AI in Recruitment: What to Automate and What Must Stay Human

AI is becoming capable of handling more recruitment tasks, but it should not own the moments where trust, judgment and accountability matter. Vanessa Raath explains how to combine AI tools with human recruiting skills — from CV formatting and sourcing to screening, interviews and compliance.

Vanessa Raath · Talent Sourcing Trainer and AI Strategist for Recruiters
Hosted by Ksenia Izotova and Timur Novopashin

Vanessa Raath is a talent sourcing trainer and AI strategist for recruiters. Her background spans agency recruitment, internal talent acquisition, employer branding and more than two decades in recruitment and HR. She works with recruiting teams on practical AI adoption, sourcing skills and the human judgment that technology should support rather than replace.

Connect with Vanessa on LinkedIn: https://www.linkedin.com/in/vanessaraath/

Recruiters should use AI for repetitive, data-heavy and administrative work — such as formatting CVs, organising interview notes, researching talent markets and supporting sourcing — while keeping people responsible for interviews, feedback, judgment and relationship-building. The right tool depends on scale, repeatability, integrations, privacy requirements and whether the task can be handled by a general-purpose AI model before the company buys specialised software.

Vanessa Raath proposes a simple division of work: AI should handle the tasks, while people should handle the trust. The strongest starting points are repetitive activities that consume time but do not deserve a recruiter’s full attention — formatting documents, organising information, preparing research and moving approved content into a consistent structure.

One example from the conversation is CV formatting. A recruitment team can create an approved workspace in an AI tool, provide completed CV examples and brand rules, and then supply the raw CV, interview notes and transcript. The model produces a first draft in the company format; the recruiter reviews accuracy, tone and omissions before anything is shared. This replaces copying and pasting, not professional judgment.

Human involvement becomes more important whenever the work affects trust: interviewing, giving feedback, explaining a decision, persuading a passive candidate or handling a difficult conversation. The purpose of automation is to create more time for those interactions, not to eliminate them.

There is no single best AI tool for every recruiting workflow. In Vanessa’s own setup, ChatGPT is the broad all-round starting point, Claude is useful for writing and complex tasks, Gemini is used for image work and Copilot supports email. Her recommendation is to test the same well-designed prompt in several approved tools and compare the quality of the results for the actual task.

Before purchasing specialised recruitment software, teams should first understand how far their existing approved AI environment can go. A dedicated tool becomes more attractive when the process is highly repeatable, used at significant volume or needs deeper integration across a large team. The decision should also account for security, data protection, existing systems, implementation effort and the cost of maintaining another tool.

The evaluation question is therefore not simply "Which tool is popular?" It is: What task are we solving, how frequently does it occur, what data does it use, what level of review is required and can the result be tested reliably?

AI is changing the work of recruiters and talent sourcers, but the episode does not support the claim that sourcing has disappeared. AI can help with market mapping, competitor research, search planning and other research-heavy work. The result still depends on the quality of the prompt, the recruiter’s understanding of the market and the ability to recognise when the obvious search is missing the right person.

Traditional sourcing skills remain useful: Boolean and X-ray search, exploring GitHub, Stack Overflow, Kaggle, Reddit and professional communities, and finding passive candidates who may not have a polished CV or an active LinkedIn profile. A screening system can evaluate only the information it receives; a recruiter can investigate context that is not present in the application.

The practical conclusion is that AI is more likely to change the balance of the job than remove the need for recruiters. People who can combine sourcing fundamentals, business understanding, candidate communication and AI fluency will be better positioned than those relying entirely on either manual methods or automation.

The value of an AI interview depends on the context. A voice agent may offer convenience in high-volume graduate hiring because candidates can complete an initial conversation at a suitable time. The same approach may be inappropriate when a recruiter is approaching a senior passive candidate who expects a personal conversation. Volume, seniority, candidate expectations and the purpose of the stage all matter.

AI CV screening also varies significantly by product. Some tools may rely heavily on keywords, while stronger systems may use broader context. CVs themselves are incomplete representations: candidates omit courses, describe experience poorly or have relevant work that does not fit the expected pattern. Human recruiters are imperfect too, which is why the tool should be tested rather than accepted or rejected on assumption.

Vanessa recommends creating a benchmark: ask several recruiters to review the same candidate sample, identify their strongest candidates and then compare the AI result. Review what the tool missed, why it ranked people as it did and whether the team can challenge the output. Do this before committing to a long contract or allowing the system to shape live hiring decisions.

The episode correctly treats AI regulation in recruitment as an area that HR and TA teams must actively monitor. Certain AI systems used in employment and recruitment fall within the EU AI Act's high-risk categories. Following the 2026 AI Omnibus changes, the high-risk rules for Annex III systems are scheduled to apply from 2 December 2027.

That future application date should not be read as permission to ignore risk today. Organisations should map where AI is used in recruitment, understand which decisions it influences, document human oversight, assess candidate-data handling and follow current guidance from the European Commission and relevant national authorities. Classification depends on the intended use of the system, so companies should obtain appropriate legal and compliance advice for their own processes.

Editor’s note: This regulatory summary was checked on 25 August 2026. It is general information, not legal advice.

This transcript has been edited for clarity and readability while preserving the speakers' meaning.

Ksenia Izotova: Vanessa, I know that you’re a certified scuba diving instructor. So does it help you somehow in building a recruiting company?

Vanessa Raath: I would say yes, most definitely on both accounts. You’ve kind of really dug into my past and dredged up my very early careers there, which is fine. There’s not a problem with that. Actually, it’s an interesting analogy because teachers make really good HR people. The thinking behind that is that they have that kind, nurturing nature. They like to empower others. They like to teach people to be better versions of themselves.

And then from a recruiting perspective, a teacher is often used to dealing with loads of different personalities, so they know how to match which children should sit together and which children shouldn’t. And that’s kind of like recruitment, isn’t it? Who should work where, who should they work for.

I always say to people that if you can read someone’s eyes 30 meters below the sea, and you can tell whether they’re going to panic, are they going to shoot to the surface, are they going to touch something they shouldn’t — you can read people. You don’t need any NLP kind of certificates or anything like that. You have a good understanding of individuals.

Vanessa Raath: So I refer to myself as the AI fluency and AI strategist for recruiters. I find myself in this position in 2026, where I’m doing a lot about where AI can improve the recruitment industry, where it should be used, where it shouldn't be used, in this day and age, because there are some people that are doing this really well and some people doing this really badly.

I’ve run my own business now for nearly eight years. I started out doing talent sourcing training, which when AI came around, I thought, I need to get stuck into this AI because I think it’s going to have a huge impact on our industry, which it has. Before that, I’ve done internal TA. I’ve been a head of TA and employer branding. I’ve also done agency recruitment. So I’ve pretty much been in the HR recruitment space for probably about 23 years now. 20 years definitely doesn’t sound short to me.

Timur Novopashin: I get what you’re saying, and I also have a quite similar career path. So I’ve done agency recruiting, I’ve done in-house recruiting, and now basically I’m empowering my peers, talent professionals and HRs, with best practices on the market. Kind of get into the more of a community, talent community builder setup. And there is a very hot topic that I’ve heard discussions happening across the board — of course, that’s AI and sourcing, AI and talent acquisition. Sourcing is dying, interviewing is dying, et cetera.

So I would love to hear your take on the role that AI is currently playing in sourcing and recruiting, so basically your areas of expertise.

Vanessa Raath: My take on AI is very simple, and it is as follows: AI should handle the tasks, and humans should handle the trust. That’s literally what it boils down to. AI is amazing. It’s a really, really fantastic tool, but it shouldn't be used absolutely everywhere in recruitment, TA, and HR. I think that those roles have become really, really admin intensive. I mean, if I look at recruitment over the last 20 years, you’re always updating something here, then putting it on a spreadsheet, and then messaging this person, and then updating it in another place.

We should be using the AI tools in order to assist us to do those skills and that boring admin which no recruiters ever want to do. We should be using AI to do that. But when it comes to building trust and human relationships, that’s when humans have to kick in because we don't want to hand that over to AI.

So I’m all for using AI to augment what a recruiter does at the moment, but not to replace what a recruiter does at the moment, and there’s a very clear definition there.

Ksenia Izotova: Can we unpack that?

Vanessa Raath: For me, whenever there’s a human touchpoint — whether that is an email, an interview, giving feedback — first prize for me, and calling me old school, is always going to be picking up the phone and chatting with someone. It could be a video call, could be an old school phone call, a voice call. Those things are skills which you need to do regularly. You’re building trust. You’re fostering a relationship.

But other skills, like for example, how long did you in your recruiting careers take taking someone’s raw CV and putting it into the company format? Here you are, you are an employed professional, and you’re copying and pasting from one screen to another into a new CV format. That for me is just a complete waste of time. Now, what you’ll probably find interesting to know — and I know this because I have such a global reach and I train teams all over the world — there are still recruitment teams that do that. They copy and paste from raw CVs, and they put it into CV format.

Now, that is a prime case for AI. It would take a human what? 15, 20, sometimes 30 minutes to do that, but the AI can do it in less than a minute. If you’re letting AI handle the tasks that take up your time… And if you think about it, think about the average profile of a recruiter. A recruiter doesn’t go into recruitment because they love admin. You’ve worked in teams, I’m sure. How many times are the recruiters being yelled at by the manager, "You didn’t put it in the CRM. Why didn’t you update the ATS?" Recruiters don’t love doing admin and those kinds of tasks, so you’re not using them to the best of their abilities by copying and pasting into a CV.

My take on that is that let’s let the AI do those boring administrative tasks, but whenever there is a human touchpoint, let’s be human and let’s spend more time with human beings.

Ksenia Izotova: That really resonates because at UnifyHub, what we are always trying to do is communicate as much as possible with the clients. If I have an issue, a question or something like that, the managers know that it’s better to have a call for 10 minutes rather than start sending emails back and forth or sending messages back.

Vanessa Raath: Exactly.

Ksenia Izotova: That’s right. Voice call, that’s the best way. I have a question here about taking someone’s CV and putting that into the company formats. How do you do that?

Vanessa Raath: There are a couple of ways, two ways to do that. There are companies that exist that do that. They are AI companies. They have built tools, they have built products which you can go and have a look at. I hope as part of this conversation, we can share a list of my AI Tried and Tested Tools. There are a couple on there that I recommend. It’s quite interesting because a lot of companies started doing that, and now they do other things, and they’ve added on to what their AI can do.

That’s one way to do it, and if you’re a big company and you’ve got a really big budget, go for it. Use that. However, what a lot of recruiters don’t realize is that you can actually set that up on a tool like Claude. I would start a project. I would recommend that you pay for Claude. I would start a project, and I would build a skill that is a CV formatting skill, and I would then feed in your completed template. I’m sure recruiters are sitting with hundreds of completed CVs in the correct format sitting on their desktop. Upload them so it knows what you’re looking for.

Create another skill around the company branding, where you can give it the logo, you can give it your colors, those #F368 numbers that go with your company. Build another skill, and then all you’re going to say is, "I'm going to upload the raw CV. I’m going to upload my interview notes." And if you have them, and you should have them in this day and age, the transcripts from your online conversation, you’re going to upload those. Let’s take those three sources of data. Let’s have a look at what the output format is, which is the company CV, and let’s write a CV, and let’s see what it looks like. Yes, you’re always going to have to go and cast your eye over it and do some editing. But it’s a lot quicker than sitting there copying and pasting, and it’s a lot more accurate because the human does forget.

Ksenia Izotova: Speaking of Claude and Claude Skills, there is something I wanted to ask you about. When a recruitment professional is choosing whether to use a Claude Skill or a ChatGPT project or a specific tool, what are the cases when you can simply go with Claude, or what are the cases when you have to specifically use a separate tool for this particular function?

Vanessa Raath: Good question, and I’m going to answer with a rather facetious, "How long is a piece of string?" Every case is unique. I want to give you a broad answer. What I found is that we’re now getting to the point where a lot of recruiters are paying for tools, but they’re still not getting the most out of the frontier models — your Claudes, your ChatGPTs, your Geminis, those kinds of models. Because they’re not pushing them enough. They’re not building custom GPTs and skills and utilizing projects and connecting apps and all of those things.

My advice would be, push your LLMs as much as possible. Your entry point is 20 US dollars. As a South African, I can afford 20 US dollars on two LLMs myself. It’s not going to be that expensive. Pay for it, and then do some training on how you’re going to get the most out of it. If those LLMs can’t do what you want them to do, then go and pay for a tool.

If I’m then talking to maybe an internal talent acquisition team, they have big budgets. They’ve got deep pockets. Then I would say, "Look, you’re going to be doing so many of these. You’re putting out 50 CVs a day because you’ve got teams sitting all across Europe. Go and pay for a tool. Get them to make it for you, and then give everyone access because that’s going to be your quickest way to do it."

So it all depends on the size of the business, what is the appetite to go and spend money on tools, what is the use case, and then the key thing is how repeatable is this use case? How many times does it need to be performed? And that would actually help me to make the decision.

Timur Novopashin: Sometimes what we’re missing is, of course, the GDPR part. Because if you’re talking about very large enterprises, big businesses, they usually have got metrics of different tools that are integrated with each other. They work in balance. And if you would like to introduce a new solution or a new tool, that’s a lot of work. That’s like a project.

Vanessa Raath: Yes.

Timur Novopashin: So there are different tech settings and different solutions that can be right for different companies and purposes. That’s also sometimes what the decision-makers are missing. They’re like, "Oh, everyone is using Claude Cowork. Let’s start doing that." Or, "Everyone is starting to implement this ATS. Let’s do that as well." Sometimes it’s wrong. We’ve got to take a step back and assess the situation first.

One of the vivid memories that I have from my agency days, the first one was doing this copy-pasting job when you need to introduce a candidate and you need to somehow make a CV. And the second memory is sourcing. I hear opinions that sourcing is dying. Now AI can do everything, so we don’t need hands-on sourcers anymore. What’s your take on this statement?

Vanessa Raath: November 2022, OpenAI launches ChatGPT. Us mere mortals get access to it. We can now use AI. Yay, exciting. Okay. Then we have a big summer holiday here in the Southern Hemisphere. I kind of ignored all the noise and all of those things, and I dived into it in January of the following year. And I went and I said, "Write a Boolean search string," "Help me to do some sourcing," and it couldn’t do it. I thought, "We're safe. This AI’s pretty rubbish. I don’t know what everyone’s going on about. It can’t do what I want it to do."

Fast forward three and a half years, it’s not so rubbish. I did have to eat my words. But what I’m trying to highlight here is that at that point in my career as a talent sourcing trainer, I thought, "That's it. My role is safe." Now, when I introduce myself, I said I’m an AI strategist and I’m an AI trainer for recruiters. Why? Because AI’s got so good, and it can do the talent sourcing.

But another interesting fact that I want to add to that, this year, 2026, we’re in July, I have not delivered as much talent sourcing training in the last two years — 2025 and 2024 combined — as I have in 2026. Which tells me something: a lot of people think that AI is doing all the talent sourcing. But if you’re not writing the correct prompts, if you’re not spending quite a bit of money on tokens, if you don’t understand how the searches are happening in the background, you’re still not getting the results you’re looking for.

So I’m back to, okay, let me teach you how to write a Boolean search string. Let’s go back to X-raying. Let’s look at GitHub. Let’s figure out how to find email addresses. Let’s look at Stack Overflow, Kaggle, Reddits, Meetup groups, which for me is super, super interesting.

Yes, AI’s great if you know how to use it correctly. It can’t do all of the talent sourcing, but it can be really helpful for things like doing market mapping, research on competitor intelligence of businesses, research on the company. It can be really helpful for that. But again, can it do everything? Should it do everything?

Ksenia Izotova: Vanessa, that’s a really interesting journey from being a talent acquisition trainer to literally being an AI trainer. What was the first training that you delivered on AI, and when was that?

Vanessa Raath: It must have been in 2023. Back then it was only ChatGPT. Dario hadn’t broken away yet from OpenAI and formed Anthropic to make Claude. Everyone was trying a little bit, like Neta was quite hot on the scene at that point. I started playing around more with ChatGPT, and I remember my first training actually I did it for a local South African company here.

Because there was new training that I’d put together, and they were amazed when I could send an email to a candidate and I could change it from English to another language. Because we have 11 official languages here in South Africa. "Wow, look at that." And then doing something else, taking a job description and saying, "Please write a job description for a data scientist." And next second, so fast, it typed so quickly, and then there was the job description. Really, really basic things.

I must admit that I look back at that training now and it just shows me how advanced AI has become, which is brilliant because it’s so much more helpful than what it was back then.

Ksenia Izotova: That’s very curious that indeed, in 2023, it was ChatGPT that everyone was talking about. As an AI expert, as an AI trainer, what tasks are good for Claude and Claude Skills, and which ones would you delegate to ChatGPT and GPT Projects?

Vanessa Raath: My advice to someone who’s never paid for any of these frontier models, LLMs, would be: if you’re going to pay for them, pay for ChatGPT, because it does the most end to end. What I mean by that is that it creates images, you can train it by building custom GPTs, you can use projects. Codex is unbelievable. Codex is the same as Claude Cowork. It’s got good integration into your apps. It’s probably the most all-round solution.

If you do a lot of writing, like me, I write that in conjunction with Claude. Claude, for me, is much better at writing. It’s much better at figuring out complex tasks. If you’re just starting out, ChatGPT is what I recommend, and then go onto Claude once you have kind of mastered using ChatGPT. You don’t have to just stick to one. I use a combination of ChatGPT, I use Gemini for my image creation, I use Copilot within my email because that’s how I answer all of my emails.

I think you’ve got to test out a whole lot and figure out what is the best use case for what you do, and which one you can use and get the most out of. The nice thing for now is that they’re all pretty good to use, even on the free version. You don’t have to go and pay for all of them. Max out your usage on the free version and see which one you like.

The easiest way to do that is take a prompt — but take a decent prompt — and pop it into different LLMs and then see what results come back and which one you prefer. Do one to create an image, do one to create an email, a cold outreach email to a candidate, do another one to your use case that we were talking about earlier to build a CV. And then see which one you prefer.

Timur Novopashin: If you look at recruiting, there are two sides, usually — candidates and companies. The candidates are also using AI. They use it to enhance their CVs so they can pass through these infamous ATS filters. But also I see some skepticism from the candidate side that we do not want to do AI-only AI interviews, because right now that’s the case. Some companies are already doing the first stage of screening AI-only. It’s not a human talking to a human. It’s an AI chatbot or any other type of AI solution talking to a candidate.

So if we switch on to the candidate side, have you met or have you encountered a very negative reaction from candidates towards recruiters using AI? What are your thoughts around this topic, a kind of skepticism from the candidate side?

Vanessa Raath: It’s actually really interesting because I’ve read a lot of case studies of candidates looking for jobs who prefer to be interviewed, the first interview, by an AI voice agent, because they can schedule it for whatever time. If they want to do it at 10:00 when the family’s asleep and the house is quiet, they can. There’s a huge sector of our population that works the night shift. They don’t want to be faffing around with interviews during the day. And also it cuts down the small talk. You get on, you answer the questions, you can ask questions. You don’t have to be overly pleasant. You don’t have to make small talk before you get started.

But for me, my feeling around this is that it comes down to the type of candidate, and also it comes down to what — the key thing would be the level. I don’t want to sound like I’m being snobby or arrogant here, but let’s think about grad recruitment. You get loads of people applying for grad roles, people just out of university. I think that those people realize that, so they are quite willing — and it’s not just an age factor — they’re quite willing to do that because they realize the volumes that are coming.

They would be quite happy to do an AI voice interview. But there’s no way if I was recruiting and I was looking for a new COO of a big company that I would send an AI voice bot after it. I think it all depends on your time and your place, and has the person applied for the job or are you trying to headhunt them? If someone’s applied, they’ve shown interest already, and I don’t see why a voice agent would be a problem. I think it really does come down to the individual, which is why it actually is quite a difficult situation to answer because there’s pros and cons to both.

I haven’t met anyone who said, "Well, I’m not doing that interview because I don’t want to be interviewed by an AI voice agent." We’re sitting in a global economic position at the moment. We have two massive wars that are raging. People are looking for work. Can they afford to just say, "I'm not going to do it"? It’s just not happening. They’ll complain about it, but they’re not going to turn around and say, "No, I refuse."

Ksenia Izotova: What’s your personal opinion on AI screening of CVs? Last year I took part in an acceleration program that took place in Saudi Arabia, and there our team was selected as a pre-investment committee, and we were working very hard on that. That was the project related to an AI screening tool that would select the CVs. You basically know what that is. So what’s your take on AI screening? Does it really always work, or… because I know that a lot of HR professionals are skeptical about it.

Vanessa Raath: I don’t think a recruiter is foolproof there, too. Something could slip through the cracks. How many times have you pulled someone in as a recruiter for an interview, their CV actually wasn’t quite right, but there was just something that spoke to you? It was that gut feel that we have as human beings. But to answer the question, it depends on the tool.

I’ve met with enough tools to be able to ask the question: your screening tool, how does it work? Is it just a keyword matching exercise? Is it just looking for "I want a Java developer, they must have the word Java on their CV"? Or is it looking at the companies that they have worked for? Does it go beyond the CV? Does it go and do an internet search on the person? Because candidates, even with using ChatGPT and other LLMs, can put together some quite rubbish CVs. They forget to add in this course which they did. They don’t always represent themselves that well.

Is a CV the best way for someone to get a foot in the door and have an interview? Or should we be doing something else? Asking candidates to put together a small video introduction about themselves and talk about their careers, and make a call that way. I think that the whole industry needs to be looked at differently, and maybe we adjust the way that we think of things.

Like anything in life, there are some tools that do this better than others. But there are some which are just keyword matching, and I don’t think that they’re going to be that effective. You’ve really got to research your tools. If you are going to use an AI screening tool — which I understand why you would, because there are candidates out there who are flooding the marketplace with their CVs using tools like JoinHandshake, AI tools, so you’re getting hundreds and hundreds of applications per job — I understand why TA professionals and recruiters would want to use these tools.

Make sure that when you test these things out, you test it out on a sample case that you’ve had three or four recruiters do the same thing on. Get the recruiters to surface, "Okay, these would be my top five candidates," and then see what the AI one comes up with. Never buy something and tie yourself in for the next year or two without actually testing it against a human capability.

Timur Novopashin: I think every TA professional has faced this situation when they need to go and headhunt someone, and usually there are no CVs. Sometimes there are even no LinkedIn profiles. You just know that this person worked for a specific company, and you’re like, "Okay, I’m going to reach out. We’re going to hop on the call and discuss the position. I’m going to offer this position to this candidate." So it happens a lot because there are a lot of candidates that are not actually actively searching, and we need to headhunt them. No CVs, no fancy keywords, summaries, ATS optimization. You just do what the company needs you to do, and you’re being headhunted.

Vanessa Raath: Exactly.

Timur Novopashin: It happens a lot with software engineers working for frontier AI companies. The minute you put OpenAI, Anthropic, Mistral AI on your LinkedIn or somewhere, immediately you’re being hunted and headhunters come.

Vanessa Raath: Yes.

Timur Novopashin: Usually they don’t do CVs. So that’s also another side of this process.

Vanessa Raath: And then how you use AI would be determined by which position you are in. Are you being flooded with applications, or are you going out to headhunt a very specific individual for the role?

Timur Novopashin: But now, let’s say in the European Union, there is a very specific agenda. The governments or European Union, they are trying to regulate usage of AI. Basically it comes down to whether the final decision — reject or accept, move forward, make an offer, etc. — are supposed to be done by humans. The human’s supposed to push the button. Human oversight. It sparks a lot of conversations around risks of using AI tools. These models are trained on real live data. And in the US, there are lawsuits happening right now. So what’s your take on the compliance part and on this kind of human oversight?

Vanessa Raath: The EU AI Act keeps on changing and morphing a huge amount. I was listening to a podcast earlier this week, and they were talking about how they want to push out the dates of one of the stipulations, which was how AI was used in the recruitment process, because it hasn’t been hard and fast. It’s still a very, very much a gray area. It wasn’t just pushed out by a couple of months. It was literally pushed out, I think it was from October this year until December next year. So it was a big push out.

My thinking is that human oversight is great, but I still think from a TA and a recruiter perspective, that’s always a bit of a worry. Where you had maybe teams of 20 TA professionals sitting within an organization, if you need human oversight by one person, what happens to the rest of the team? That for me is also something to think about: how many people are you actually going to need in these TA teams?

It is a fine balancing act because another thing that hasn’t happened in our space is that there hasn’t been a first major court case where a company, HR or TA company, or division within a big company has broken the EU AI Act. Someone — a disgruntled candidate — has gone and taken them to court and said, and then they’ve set a precedent, and this is what it’s going to be like going forward.

My advice at the moment is to understand the EU AI Act. You can download it online, get the PDF, put it into an LLM and say, "What does this mean for me? These are my processes. Am I in contravention of any of these stipulations here?" And see how you go. I think that it’s still too early to tell.

Timur Novopashin: They definitely pushed the implementation of the first and second phases of the act to 2027, and those phases are really affecting high-risk tools. Everything that we use in recruiting in HR is labeled considered high risk. 

Vanessa Raath: But it should be like that. I’m glad that these guardrails are in place.

Ksenia Izotova: Let’s switch from finding people to people finding you. Your LinkedIn profile is very impressive. You take part in so many different events. Your posts are very friendly, warm, I should say. Can you elaborate a bit on building your personal brand? Because obviously when you started your own company in 2019, this is when you started building your personal brand.

Vanessa Raath: I started my own personal and company brand in 2019 when I started my own business, but I built a brand for 10 years before that. I think that that’s the key difference here. A lot of people wait until they are retrenched or they’re starting their own business to start working on their brand, and then it’s a little bit of catch-up because it takes a long time to build a brand.

I do post a lot on LinkedIn. I try to post three to four times a week. Do I wake up every morning and leap out of bed and say, "Yay, I’m getting up. I have to post on LinkedIn"? No. I sit there and I sometimes think, "What the hell am I going to post on LinkedIn today? I don’t really know what to say," because I’m not doing anything that’s particularly exciting. Yesterday I didn’t post because I was reworking my website. So I was like, "What am I going to say about this?"

But let’s look at the flip side. If I didn’t have such a good worked LinkedIn profile, and I didn’t post regularly, and I didn’t have an opinion, would I then be sitting on the phone cold-calling people and saying, "Hey, can I do some talent sourcing training with your TA team," or, "Can I come and do some AI training with your team?"

I think that answers the question. I’m not an inherently salesy, let’s go and sell, sell, sell person. I would rather make a presence on LinkedIn, be what I refer to as the SME — which isn’t a subject matter expert because I don’t believe anyone can be a subject matter expert in AI — I would be the subject matter enthusiast. Post and share as much as possible of what I know. That’s why, this morning, my newsletter went out, sharing about business development using AI. That, for me, is what will help with people’s attraction.

I want to be as busy as possible. I want to work with companies to work on their AI strategy.
I want to deliver training. For me, it’s a no-brainer to build a brand. But it’s not just because I own my own company. When I was internal and I was head of TA, I had an amazing brand, so that when anyone walked into my office and said, "Vanessa, I need an X, Y, and Z person," or, "I need this," I already knew who to contact. And if I didn’t know who to contact, I would have just had to do one post and the right person would pop up.

That’s what I try to explain to recruiters through my training all the time: it’s not for now. You want to build a brand so that it makes your recruitment life a lot easier in the future.

Ksenia Izotova: Do you specifically work on AI visibility so that the other companies would ChatGPT you?

Vanessa Raath: We’ve all worried about SEO for many years. Search engine optimization. Now people are talking about GEO and AEO, which is to do with artificial intelligence. Yes, the structure of my blogs has completely changed. The way I structure them is different because I work with Claude. I send Claude out to the internet and I say, "Go and do some research about what a blog should look like to be optimized for AEO and GEO so that when I post them, people will find them." It’s a completely different ballgame. I focus on that a lot, and I have probably weekly someone pop up from weird places like Brazil, US, New Zealand.

I asked ChatGPT who I should speak to about doing AI for recruiter training, and your name came up. I’m like, okay, I can’t really argue against that. Yes, let’s chat, let’s do some training. It’s clearly working, and I think that this is the future. Your SEO still has to be good, but you’ve got to make sure that the AEO and GEO are just as good.

Ksenia Izotova: It’s now a hot question for everyone because I can share my experience. Yesterday I was brainstorming with Claude again on AI visibility. We do corporate language learning, and obviously it’s a thing for us. We need to be visible on ChatGPT, Perplexity, and everywhere. And Claude said, "No, LinkedIn doesn’t work on AI visibility. It’s your visibility for a LinkedIn audience. You should go on Reddit." There are so many different points of view. Some people say LinkedIn is crucial, so keep posting LinkedIn newsletters with the needed keywords.

Vanessa Raath: You know what? You are so right. I think it’s about LinkedIn and everywhere else. Let me give you an example of what I do. I write the blog that goes out on my website. That blog then gets made shorter, and it becomes my LinkedIn newsletter. I then can take, for example, here we are recording this podcast. With recording this podcast, I’m going to take the podcast, I’m going to cut it up using AI, and I’m going to put it on my TikTok channel and my YouTube Shorts.

It’s about being smart. You produce content all the time. Where do you put that content? I really feel that’s what we’ve got to get better at. LinkedIn has a page that sits in front of it, it’s got a robots. txt file, which means that outside LLMs like Claude and ChatGPT can’t go into your profile and read it because a lot of that is behind a LinkedIn Recruiter license or something.

You need LinkedIn because when people go and look and they say, "Okay, who’s Vanessa Raath?" and they go to my LinkedIn, then they see that I have a newsletter. They see the readership. The newsletter is slightly different for LLMs because the LLMs can access the newsletter, so they might not be able to access individual profiles. So it’s literally just about being in as many places as possible and being smart with your data.

For example, the transcript of this podcast over here, you guys should be taking and building it into a LinkedIn newsletter because not everyone wants to sit and listen to us talk for an hour. Some people might want to read that information because some people take in that information by reading. You should take these conversations and that could be a blog in itself. The newsletter that I posted this morning was from two sessions I did on an AI for recruitment webinar that I did about two weeks ago that I had recorded.

There’s power in conversations because you’re just repurposing data using AI, which is very quick to do these days. But the data’s still yours. It’s us three having a conversation.

Ksenia Izotova: Timur, are you making notes?

Timur Novopashin: No, of course I don’t do that on podcasts, but I’ve just decided to give it a go. So I just went to ChatGPT, and I asked to find experts in international talent acquisition living in Europe, and my name popped up. So apparently I’m doing something right.

One question that I want to ask you, and our audience would love to hear the answer. Let’s say if you are an early career professional starting in TA, sourcing, HR, and you would love to explore the AI capabilities, how you can augment your skills, your jobs, your work setup with AI. What would you recommend? What tools or what would you recommend? Maybe take a course or follow a specific speaker or public figure? What would be your recommendation for an early career professional starting in TA?

Vanessa Raath: Early career professional in HR and recruitment, there’s lots and lots of free content out there. Go and find someone like myself who puts themselves forward as an AI strategist. I’ll always post when I am going to be on a webinar. I’m going to share a link to this podcast. Tomorrow I’m doing a LinkedIn Live. I know I have one the week after. You need to be involved in as many of these as possible.

Something else to look at is Recruiting Brain Food, Hung Lee. He does a show every Friday. It’s often AI-focused these days. That’s a great place to go and learn. There’s no excuse at the moment for not keeping up. For me, the early career people, you’ve got to get really up to date with AI because I think that they could be the most impacted.

You need to find out what AI tools your company allows and then get really good at using them. A lot of big companies have Copilot, and Copilot’s actually really good now because it’s got Claude and ChatGPT underneath it. Go and educate yourself. Go and spend some hours on the University of YouTube. Learn how to use Copilot really well and become a master at it because it’s going to make your life easier at the end of the day, and you are going to be able to show other people in your team how to use it. It’s not going anywhere. It’s not going to disappear. You may as well embrace it and use it to the best of your ability.

Ksenia Izotova: So what do we promise? We promise a list of tools.

Vanessa Raath: I’ve also built an AI assessment tool, so I’m happy to share the link with you too. Recruiters can go and do an assessment and see where they are compared to other people in the world in their AI journey, and what they have to get better at. So that’s something that you built with Claude? That was my first ad. So I’m happy to share both links with you.

Ksenia Izotova: I have so many questions, but let’s put a dot here. Let’s park and let the listeners enjoy.