time. Now it's generating those as quickly as I need them done, but with content that's at least relatable to everything needed. From your side, I used to hate it when I'd get these generic...

[00:46:52] Adriaan: Sorry.

[00:46:53] Marcus: ...ones coming through. No, you guys are pretty good, actually. But some come through and it would be one or two slides different. It's like, "Who is Nebius?" I'm like, "Yeah, I know who Nebius is, I work there." Or it would be something different. If you do a fully tailored deck, the person on the other end knows you've done it fully tailored for them. We've seen so many that I think in your industry that's going to be super important. So kudos for spending the time on it.

[00:47:17] Adriaan: Yeah. No, super interesting. Marcus, anything else you want to chat about? We're 47 minutes in, and normally I do 20 to 30 minutes, but this time flies by.

[00:47:33] Marcus: No, listen, I think from both of our sides, we're trying to give information back to the TA world, whether it's from leaders or other parts. What I would ask is that anyone who listens to this, anyone who sees it, share it, because we're trying to give advice and help people. It's nothing other than that. So help the people around you. The job market is super tough right now. People are still struggling to find jobs. The more we can do to support, the better the industry will be. We do still have a bit of a bad name in the industry, whether it's from recruiters themselves or the industry we're in. The more we can do this, the quicker and easier it will be to change that narrative.

[00:48:18] Adriaan: Amen.

[00:48:20] Adriaan: How is the cooperation going so far with Matchr and Nebius?

[00:48:25] Marcus: Wonderful. Do you want me to quickly share how it started?

[00:48:29] Adriaan: Yeah, share a little bit. Now that I have you on stage.

[00:48:34] Marcus: So Adriaan and I worked together. I brought Adriaan and Matchr over to Miro when I was working there and we were growing the team. When I left and joined Nebius, we were looking to expand vastly across the whole business, and we were looking for an embedded solution. We met with 10 agencies or companies that could offer us this across the globe. I interviewed all 10 of them, which took me a while, and looked at all of those decks that got sent to me, which were generic, to say the least, from most of them. Then I whittled that down to three, did a full interview with all three, and really went deep into what our needs are and what people could offer.

[00:49:13] Marcus: I already knew you guys would be able to deliver way beyond what we needed. At that time it was four or five recruiters we were looking to bring on board. That quickly escalated to eight, then 12, and now I think we're at 16 or 17, somewhere around that number. So the numbers have escalated way above what we were thinking. The reason we went with you guys is that your embedded solution is everything we need. You can do it across the globe, in the US, in Europe. We're looking at very different types of profiles: sales, engineers, and, like I said, people who can build data centers. That's not an easy skill set to look for. And you guys always bring us the right data. You tell us how the team is performing without us needing to ask. We know what the market is like, and it feels like a true partnership, where you want to know what our challenges are, how you can improve, what we can do better, and help us evolve to a 3,000-person org.

[00:50:20] Marcus: I always feel like previous embedded solutions were all about "give us the recs, give us the jobs, we'll fill them, move on." Whereas with you it's more, "Okay, what is the biggest problem you're trying to solve? How do we do it all together? How do we find solutions, and how do we make this a real working partnership?" I think that's why it's working so well, and long may it continue. For me, it was a no-brainer. I'm just happy you guys are doing such a great job. So kudos.

[00:50:51] Adriaan: Yeah. And honestly, why do some clients work better than others? It really is a partnership. When people ask me in our sales calls, "What defines a good cooperation?", it really is being treated as a full-on partner. You celebrate the wins, you battle through the challenges, and that's how you come out successful on the other end. Testament and kudos to you, Marcus.

[00:51:18] Marcus: Yeah, but that's how this works. We've had honest conversations where things aren't working on our side, or on yours, and then it's "this is going great," and we just make sure we talk. We've all got the same end goal, which is that we need to hire amazing people for Nebius. How do we go about doing it? We both want to get there in the right way, so we just work together to get it done. So I think it works great.

[00:51:40] Adriaan: Marcus, where can people follow your podcast? Where can people follow you?

[00:51:44] Marcus: Yeah. On Spotify and YouTube, Talent Unplugged Across the Pond is the podcast with my co-host Megan. And if it's any sort of coaching, it's shapingyourpath.com, where you'll find my website for all of my job search coaching and career coaching as well.

[00:52:00] Adriaan: Amazing. We'll put it in the show notes as well. Marcus, such a pleasure talking to you. I appreciate you.

[00:52:05] Marcus: Thank you for having me as a guest. Good luck. Thanks, Adriaan. Cheers.

“

Need external recruitment support? Embed Matchr's expert recruiters directly into your team. Seamless integration for immediate impact on your hiring process.