How Recruiters Go From Dabbling With AI to Real Results, with Vanessa Raath

0:00 If you look at the recruitment process, what is your bottleneck? What is slowing down your time to hire? What always comes up is the raw CV, getting that into an agency format. When you can do all of that in under five minutes, some people are saving 24 hours, 48 hours. How did you end up becoming one of the world's leading trainers when it comes to AI for recruiters?

0:23 If I look at my week, I think I've trained to every continent this week. It uses it as a real assistant. When you cross that bridge, you suddenly see people saying, wow, now I know what all the fuss about AI is. Can you explain what you mean by creating a skill? I need to do this three or four times a day.

0:42 Let's build a skill around it. I don't even have to give it instructions. I'm not repeating the prompt. It knows the process. The best result, hands down, came from Claw. What are some of the best use cases that you believe for recruiters specifically in order to go from chatting with an LLM to actually doing productive work with it?

1:05 Fantastic. Last time we caught up was at the NPA conference in Dublin. Loved your presentation. And that's when I knew I have to get Vanessa Roth on the podcast. So thank you for doing this. Awesome. And I can say the same for you. I loved your presentation and actually had some great takeaways myself, which I have implemented.

1:28 So thank you very much. Fantastic. Do you mind sharing what's the number one thing you implemented? I think what came across strongly in your presentation was don't just assume things when it comes to sales. You know, be more intentional about it. And that was the first thing. And the second thing was the follow-ups.

1:44 I don't think that I was that concerned about the follow-ups. And that was something that I was like, I must actually be more intentional about this because I'm missing out on opportunities because it might have been like a miscommunication or someone just got busy. And I've actually been, should I say, reinvigorating a lot more business opportunities by doing that.

2:02 So thank you. Oh, fantastic. I'm delighted about that. That's cool. So how did you end up becoming one of the world's leading trainers for when it comes to AI for recruiters? Sure. Good question. I like to, and that's a huge accolade. I'm quite humbled by being one of the world's leaders in AI trainers.

2:25 I think it was before that, though, to be honest with you, Mark, because I was actually well known as globally, thanks to COVID, as the go-to talent sourcing trainer in the recruitment space. Right. Then when OpenAI released ChatGPT in November 2022, and I quickly, within six months, realized that ChatGPT is going to be doing a lot of the talent sourcing for us, I needed to pivot.

2:51 That's when I started dabbling myself in a lot more AI, and I figured out, hey, I can train. I've already got a good name in the industry. I'm just going to start talking about AI. It was probably the best move because it's kept me incredibly busy over the last couple of years. Let's rewind then. So you were already very well known globally for training on sourcing, right?

3:12 Yes, yes. And then you perceived, okay, AI is going to disrupt this and how can I pivot in a way that – allows me to leverage the network and the training experience I already have, but in a new direction. What was it about AI that captivated you? Because when I saw your presentation, I could tell you absolutely geek out on it and you've really gone deep on it.

3:44 So what was it that you saw that made you think, I am going to master this? Sure. Well, first of all, don't blow my cover about being a geek, but you're right. I really and truly do. So, I mean, let's face it. When it first came out that first year and you gave it a prompt and it was like, well, I'm sorry, I can't assist you with that.

4:04 I've only been trained up till April 2020. It was rubbish. I mean, I couldn't write a Boolean string. But I think the one thing that captivated me and encouraged me to keep going is that I have ADHD and I have what I call the passionate curiosity. So anytime anything is new, and let's face it, AI is new from a global perspective.

4:25 I mean, it's the biggest new we've ever experienced in our lifetime. I figured out it's going to change everything. Let me see what the impact that's going to be in the recruiting space. Okay. So I figured out if I could get my head around it, I'd have got a big enough audience that would listen to me if there were any warning bells or anything people needed to do to stay ahead of the curve.

4:46 And the more that I got deeper and deeper into it, I realized, hey, this is learning. It's got this continuous learning function. It can be amazing. We just need to spend more time with it. And that's when I realized, hold on, there's some great user cases for the recruitment industry with AR. And I never actually turned my back on talent sourcing.

5:07 I mean, talent sourcing is still a thing. If I had solely relied on talent sourcing training, let's say through 2024 and 2025, I'd have a very hungry family right now because I didn't do a lot of it. But in 2026, I'm actually delivering 50-50 talent sourcing again, talent sourcing training, and AI for recruiter training, which I actually find really interesting.

5:30 Oh, that's cool. All right. So a few things to unpack from what you just said. And first of all, I'm sure this is what you meant, is that AI isn't really new. It's generative AI that's new specifically, right? Absolutely. Whereas traditional AI has been around like... has been developed since the fifties, you said you saw lots of use cases.

5:49 We're going to, we're going to dive into those. And you also said like, because you had that curiosity, you just naturally wanted to spend time to figure it out. So on that, on the time side of things, like how long do you think it takes to really become fluent with AI to the point that you're actually, um, you know, making an impact and achieving business outcomes by leveraging AI rather than just dabbling with it.

6:22 That's a really good question. And I could probably say with guidance, you could get up to speed in the morning. Because for me, it actually, the only thing holding me back at the moment is my token usage. So I'm sure that you realize that we have this five-hour token window. So I now pay for ChatGPT and Claude.

6:42 I found that I very rarely max my token window on ChatGPT because they're really generous. We're clawed our max probably two or three times a day. Okay. And it's literally how much you can push it before that token window runs out. But I take it back to we don't know what we don't know. And I find with the teams that I train, I was actually working with a team out of Canada last night, is that they thought they were using it quite well.

7:13 They all had the paid for version. But if you don't know what it can do, then you just don't know. And I think it takes, I run an hour and a half or two hour training workshop, depending on the size of the team. And once we start talking about things like building skills, MD files, building projects, All of those things just, I can see the light bulb moment happening.

7:40 So once you can show people what it can do, it is such a marked difference. Then the ball's kind of in their court. And then it's up to people to go and say, right, I've seen what it can do. I have Vanessa's notes and workflows next to me. I now need to go and do this for myself. And that is the key thing.

7:59 If people take the time to do that, they're going to be up to speed very quickly. Let's then captivate people's imaginations like, You said once they find out what it can do, then they'll be off and running and really pushing the model every day. So what are some of the best use cases that you believe for recruiters specifically in order to go from dabbling with chatting with an LLM to actually doing productive work with it?

8:38 Two spring to mind straight away that have really come to light over the last couple of months. The first one is that every recruitment agency that I chat with have got, we talk about bottlenecks. So if you look at the recruitment process, what is your bottleneck? What is slowing down your time to hire?

8:55 And what always comes up is getting that CV, the raw CV, getting the notes from the interview and getting that into an agency format, which can then be sent off to a client. Now, that, in my opinion, is a repeatable task. So anything that is a repeatable task, I encourage teams to go and build a skill around that.

9:17 So, for example, let's build a skill. We'll train the generative AI. This is what the final product looks like when the CV is ready to leave the agency. This is what needs to happen in order to get to that final product. So we're not repeating the prompt. We're not waiting for someone to type up a CV.

9:38 We can take different sources of information, someone's PDF of their LinkedIn profile, the raw CV, the interview notes, the transcripts of the conversation. And we can create not only a CV, but as well as an executive summary. And that is a complete game changer because when you can do all of that in under five minutes and have the final version that you want to send, some people are saving 24 hours, 48 hours, and that's a complete win.

10:06 So that would be the first example. Can we just double click on that one for a second? Because I agree. That's like the low hanging fruit here. That's a tremendous use case because, well, for a couple of reasons, number one, no one really actually enjoys doing those, right? And is very time consuming.

10:24 So there's Two great reasons to let AI handle that, right? Something you don't enjoy and something that's time consuming. And as you said, it's a repeatable, like it happens. There is a repeatable process, you know, that you go through every single time. But can you explain what you mean by creating a skill?

10:43 Because I think that will be new to a lot of people who may be already doing part of this. They may be taking, for example, they're recording the... client intake call, they're recording the candidate interviews, they're getting the transcriptions, they're maybe having, you know, Claude or Chat GPT create a summary and then sending it off where they're going in and uploading, for example, the transcript and saying, please create an executive summary for me based on this CV and this transcript.

11:14 But you're taking it much further than that when you talk about creating a skill. Could you explain what that means and what it involves? So I also want to add in a caveat to that, and that is please never take someone's raw CV with things like their social security number, their phone number, their home address, and put it into any generative AI.

11:36 So you need to redact that information first, and then you can put it in. And you also need to have your candidates aware that you use AI in your recruitment process and how you use it. I mean, this is to be in line with the EU AI Act. And even if you're not in the EU, those are very good guidelines because every country is going to have rules and governance in place around AI.

12:00 And let's start getting it right from the outset. So for me, if I needed to build that, and I've actually done this exercise, I did it with the team I was working with recently. We built the repeatable skill using CoPilot, ChatGPT, and Claude. Now, out of the three, they all have different terminology.

12:21 So in Copilot, they call it building an agent. In ChatGPT, they call it building a custom GPT. And in Claude, they call it building a skill. Now, the best result, hands down, came from Claude. And that is why I pay for Claude, because it's fantastic at generating content. It's great at putting that in a Word document, which I can quickly edit and then PDF and send off to a client.

12:47 So how you would do that? is you could either actually just do it on Claude and go through the process, go through the motions, and then say, right, I need to do this three or four times a day. Let's build a skill around it. Or what you could be is more intentional. You could go to the customize setting on your Claude, and you can say, build a skill on Claude.

13:12 I like to build a skill with Claude because then I can interact with it and it forms the instructions and everything that I need. And I say, this is a repeatable task that I do day to day. I take a raw CV. I take a transcript from an interview. I sometimes have a PDF of a LinkedIn profile, and I want to combine all of these together into a CV format that looks like this with an external executive summary that I want to put into an email.

13:43 And once you do that a couple of times and you go through a couple of iterations and it might make a mistake in one of them, you can go back and correct it. But then basically the skill is built. Because the next time that I have all of that information, I have the raw CV, I have the transcript or the interview, I literally need to go to Claude.

14:04 I need to do backslash to open up my skills. I can choose the CV formatter skill or whatever I've called it. I can upload the raw CV. I can upload the transcript. And I don't even have to give it instructions. It knows what it needs to do. So then again, I'm not repeating the prompt. It knows the process.

14:24 And that for me is a complete game changer for recruiters, is when you identify those repeatable tasks that you can build a skill around. So to give you an example, I have one that builds a blog that is SEO and GEO enabled, but it knows to ask me lots of questions so that I can give my own views and my own stories into the blog.

14:48 I have another one that is my voice, so it knows how I sound because I have fed in loads of transcripts of things like this, being a guest on a podcast or, you know, meetings that I'm in, all the way that I write my LinkedIn posts. And all of these are incredibly handy because suddenly... So the recruiter moves from using generative AR as a glorified Google backwards and forwards, but uses it as a real assistant.

15:17 And that is a game changer. When you cross that bridge, you suddenly see people saying, wow, now I know what all the fuss about AR is. You've given some really great examples there. Yeah. Let's back up a second. What are the basic building blocks or, and that could be tools or it could be the other things that someone needs in place in order to be able to do what you've just described?

15:44 That's the thing, Mark, is I don't think people realize is that AI isn't here to cost them money. I mean, all of these generative AI tools literally cost a princely sum of 20 US dollars. I mean, if me as a South African can afford to pay for two, it's really not that expensive. So all you need is a subscription.

16:03 And the reason why you want a subscription to Claude, if you're building your skills in Claude, is because you get usage of better models and you get more tokens so you can use it more before it pops up and says, actually, please, can you come back on Tuesday at 8 a.m. when you have more tokens that you can use, which is my pet hate.

16:23 Um, so that's pretty much all you need. The one other thing that I would recommend that I'll, you know, a lot of people are using them, but are they using them to their full capability is something that records all of your meetings. So whether they're a Teams meeting, a Zoom meeting, a phone call, a cell phone call, whatever it might be, because that information is gold.

16:46 Right? I can take a lot of the content that I say in presenting a podcast, having a conversation like this could be my next blog. Because what we've been doing a lot without recording this content is we've just been repeating ourselves. And kind of generating content twice, sometimes three, sometimes four times.

17:07 So Mark, to answer your question, when it comes to tools, the time has come. People should be paying for at least one large language model. Kind of like pick your weapon, you know, whichever one's going to work best for you. And then maybe something that is going to record your conversations. And I think that, you know, I'm training a lot of teams at the moment.

17:30 A lot of them are realizing that they're paying for tools, which maybe they don't actually need to pay for because generative AI is becoming so capable these days, but they need to learn how to use it correctly in order to replace the tools. So I don't want people to sit back and think, oh, AI is just going to cost me a fortune because it's not.

17:50 Follow up on your use case of preparing CVs and executive summaries to be able to submit to jobs. Would you recommend setting up connectors so that you don't even have to upload the transcripts and the resumes, but Claude can just go and grab them for you? Yes, absolutely. So for me, Not really connectors.

18:13 What I would do is I would use plugins. And this would be for advanced users, Mark. So this is going to confuse a lot of people when I talk about this. So there's a great connector called Xer, which is a website, which is fantastic for people information. And what it is, it's a website that AI agents can go to to find data.

18:34 Now, when AI agents go to find data on individuals, it actually reads code better than English. So EXA has got coded information and loads of it that when you put that plugin either into your chat GPT or your Claude, it actually brings back a better set of results. That was actually the second example I was going to share with you about what I'm seeing in the markets of, you know, So I will get onto that.

19:02 So yes, and then also they can build projects where they could say, when I upload into this folder and I upload a CV and a transcript, you can go to this particular folder and you can do what you need to do to format a CV. With that, I find that I just ended up with loads of folders all over my laptop.

19:24 And I was like, you know what? It's actually easier. It's two files. I can drag and drop them into whichever LLM I'm using. That is fine. And then I can pull the final copy. And that's what I'm going to save somewhere on my desktop. But yes, plugins are amazing. I mean, you can plug in your TripAdvisor.

19:44 So, you know, with us leaving for Namibia on Monday is that I've used a lot of my TripAdvisor, but through ChatGPT, so I can have a conversation with it and say, right, we're staying here for two nights. What should we be doing and seeing? which is great. Love that. That's cool. So Vanessa, I think it's exa.ai.

20:03 Does that sound right? That is perfect, yeah. So that's one of them. There is another one. I like Tavoli, but I find Tavoli is actually better for competitor data. So market intelligence, market data on other companies, more of a BD kind of tool for recruiters. And Tavali can also be added to ChatGPT or Claude as a plugin.

20:25 You just need to make sure that you go to Exa or Tavali first, you sign up there, and then you can connect it with your plugin option or your plugin option on ChatGPT or your connector option on Claude. Okay. So I'm already using Exa, but this is a new one. How do you spell that? T-A-V-I-L-Y is Tavily.

20:49 Tavily, T-A-V-I-L-Y. I don't know if it's Tavily.com or Tavily.ai, if I'm honest with you. That's all right. Yeah, it's really good. Could you explain a little more about the kinds of things that you would use Exa for versus what would you use Tavily for? So EXA, I would definitely use for candidate information.

21:10 So if I'm doing a talent sourcing exercise, so I'm looking for, let me explain, a company I've been working with the US this week, they are looking for veterinarians for a veterinary practice. So I use the EXA plugin to go and do some research about, for example, who are the veterinary surgeons, the vets in like a private practice, the veterinary assistants in, let's say, New York State.

21:40 Yeah. And Exa is great at finding people data. Then what I wanted to do for the same company was where could I go and have a look at competitors where some of these people might be currently working? And then I activated Tavoli. So when you go into ChatGP team, you'll see now it's changed because these things like to change and keep us on our toes.

22:03 And at the top of the screen, you've got your chat and your work option. Now, if you go to your work option, what you're going to notice is you're going to notice that it gives you a plug-in option underneath the text bar where you go and write and communicate with ChatGPT. So under that, you'll see it has a plug-in option.

22:23 And there you can choose your plugins. So not only do you have Exo or Tavoli, there are others, but you can also connect your Gmail, your Outlook, as I said, your TripAdvisor, your Spotify. There's lots of different plugins that you can put in. But that is where I would choose it to say, okay, I want you to use these tools to help me to get the results that I'm looking for.

22:45 Love that. That's really cool. One of our partners, Vanessa, is a company called CoRecruit. They're a transcription software for recruiters. Two Canadian guys, Ryan and Josh. Yes, I met with Josh this week. Oh, did you? Okay, cool. Small world. Now, what's cool is you can build in your templates and your formats within the tool so that when you do a call and record it, it automatically puts it in the format you want.

23:19 But the other thing they have is an MCP server so that you can connect directly with Claude. And then you can work with the transcripts from Claude, which I think is super cool. That is very cool. And actually, if I wasn't heading off to Namibia for three weeks on Monday, I would have that up and running already.

23:38 But I've set up a meeting when I get back to actually get that up and running because things like, you know, also business development cases, taking those transcripts and generating, you know, proposals for companies that you've met. Right. For me, like making training notes after I've trained people.

23:55 So I'm super excited to try out co-recruit. They're on my list of tools. Oh, you're listed. You have a famous list of tools. We'll link to that in the show notes, but what's the URL just so that people can go and find that, Vanessa? You know what? It is actually, let me quickly open it up. It's a start.me page because that was the easiest way to make it accessible to people.

24:18 So it's quite long. Yeah. So let's send people to your website instead because that's... They can find it from there, right? Yes, yes. They can find it. Or just go to Google and type in Vanessa Roth tried and tested AR tools. There you go. All right. Fantastic. It's simple. We went off on a tangent talking about transcription and then turning that into notes for submissions.

24:44 But then you also shared a couple of other use cases. What... If that's your number one repeatable process, let's go back to that sort of thought experiment where you're thinking, okay, what are the things that we're doing repeatedly that take up a lot of time and where the human isn't really adding a lot of value here?

25:05 So what would be next on the hit list? One thing that I'm finding quite amazing is that people aren't getting their prompts right. So I see a lot of people going into any LLM, whichever one is their preferred weapon of choice, and putting in a prompt that's 15 words or less. And I'm trying to say anything that's 15 words or less in a prompt is actually not enough context.

25:30 You haven't given it enough of a brief. It's going to come back with something, but it's going to be very generic. It's not what you're looking for. So my training now is heavily involved around working with the large language model to generate the prompt. So not diving straight into the task, but going to chat GPT and saying, right, here is a link to the company that I'm recruiting for, popping it into the text box.

25:58 Here is the role description that I'm recruiting for. There are a couple of non-negotiables that I'm looking for. I want to prioritize females. I want to prioritize people who live within, let's say, a 100-kilometer radius of Seattle or wherever it might be, London, wherever you are recruiting. You know, the things that don't always appear in the role description.

26:20 And I say, let's work together. And let's generate a prompt. And bearing in mind, this is all done with voice. I very rarely type into my LLMs because as recruiters, no one went into recruitment because we love admin, right? So we all are very good communicators. We pretty much over communicate. But this is our strength.

26:43 So let's go and... you know, dictate what we want to happen. And then what's going to happen is that the LLM is actually going to do the heavy lifting. It's going to give you all the potential synonyms for the job title. But also at the end of what you're putting into this prompt is to say something like, do not go ahead and just build the prompt.

27:04 Please ask any questions that you are not sure That is a great exercise because no matter how much context I give it, it always has questions and the questions are valid. So I will go back and I will answer all of the questions. And then what it does is it actually comes back with a phenomenal prompt.

27:31 It's long. I mean, it's a good couple of pages. But once I've got the prompt, it makes my life a lot easier because then even though I pay for ChatGPT or Claude, I actually take that prompt and I drop it into other LLMs. I drop it into Gemini. I drop it into Copilot. I drop it into Grok. I drop it into Perplexity.

27:50 Because the heavy lifting is done around building the prompt, I want to see, because they all search differently. They've all been trained on different data sets. I can then go and have a look and say, okay, great. Okay. Now, who's done a great job of this? And how do I combine all of the information that's come back?

28:10 And that's easy to do. You know, it's a copy of the information. Let's create one spreadsheet on the results. Let's get rid of duplicates. Let's highlight who we think is the most, you know, the person that we want to chat with first. Give me the top five. And that in itself is a really, really worthwhile exercise.

28:29 But because it maybe takes an extra 10 minutes to build the prompt, a lot of people would rather just, and I see it happening all the time, drop in the job description and just say, find people for me. And that just doesn't work. And then a lot of people say, oh, generates of AI is rubbish. It doesn't work.

28:47 And I'm like, well, how are you using it? Yeah. Who better to, or what better to help you create those prompts than the LLM itself, right? And so rather than you trying to create a prompt from scratch, having that initial context of what you're trying to achieve and then getting it to ask you questions and then getting it to write the prompt for you.

29:11 So it is all about keeping up to date with prompt frameworks and what works well and what doesn't work well. You know, when we first started doing this, let's say a couple of months ago, it was always say like, you are a senior recruiter with 20 years experience recruiting finance candidates in, you know, wherever it might be.

29:31 Now, apparently that is restricting a lot of the results. So you've got to know what does good look like. Because as part of my training is this is what a good prompt looks like. Because just because it's a lot of text doesn't always mean that the LLM has actually given you a good prompt. You need to understand what am I looking for as a good prompt.

29:53 Fair point. Yeah, totally. Absolutely. Okay, Vanessa. And what would be next on your list of things that are extremely high ROI for recruitment agencies specifically to be doing with LLMs? So we've spoken about the first one, which was the CV formatting. Then we spoke about using it for sourcing with plugins and those kind of things.

30:19 I'd say the third ROI that is a huge one is a lot of people don't realize that Microsoft's co-pilot, you actually get for free. Because you can pay for it. It's a substantial amount of money. But if you don't pay for the co-pilot add-on, you can still use it within your Outlook, your Word, your PowerPoint, your Excel.

30:43 Okay. So everyone is seeing it now. You go to a Word document, bottom right-hand corner, that little colorful co-pilot pops up. A lot of people aren't using it because they think that you have to pay for it. But the best user case that I'm finding for recruiters, because of the volume of emails that we get, is to use it in Outlook.

31:01 So, Mark, I don't know when I last typed an email. I'm not one for admin, as you're probably picking up. But, you know, we all inundated with emails. So I dictate all of my emails. And I mean, I have tried with ChatGPT to do the plugin and let it answer emails on my behalf. didn't love it. They didn't sound like me and they were very robotic.

31:23 And even though it didn't send them, it still put them in my drafts folder. I was like, nah, that's not right. Then it was answering emails that I didn't need to answer and I just got frustrated. But if you go to co-pilot and you open up, I mean, to Outlook and open up the co-pilot window pane, you can actually dictate.

31:40 So now I can read the email and while I'm reading the email, I'm answering Copilot polishes my result, maybe takes a little bit of the edge off if it needs to, makes it sound very professional, doesn't include profanities, which I must admit I don't use a lot. But what it comes up with is brilliant. And it's literally in a fraction of the time, because I'm not the world's best typer.

32:06 And I found that You know, recruiters always don't love admin and they have the best intentions, but halfway through an email, you know, they get tired and bored and this is becoming a bit of a pain. So being able to dictate and then also training an agent to sound like you, you know, it's a game changer because that's admin.

32:27 And my take on this whole AI thing is for the recruitment industry is let's let AI handle the tasks and let's human, you know, handle the trust. And we can do that by spending more time with human beings. Absolutely. And we're going to come on to that because I think that also is an interesting topic is like, what are the things that, you know, how do you, I think your expression is how do we humanize?

32:51 Am I saying that right? Yeah. It's normally my hashtag up your humane, which is. Up your humane, that's it. Yeah. Yeah. Okay, I love it. So, Vanessa, something that we've taught our clients is if they're using Claude to use Claude projects as opposed to starting from scratch and re-prompting it every time for something that's a repeatable process.

33:19 But you seem to favor skills. So in what case would you use skills versus projects? So for me... Projects are also for repeatable skills, but in a different way. So the way that I train it is for my recruiters that I'm training to have a project per client. Because basically what it does is it just has all of the information.

33:41 For me, a project is more like a data repository. So that's where for one client, I'll have all the terms and conditions. I'll have the transcripts of our meetings. I'll have the training proposal that I put together. Yeah. All of that bit of information. But when I go to prompt it, I still activate the skill because building training notes is something that I do across multiple different companies.

34:05 So I will activate the skill within the project. The project is there for me to have something to refer back to when I want all of the documents and all of the information that is related to that one particular client. Okay, that's really interesting because we use it almost the opposite way. So we use it for a repeatable process and we'll have the MD file with the project instructions and we'll have the examples of what good looks like and that kind of thing.

34:37 And then we will upload or prompt the specific like for this client or for we're writing a LinkedIn post about this. use our instructions to generate the LinkedIn post, that type of thing. Interesting. And I could see that it would work. Just for me, from getting it right in my brain, The repeatable tasks are the skills because you repeat the tasks across multiple clients.

35:01 That's the thing with this is like, there's a lot of different ways that you can use this. So is there one final use case that you find when you teach this to your clients, then that's a game changer for them? I think it's not one particular thing, but I think it's a culmination of how what it can do the more that it knows about you.

35:25 So even things like updating the personalization. I mean, it's such a simple thing. But you can get ChatGPT to help you to write the personalization or Claude or Copilot. And as soon as it knows that you're a recruiter, you work on these type of roles, the outputs are so much better. Yeah. And, you know, you brief it with, you're my assistant.

35:46 You are, you know, my EA. And this is how I want you to act and behave and answer. Then people realize, okay, wow, this is great. Because what I see a lot is people generating content. They don't have the memory enabled on their LLM. And then what they do is they take the first response that comes back, they copy it, they throw it into Word documents, and then they waste two hours editing that.

36:13 And then I'm always like, you know, would it not just be better to write this yourself from scratch instead of using so long data? But it doesn't, it's not getting what I'm saying. And it's like because you haven't trained it. That's the big issue. And once you get past that kind of barrier and people start seeing way better results, they're like, okay, now I know what Vanessa meant.

36:35 So what are the essentials? I know you could do a whole course on just this, but top two or three things that everyone must do to train the model to be personalized to them. So they must fill in the information. So if you go to the bottom left-hand corner of any generative AI that you are using, there's normally a personalization setting there.

36:57 Fill in all of the fields. Make sure your memory is enabled. And then talk to it. imagine that it is a junior member of your team. Like, for example, I actually don't normally have meetings on a Friday. I'm doing this because I wanted to be on your podcast before I disappear for three weeks. I actually said to it, okay, I've got a quiet day.

37:18 I said to my crowd, what is left that we still haven't finished? Like, are we halfway through building a skill? Have I got any agents that need updating? What are we going to do this morning? And I'm going to It said, okay, well, we were halfway through building this skill. We're halfway through this agent.

37:36 Do you want to complete it now? It asked me, where are you going traveling? I said, I'm going to Namibia. It gave me a list of, have you got your passport? If you're going up north, you need malaria medication. So you've got to almost treat it as the assistant and say, you know, over to you. You know, this is what I do.

37:54 I'm giving you enough context. How can you suggest improvements? Another thing around the personalization piece is that for people that have been using it for a while, and I mean, a while could actually just be a month or more, type into them and say, what do you need to know about me in order to sound more like I do?

38:13 Where are the gaps in your knowledge that I can fill in? And spend half an hour having a conversation with it. You know, things like that are just going to add so much value down the line because we just assume that AI knows everything, but it honestly doesn't. Something that we teach our clients, Vanessa, is there's some core files that you want, like context files that you want clawed specifically is what we typically use, but also for chat GPT.

38:46 to refer to when it's doing things for you. And one is we call it voice.md and it's your style guide that captures your tone of voice, how you communicate in different contexts. So it could be in a blog, could be in a LinkedIn post, could be in a email. So it knows your style and how you like to write and what you absolutely like don't want it to sound like.

39:08 What you do want to see and what you don't want to see. Exactly. The other one is your differentiators or your service, like talking about your service, what you offer. So like, what are the services you provide? What makes you different? And so on. The third is your ideal client avatar or ICP. So it knows who you're...

39:31 selling to. So now knows three things. It knows who you sell to, what you sell, and how you sound, right? What your tone and your taste is. And then the other one that we add in is proof.md. So that's all of your testimonials, case studies, success stories, that type of thing that it can draw from. Is there anything I'm missing?

39:57 What else would you want Claude to know about you? We're talking about the same thing because remember I spoke about building skills in Clawed, custom GPTs and chat GPT and agents and copilot. That is all of the information that is stored in MD files. So the MD file stands for markdown. So what I find is that I will build the skill in Clawed.

40:20 Because that is my go-to. So I have also, you know, Vans voice that sounds like me. But then it's very easy to download that file, put it somewhere on my desktop, and I can use that to build my new agents in Copilot. I can use that for my custom GPT in ChatGPT or my GEM in Gemini. So what we need to be doing is those three files that you mentioned were gold.

40:44 Absolutely love them. And I think that that's a fantastic idea. But people need to realize that those skills are now transferable. They don't have to go and retrain ChatGPT when they've already trained crawled. And that's the beauty of it. If they're on your desktop, you can upload them anywhere. So to think of others that you could use, what I did from the proof, what was a really worthwhile exercise, is I went onto LinkedIn and I downloaded all of my recommendations.

41:14 So that data is so easily available. I mean, I'm really blessed. I think I'm sitting on just under 600 recommendations now. Holy smokes, Vanessa. I've never met him over 600. I thought I was doing well. I've got over 100. And I was like, wow, 600? That's a crazy number. Mark, I'll be honest with you.

41:33 And I say this in my training. It's not my sparkling personality. It's because I've asked for about 2,000. So if you don't ask, you don't get. Right, absolutely. But then people still need to have seen the value in what you've provided in order for them to agree to do it. So it's a combination of being brilliant at what you do, but also remembering to ask every client for a recommendation on LinkedIn.

41:58 100%. So that is gold. So do that with your LinkedIn. I mean, you can download all of your LinkedIn information. So you can download your posts, your newsletters, your articles, whatever. And all of that can just be fed into these different skills or custom GPTs or agents in order to be able to replicate yourself.

42:18 It's almost like you're building a second brain out there. Yes. Of everything about you. You've just reminded me of something though, Vanessa, which is one of our members, Rory, shout out to Rory. I hope he hears this and he feels good that he's getting a mention on the podcast. So what he did with his LinkedIn archive is he downloaded it and he used that to get Claude to write his style guide for how he, specifically for things like LinkedIn messages.

42:49 And it analyzed all of his messages and then wrote a style guide of this is how Rory communicates on LinkedIn. It also analyzes LinkedIn posts and so on. And instead of because, you know, what's hard is if you ask someone, how do you like to write? How do you sound? How do you like prefer? How would you describe your communication style?

43:11 People are can't answer that question, right? Not on the spot. And so it's better to look at real examples of like, as you said, transcripts of conversations or examples of your actual writing. And then Claude or ChatGPT can use that to say, well, this is how I think you write. And then you can correct it rather than starting from a blank page.

43:35 And it's also great to refer to. I mean, if you are, you know, in a small team or a solo and you sending a proposal to a client or something, you know, you want to reference those recommendations. So they're good to have so that they can be referenced. Love it. So let me see. What haven't we talked about yet that you wanted to be sure that we covered?

43:59 I think we've done very well, Mark, actually. The key things I think would be AI doesn't have to cost money. And I think that there's a massive misconception globally out there is that as soon as I start starting this AI journey, it's going to cost me money. And I think that that really is something that has to change.

44:16 It's an incredibly helpful tool. Yes, if $20 a month is a lot of money for you, then, you know, it's going to cost you a lot of money. But it doesn't mean you have to go, I mean, if you look at the cost of some things like a LinkedIn recruiter license and how long you must sign up for, that's a lot of money.

44:34 Paying for an ARLLM and getting the most out of it is so small by comparison. And I think that that's a misconception that I really want to debunk out there. You know, pay for the training because you don't know what we don't know. Suddenly it's like recruiters are expected to be technical AI geniuses overnight, and they're not.

44:55 I mean, I'm lucky that I have the time on my hands to study a lot at the University of YouTube because that's the only place that is the most up-to-date experience. And X on what's happening in the world of AI and new models and how we can push them. And then I go and think of the recruiter user cases.

45:12 So I can do that. Recruiters are running a desk. They're busy. They haven't got time to figure out all of these things. Go to someone who can assist you. It's a no-brainer. You can't just be doing all of these things by yourself. A hundred percent. Vanessa, I know I've heard you talk about, you had an experience where you scaled a TA team to a really, really big team.

45:38 And what are you seeing? Well, first of all, what did you learn from that experience? But how does that apply to recruitment agencies in terms of companies that want to scale? How do you see companies growing in the age of AI? Really interesting because I scaled a team prior to AR. So we didn't have, you know, the digital team members, which you could potentially have.

46:02 And it was actually really interesting because I was listening to a podcast on my run this morning, and it was about the future of HR and recruitment is building data. digital team members that are going to do a lot of the admin. So what's going to become really important for the recruiters is what I keep harping on about is up your humane.

46:23 So a lot of my training, I'm leading with what do you look like online? You know, when people Google you, are you someone that they know who has an opinion, who's connected to other people in their network? Because AI is so good. It literally can go and it can source the candidates. It can send the outreach.

46:44 It can do the first round of interviews. It can set up next meetings at companies. Like you don't essentially need a recruiter if your company wants to go down that route and they want to fully put AI into your whole recruitment process. So what is it that you are bringing to the table as an individual, as a human being that's adding value?

47:04 And that's a really big question which recruiters need to start asking. So I started my business eight years ago, and, you know, a lot of it was – Let's build a brand. That was the first course that I ever wrote was let's build a brand as a recruiter. And it was always like a nice to have. And I trained it.

47:23 And it was probably the least asked for training where I always thought it was the most important. But now it's like it's no longer a nice to have. It's imperative. You have to have a brand as a recruiter. Because how else are you standing up? Because AR can do all of the admin that, you know, if you're good at admin, so what?

47:42 So is AR. Yeah. There's a few things I wanted to take away there. So first of all, the question, what do you look like online? And are you building a brand? Can you elaborate on why that is increasingly important? I think for awareness, because I mean, the first thing, and I mean, I do get headhunted.

48:03 I'm very blessed. I get headhunted a lot. People just don't realize that I don't think I'm employable. And I don't think I could ever work for anyone else ever again. Me neither. Yeah, exactly. You have the same position, Mark. But the first thing I'll do is when I get that message is I'm going to go and look for that person.

48:18 I'm going to Google them, look for them on Instagram, look for them on Facebook. And the number of times these days I actually can't find that person because it's actually an AI bot that's reached out to me. Sounds like a human. It's got the personalization right. But there's nothing to substantiate or back up that this person really exists.

48:36 And that's a problem for me. And it's a problem for a lot of candidates out there. So for me is that if someone's reached out to me like that and they are active on LinkedIn and they have a newsletter and they have a following and I can see them commenting and adding value in my chosen, you know, industry niche, I'll probably still go back to them and say, look, thanks, but no thanks.

48:56 This is not on my radar at the moment. But if I can't find anything about them, I just ignore them because I'm like, it's just a bot. I'm not going to, you know, hurt anyone's feelings. And that's becoming more and more prevalent. It's so interesting. I was actually talking to Matt Alder about this. Do you know Matt?

49:12 Yes, I do know Matt. I also like his podcast. Yeah, another great... I think his is the biggest podcast in recruiting and talent acquisition. He lives like about 30 minutes from me, coincidentally. Great. Yeah. So met up with him a couple of times for coffee. And then he's been on my show and I've been on his show.

49:31 So what we were talking about is that really one of the differentiators is trust, right? That... Who is the candidate going to respond to, right? Are they going to respond to the bot where there's nothing behind it, even if the email is really well-written and everything? Or are they going to respond to the person who is very visible in the market, who has got a recognized brand, 600 recommendations, right?

50:01 And is sharing, you know, posting, sharing market insight, obviously knows, you know, their space and is credible, right? Which person are you going to, or not even person, which of those kind of outreach messages are you more likely to respond to? And so building that brand is because people, especially, I didn't tell you this, Vanessa, I'm doing the Oxford University AI program at the moment.

50:31 Fantastic. And it's super interesting. And, cause it's not, it's less about the stuff we're talking about today. It's more- Like big picture, like the fundamentals about the whole history of AI and then thinking strategically about the future and how your business needs to adapt in order to continue to prosper in the future.

50:58 I think you need someone to interview you about your findings on that on your own podcast because that would be so interesting. That's a great idea. All right, cool. Thanks for the suggestion. But one of the topics that came up was how good, it was the topic of deep fakes, right? And how good the generative AI now is not just at producing text, but producing audio and video.

51:25 Yes. And images. And to the point now that you can't trust what you see, right? When I was growing up, there was expression like seeing is believing, right? Yeah. If you see it, then – and almost like you can think of like in court, right? If there's photographic evidence of someone having done something or whatever, then that is like – that's almost irrefutable – Exactly.

51:49 Evidence. But now you can't believe what you see with your own eyes. And so that is creating an environment where it's very low trust. And that's why it's even more important that you... So I want to talk about up your humane, right? Because... And I think that links to... being visible online, building that brand, being that trusted advisor, which is a term that gets thrown around a lot.

52:17 But so how does Up Your Humane fit into all of this? That's exactly what it is. I mean, Up Your Humane, for me, includes all of those things. You are a trusted voice. People interact with you. You have some followers. It's not just that You know, how do people perceive you? And that's why, I mean, yeah, 600 recommendations.

52:42 There's a reason why they're there. You know, that's 600 people who've trusted what I have done for them, you know, in the past. That speaks for itself. It's almost that kudos that you are looking for. And recruiters need to work harder in order to be able to build that. I mean, I use that as an example personally.

53:02 I live in a very small town in South Africa. I mean, I live on the coast, yet I travel all around the world speaking at conferences. And I put it down to my brand is I have an opinion. I share a lot. I comment on other people's posts. I make good connections. You know, I make sure that, you know, if I'm having 30 minutes coffee introduction with someone online, it's a good conversation.

53:26 So it's all about up your humane, being a human being, whether you are in person or whether you are online. Yeah. And that immediately makes your job as a recruiter so much easier because people already have that trust. And then you reach that, what I like to call the utopia in recruitment, where you are doing far less, I need to go and find clients, I need to go and find candidates, because it's all inbound.

53:52 And I think that once a recruiter has got to that point, they really can sit back and pat themselves on the back because all of those tasks and tasks, exercises, which they've done for years to build a brand on LinkedIn, you know, upping up your humane. And they've done that online and in person, and they've achieved it.

54:13 And that for me is where every recruiter needs to get to. Vanessa, I agree with everything you said, except for one thing I would tweak, which is I don't think we can ever sit back and feel like we've arrived. We'll sit back for five minutes and then get back to work. Exactly. Exactly. But no, I don't think you meant it that way, but I get what you're saying.

54:34 that that is what we all want, right? Is to be the person who the candidate calls before anyone knows they're on the market. They've not even updated their CV. They just want advice on their next move or the client is planning a reorganization or their succession planning or there's some project coming up and they haven't even written the job spec yet and they want your advice and benchmarking and shaping the role and everything.

55:04 And then I think that that is, I mean, it's very, you know, a lot of people battle to measure their success in recruitment. But for me, if those two things are happening organically, you really can, you know, as I say, sit back five minutes only, give yourself a pat on the back, because that's the ultimate goal, isn't it?

55:22 For sure. Something else you said, which I want to pick up on though, is it's not just about what you look like online. How many different cities have you traveled to this year to speak and attend conferences? Probably about 10. And I actually just booked another trip. I'm going to be in Australia in November, you know, to speak at a conference in Brisbane.

55:42 So, yeah. And that's the thing is transferable skills. I think that being in person is going to become like... even more valuable in the age of AI, because like we said earlier, like if I see a video, Vanessa, is it real or is it fake? Right. And, but if I see you in person, like I met you in Dublin, we were able to have a hug and shake hands and everything.

56:09 I know you're real, right. And I can have a real relationship with you. So I think being like getting in front of clients, Absolutely. Ironically, I know we're talking about AI today and how fantastic it can be if you use it correctly, but I also think that by contrast, it makes... almost the things that we've done in the past become more important, like going back to meeting up in person.

56:43 Yes, where possible. I think COVID had a big play in that. Then straight after COVID, we were smacked with November 2022. Now, suddenly, OpenAI lets ChatGPT be accessible to all. Now, everyone can access generative AI. I do think that there's a big drive to back to in-person, which is great. Love that.

57:05 But I think people need to be willing to get back to in-person and traveling again. I think there's a lot of people that are like, nah, I've moved. I'm just going to stay where I am and see how I go. I don't think it's going to end well for those people. Yeah, well, exactly. It depends, I guess, what your goal is, right?

57:23 Yes. And what stage you're at. But if your goal is growth, right? I want to reach more people, I want to have a bigger impact, I want to grow our business, I want to increase my income, then putting yourself out there online and in person is, I think, going to be important to that particular set of goals.

57:44 Yeah. I think you need a bit of a balance. I mean, it's Friday. If I look at my week, I think I've trained to every continent this week. So, you know, Australasia, yeah, Asia, Europe, you know, North America. I haven't done South America this week. I couldn't have done that if I had to fly around and deliver training in person.

58:04 I couldn't have done that. But a lot of those people that I've trained this week, I've met in person at an event somewhere around the world, and they have then followed up with me and said, okay, now I want you to train my team. So I see the travel worthwhile because that is my marketing. That is my business development.

58:22 And then let me come back home to Sleepy Hollow where we live, and I can deliver all the training. Makes sense. That sounds like the perfect balance, Vanessa. If there was one key idea, takeaway that you want people to get from this conversation, what would that be? Look, I think that we passed the point of don't be nervous of AI, just get started, go and experiment.

58:45 I think my advice would be that generative AI is actually far more powerful than you think it is. So I want you to think of an end goal and I want you to go to your LLM of choice. And hopefully that's one that you're paying for. and work alongside it. Say, I am a recruitment agency or I'm a solo recruiter.

59:08 This is what I do. This is what I want to get better at. My goal for the end of 2026 is this. How can we work together in order for us to achieve this goal? And I think you're going to be really amazed at what you're going to get out of those conversations. Love it. That's a great takeaway. Thank you so much, Vanessa.

59:28 This was a lot of fun. My pleasure. Thank you so much, Mark. Enjoy your holiday in Namibia. Thank you. Looking forward to it. All right. Talk to you again later. Take care. What stood out the most for you in this one? Let me know your biggest takeaway on LinkedIn or in the YouTube comments. And more importantly, what are the one or two things that you're going to put in practice this week?

59:52 If you're getting value from these conversations, do me a favor and hit subscribe so you never miss an episode. I really appreciate you listening. Make it a great day and I'll see you next time.