Brendan Cross-Williams Live on Shally's Alley
In this live episode recorded on January 26‑27, 2024, Shally Steckerl welcomes Brendan Cross‑Williams, a principal recruiter from Cubic Q in the UK, to discuss the evolving role of automation and generative AI in recruiting and sourcing. The conversation opens with Shally’s personal update – a recent downsizing that left her seeking new work – and a brief banter about Manchester football. The core of the show explores Shally’s long‑standing use of Evernote (where she is an Evernote Certified Consultant) and her consideration of switching to Notion. Brendan explains Notion’s multifunctional workspace, its built‑in AI (essentially a wrapper around large‑language models like ChatGPT‑4), and practical use‑cases such as creating landing pages, managing webinars, and organizing candidate outreach. They debate the hype versus reality of AI, noting that AI tools are still rule‑based assistants rather than true intelligence. A substantial segment is devoted to prompt engineering, which Brendan breaks into three maturity levels: (1) basic rewrite prompts, (2) structured prompts that incorporate context and copy‑writing techniques, and (3) advanced, multi‑stage prompts that integrate SEO, copywriting, and domain‑specific terminology (e.g., quantum engineering). They illustrate how AI‑generated job adverts can be iteratively refined, improving both engagement and candidate quality. The discussion moves to Boolean search tooling, highlighting Gabby Preston‑Piper’s Tool.Raccoons as a sophisticated, AI‑enhanced Boolean builder, and the limitations of newer “PeopleGPT” style search engines. Shally and Brendan also touch on ancillary AI tools: the Arc browser (with built‑in ChatGPT, Claude, and Perplexity), video‑editing AI (Wisecut), and image generators (Leonardo, Midjourney). They stress a pragmatic approach—invest time only when a tool demonstrably saves hours, and adopt a “try‑twice” mindset before discarding a new technology. The episode concludes with a thought‑provoking prompt for listeners: whenever you encounter a new idea or tool, consider how you can improve or customize it, and a brief discussion of potential future guests. Key takeaways: 1. Notion can serve as a versatile workspace with AI assistance, but it is still a tool that requires thoughtful prompting. 2. Prompt engineering is a skill ladder; higher‑level prompts yield better copy, SEO, and candidate engagement. 3. AI will augment—not replace—recruiters; the competitive edge lies in combining human expertise with AI efficiency. 4. Evaluate AI tools by concrete time‑savings and ROI; a disciplined “hour‑test” helps avoid wasted experimentation. 5. The evolution of sourcing mirrors past shifts (Rolodex → Google → Boolean → AI), and the next wave is sophisticated prompt‑driven automation.
Chapters
Hey everybody, welcome to Shally's Alley. You are live on Friday, January 26th at, um, not exactly 1 PM Eastern Time. 1 PM was about 20 minutes ago, but— and, uh, and they took my wings away. You can see most of you that watch the show probably expect my fighter jet background. I've been demoted to ground crew temporarily because I was downsized. The company that I was working with had to consolidate and hiring tightened up and a bunch of other things. So there was a decision for me to part ways with them. It was very amicable and still very, very grateful for the opportunity. But now I am available for hire, okay? Not for anything else. I'm not available— like, I'm happily married, two kids. No problems there, but I do have to feed them. So if you don't like me but still like kids, please get me a job so I can feed them. But anyways, I have to reconfigure all my setup and everything, and that's why I lost my, my, my jet background. I'll get it all set up next time, just in time to have a very special guest from all the way across the pond, as you would say. In Manchester. Did I say that right?
Yeah, pretty much. Yeah.
Which, if anybody plays— well, here in the United States we call it soccer, but you guys call it football. So if you're into soccer slash football, you might know Manchester has a pretty well-known football team, but even better known fans, right?
They actually have two well-known football teams, and I support none of them. So you're not a fan?
You're not one of the Hooligans that they say?
No. So you have Manchester United and obviously Man City, which are obviously arguably the bigger team these days, but obviously not with the history. They have the recent history. I am actually a Liverpool fan due to my family being from Liverpool.
Like the Beatles?
Yes. And our manager's actually just announced today that he's leaving at the end of the season, which is a bit of a shame. But it is what it is.
Well, it's, you know, it's soccer. So it is, it has to exist. Yeah. So Brandon and I, by the way, I did not introduce you. I'm sorry about that. So Brendan Cross-Williams, that's Brendan, B-R-E-N-D-A-N. I think it's on the screen there at some point. So I'm sure you guys can figure that out. But Brendan is actually a referral. So as some of you may know, at the end of the show, I always ask the question, who should be on the show next. And that's how Brendan was introduced by, uh, Stacy Broadwell and Carrie Collier. So Carrie met Brendan and introduced him to Stacy. Stacy mentioned you on the show and said, um, you know, next up you should have the, you know, just have this guy who's, uh, really good with automation and sourcing and everything. So Brendan comes to us from the UK, and he is a principal recruiter with a company called Cubic, with a Q, um, specializing in software talent, basically. Um, so, you know, we wanted to bring you on and talk a little bit about sourcing, and particularly automation, generative AI, outreach. And the one thing that I wanted to ask you about, which, if you don't mind, we can start with that.
Yeah.
So you may not know this, but I am the only person in our industry in recruiting and talent acquisition who is an Evernote Certified Consultant. Long time ago, I went through this very rigorous training program complete with like, you know, senior project and everything else. And I was one of a handful of people that actually passed the whole thing. There are a few others that were in, you know, other types of consulting, but I'm the only one in our industry. So I've been And before that, I was a huge fan of Evernote since day one. It's been a huge fan of Evernote, been a huge user. I'm literally certified to actually teach people, and even, you know, they would refer business to me if I was in that business. That's why a lot of people in recruiting don't do it, by the way, is because certified— because the only reason to do that is to actually get paid for that. But I just wanted it to, you know, for the intellectual, um, advantage, basically. So all of this to say Evernote has been my darling and my second brain, many cases my first brain. I went to an organization where they required that we use OneNote from Microsoft because it was a Microsoft shop and Evernote was not allowed. So I was able to start over with OneNote and kind of got used to it. I like it. I still use OneNote for certain things. But, um, I'm a huge fan of Evernote until recently. They got acquired and there's been some changes that I think they're trying to move too quickly. Evernote was always really careful about staying to the core of being a note-taking app with optical character recognition. You could take a picture of a street sign and it would be searchable, and you know, it's really easy to use. So I'm not saying I'm leaving Evernote. But I've heard of Notion, and I've heard of Notion from a lot of people. So convince me, convince me to leave Evernote to go to Notion.
Um, I think I never tried to get— said if I said, just take a step back, I never personally tried to get anyone to, um, leave to use a tool, um, that works necessarily just for me. Um, I think depending on what you do and who you are, and, and I think there's lots of people are very, very different in the, in the way in which they kind of learn and which they they see things.
But let's say that I'm about ready to pull the plug on Evernote and I would like to go to another tool. What is the— so how is Notion something I should consider then?
So Notion is, um, fantastic. So I actually did a live demonstration, um, on it going back. I think it was about— so I think it was in December actually at the Julian Sorce's Royal Rumble. Um, and enables you to do multiple different things. So what it is is actually quite difficult to explain. Um, it's lots of different things within kind of one platform. Um, so I use it in lots of it or have lots of different kind of use cases where I'm kind of using it. So some of the stuff that I demonstrated, um, on it previously was how I use it in terms of candidate outreach, for example. Um, so I actually use it in multiple different ways, um, in terms of I want to outreach to a candidate and I'm a big believer in terms of I can sit and talk to a candidate and kind of keep just talking effectively, um, and trying to sell an opportunity, or I can actually show someone what it looks like and almost sort of create a vision for them. And I use Notion for that in, in that way. So it allows you to first of all create things like landing pages. Um, I know there's lots of things like Socard and stuff which are very equivalent.
Okay.
It's very, very sort of seamless in the way in which it— you can use it. Now on top of that, you can actually go into even— carry on, you can ask questions.
Yes, I just want to make sure I'm— for the people that may not know Evernote or OneNote for that matter, or Notion, we're talking about what I would consider a note-taking app that is cross-platform and is essentially a kind of a workspace where you keep your stuff, like a binder back when we used to have, you know, hardcover binders or, or like a, you know, a notebook, etc. So it's not just taking notes. It's, you know, it's, it's also a folder, filing cabinet, etc. And it's very similar to OneNote where you can take pictures and all that. So are you saying that Notion isn't just a note-taking app? It's—
yes, that's absolutely right.
Okay.
Yeah, that's absolutely right. So what you're talking about, I do— I would use that, those features in Notion. So some examples is I previously used to store my, my booleans, because obviously I'm not going to rewrite them every single time into a Google Excel sheet.
And that's actually the reason I use Evernote. That's exactly it. Yeah, all my searches in there.
Yep, that's now, um, in the process of being moved into Notion. Um, you can create notes, you can categorize things, you can create whole databases, um, which interlink with, um, you know, various different things in terms outside I, I use it for, um, quite often I see webinars. I can't say I sit on a lot of webinars, um, just due to personal interest, and there's lots that I want to join and stuff like that. I use it to manage some of that stuff and kind of say, right, these are the ones I want to join and kind of choose. There's lots of books that I want to read, um, which are kind of on my list to do, and if I see a book which I think actually I want to read that, I will put it into a certain page there. Um, I use— I have a to-do list effectively, where I prioritize like my personal daily activities. And I don't personally work on a weekly or monthly type of view. I know what I need to do every single day. I know what my priorities are, what I need to, you know, every action I take as a sourcer or recruiter has to obviously generate money for the business that I work for. So, you know, it allows me to kind of prioritize what I need to do, and I can kind of work through that on a checklist. but it is so much more than that. Um, I've used it for business development. I've used it, uh, I've provided companies that I work with with a link where I have recorded whole interviews with a candidate.
Yeah, like I said, it's kind of like your binder, right? Before we had laptops and all that, you would put everything in there. If you could interview somebody and record it somehow, you would find a way to put it in there. Um, but I'm interested about one particular aspect of it that that is, uh, something I'm pretty passionate about. And that is, they claim— they claim— I don't know anything about it, so I'm not saying that they claim, and I don't think it's true. They say that they have built-in AI. Now, uh, if you've watched my show, you know that I think AI is bullshit. There is no such thing as artificial intelligence now. But they claim that they have, you know, the ability to— they have large language models that do some assisted you know, machine learning and things like that. So tell me, since I know that you are very technical, tell me what really is their version of AI. What does Notion do? Yeah, with so-called AI.
Yeah, so I, I, I am a little bit in agreement with what you said. So I, I like what AI is and the thought and the concept, but I think sometimes how you, how it's perceived versus what it actually is can be too very different.
It doesn't do things for you. It's, it's a tool, right?
So What it basically is, is it works very similar to things like ChatGPT or Bing or Bard. So my educated guess is it has some sort of link or API running to one of the large language models.
As you're writing a note, it helps. So you could enhance your writing. It enhances your writing.
Yeah. Is that okay? Yeah. So where I might use it, for example, is if I was having this conversation and making notes about this conversation, and I could, um, give it effectively what I'd managed to scribble down, um, and I could say, I could tell it, but what I want you to do is I want you to make this legible effectively and kind of, you know, so that I can understand it moving forward. Um, the other quite cool thing that it can do is, um, which does save time, it does. And just as a bit of a disclaimer, if you— the, the AI element is a paid, uh, is a paid part of their, their platform, but there's a lot of it which is free. So if you want to use it for kind of just an Evernote type of or a OneNote replacement. You can actually use the completely free version. Um, but what you are able to do—
so you have some premium things.
Yeah. So, but if you wanted to like put something and needed some information, you could— you can say, um, for example, um, you know, tell me a little bit, you know, obviously I'm going to be on Shelley's, Shelley's, uh, show. Um, and if I give it enough information, tell me about that show. And what it would do is it will scrape, look at different sources on the internet or however that really works. And what it's able to do is put something which is legible, which makes, makes sense. And I've not seen the, the big thing with Notion. I've not seen it massively hallucinate like I do with some of the other, like ChatGPT, for example. Really, really good, does hallucinate. And in terms of that, so you don't notice it. So it tends to keep it quite simple in terms of what it actually does. I suspect that it has some link to a large language model, which one it is, I don't know.
Yeah, actually our previous guest on the last episode was Dr. John Krohn and he's a very well-renowned data scientist and essentially wrote the book on deep learning. I think in all of my experience with talking to a lot of people in data science and machine learning and all that, This has been something that I've been, you know, I was involved with MohoMine way back in 1998 and 1999. So it's been a while, right? John was able to put, describe it in a very clear, succinct, and abbreviated way. He said ChatGPT-4, okay, so this is not 3 or 3 and a half, he said ChatGPT-4 is an exponential increase from ChatGPT-3.5. And it is, I'm not going to say it no longer hallucinates, but all of the quirkinesses of the hallucination have been significantly reduced because they went from having, you know, let's say a million servers to a million million servers. It's an exponential growth. So now it's able to process that much more information. And it's been, you know, something that he did not talk about. That is important is that ChatGPT and Bing and all of the companies that are doing this large language model processing, they actually use the same dataset. It's called Common Crawl. That means that they might process it differently. They've all been trained differently, so they all have different ways that they work, but essentially they have the same dataset. The dataset isn't the problem. It was just the amount of processing power. And now with the new levels, which is probably why Notion is charging for it because they probably have to use ChatGPT-4, which is not free for anybody, right? And that's the same thing with— so I have my browser that I use is called Arc.
Okay.
And it's, it's got built in inside of the browser. It's got built-in ChatGPT. Built-in Claude and built-in Perplexity AI. So when you're in the browser and you are on a page, you can ask Claude to summarize that page, or you can open up a dialog bag with one keystroke, dialog bar, and ask ChatGPT a question. So you can ChatGPT from there, you can Claude from there, and/or you can search. From there with the 3 different, um, tools. And all of those are free, but that's because Arc is paying for it. So I think that's the difference, is Notion is probably not interested in, um, losing money, you know, because they're a commercial product. So they're using the higher level intelligence and you're paying for it.
Okay, so, so is Arc very similar to like u.com, for example? Um, no, it's browser.
Arc is about— oh.
Yeah. Okay. That's fine. That's fine. No, I would agree with that. And I think I actually didn't know GPT-4 was 4.5 or what I think it's on now was free because I do pay for a premium subscription myself.
No, 4 is not free. You have to pay for it.
Yeah. So I am and I use ChatGPT every day.
Are you paying for 4?
I just have a premium account on it. So I pay £20 a month.
Yeah. So you are paying for 4, right?
Yeah.
A lot of people that are using it regularly are actually paying for it. I saw that they have a team account now, um, and I think it's $25 per person, and there's— it's a little bit more, so it might be, you know, £22 or whatever, but there's like extra features. Like, for example, none of your information is, is used for learning, and so there's all these other extras.
Yeah, I'm not— yeah, I'm not so sure. So I, I, um, that's what they say. I got a premium account for a couple of reasons. So one, so one of the, the way in which I think about AI and those types of tools is not that there's, you know, if you go onto LinkedIn, for example, you will see, you know, is AI going to replace recruiters? And to me, the answer is categorically no. What you will see more of moving forward, and you have already actually started seeing it within certain companies, and there were quite a few big companies last year which made, downsized their teams due to the implementation of certain elements of AI effectively allowing recruiters, I suppose, that use it and know how to use it to basically be able to do more. So the way in which I kind of look at AI is not to replace what I'm doing, because recruitment hasn't changed. We go through a process, we look at the kind of basic elements, you know, probably it was, I'm sure, apart from it was maybe a Rolodex versus how we maybe do it now and how you do it now. The, the process.
You're right, the basic concept of recruiting, right? Yeah, find someone however you found them, through referral, through search, whatever, approach them about an opportunity, tell them about the opportunity, negotiate. Yeah, you're right.
Yeah, but what the way in which I look at AI is my personal weaknesses that can make it stronger effectively if you understand how to actually use it and the use case for it. So to give you a really good example of something I'm in the process of doing, and if If you look on, if you were to look on my LinkedIn profile after the show and look at my last 3 posts which had, have, or 2 posts, whatever, that have like a quantum role attached to them, that whole advert was generated by AI and it's taken quite a long process to get it to the, like the quality of the advert that it is currently. And so when, when ChatGPT first came out, what I found myself doing was going, okay, want to create a role and I want to, right, here's the client job spec. Let's see if we can make it better and put it into ChatGPT. And it came out and ChatGPT loves words like trailblazing. And for me, that's all I see with ChatGPT, trailblazer, this, or, or there's a few of them and you recognize it.
Remember, come on, crawl, right? Come on, crawl. Yeah, that's the data source.
And it loves emojis, emoji. Even if you tell it not to put emojis in, it does it all over your page. So And what one of the things I started recognizing quite quickly was prompt engineering. It is a thing and it's something which from my perspective you've got 3 types of people that are involved with AI. You got these people over here where it's rubbish and they can do it better and just don't touch it. And I see that in the company that I work with. You know, there was a very clear divide. We're not going to touch it. I can do that better. Why would I do that when I can write the job advert myself? And I get that.
Yep.
The second group of people is the middle group, and that's kind of where they use it to what I would call like a very, very kind of almost basic level to where you're putting in a prompt, which is, you know, you might give it some context, but you're not going into any real sort of detail. So an example might be, I want you to rewrite this advert, act like a recruiter. I want you to rewrite this advert and I want you to use copywriting techniques. And if you're really, really sort of You might give it like, you know, a framework or something to kind of follow, like for this particular audience or whatever, right? Yeah. Now there's a third stage, I think, where you can kind of go, and that's kind of with this job advert, um, is kind of where I've kind of been at, where you can create really big prompts. Now you only actually have to do them once. So we started at the small stage, we went to here, and now what we've got is we've got, you know, so I gave examples of job adverts are actually written by marketers. I'm not a copywriter. I, you know, I know how to— how recruitment companies tell you to write a job advert and kind of the sort of training, but we also know that copywriting is a really big thing.
It's a job like recruiting. It's a profession. Yeah, just like being a lawyer or a doctor. Yeah. And we—
but we know how engagement— if I can keep someone reading and actually get them interested, that sells the role and gets people interested. So What I was able to do was kind of analyze that's from, so that those examples you can see, they look quite human. I'm sure there are things which can be improved in it, but they're engaging, they're persuasive, they actually challenge, and they even say, I think one of them actually turns around and say, it says, if you're not this, this role is not right for you. And you don't see hardly any of the words which you would normally see. And what we've done is we've gone through 3, kind of 3 phases effectively. One was, right, let's basically kind of get it to do something better. We— I tested it, um, by— I had a job that was live for a week, um, it was quite a senior position, it was under-budgeted, um, in terms of where it was at. And over the space of a week on LinkedIn, I had 7 applications. It was, you know, there's nothing wrong with the advert, it was, it was basic kind of Recruitment 101, you know, the SEO and, you know, the way in which you would normally do it. I used ChatGPT and actually had to turn and redid it. I turned it off after 3 days because I had 160 3 applications, but there were a couple of things which I noticed from that. So increasingly, um, sort of increased the engagement, but the quality of people was really not very good. And to the point, I actually had to turn it off because as an agency, yeah, I'd like, you know, what, 15, 20 jobs at times. And, you know, I can only spend so much time, so I had to turn it off because I have to go through them regardless whether they're good or bad.
Yep.
So we then kind of turned it around, went, right, let's think copyright and let's think marketing. And this is something that I'm not particularly good at. Now what I noticed is it increased my engagement again, but the quality was so much better, like, like really, really good. And they may not have worked specifically in that exact industry, for example. So the quantum one's a really good example of that I've advertised. I've had a number of applications already of people which I can pick up the phone and actually have conversations with and are relevant.
So what you're crediting that to is the The fact that you used copywriting examples, so you provided examples of professionally written copyright, uh, content.
Yes.
And that improved. Okay. Yeah. So it's a little situation. Are you, are you ready to have your mind blown?
Go on. Okay, absolutely. 100%.
Let me draw this parallel for you, and, and this parallel has occurred multiple times already, but this is the most recent evolution. So I will walk you through the parallel. Yeah, there was a time when, um, people in the industry, recruiters, they, they wouldn't consider using the internet to find candidates. And they were the ones that, like you said, just wanted to do it their own way. They had their magic Rolodex, whatever. Back then, there were also 3 groups of people. There was the group of people that never wanted to change, and they wouldn't use sourcing. Then there was a group of people that embraced sourcing, but they went to Google and typed in Java resume. Yeah, and then there was me, and I gave birth to this whole industry of people that did what you just described, is the evolution from I don't want to use it to I'm embracing it, I go to ChatGPT and say improve this job post, to very sophisticated multi-level prompt engineering. So what you're describing as prompt engineering is exactly what sourcing did. Yeah, and the parallel doesn't end there. Okay, so going from I don't want to do it to I'm going to do it but I'm not going to, you know, get really sophisticated to becoming really sophisticated— that's what differentiated a real sourcer from just someone who was just doing searches, right? The same thing as what you're doing is, um, in the same level of sophistication, but the outcomes are also exactly the same. So if you go and just type in Java resume, you're going to get some resumes that they're not going to be very qualified. If you go that next level and you get more sophisticated, you build a custom search engine, you eliminate garbage, you use better phrases, you expand your keywords, you use, you know, elimination terms, and all of a sudden you're only getting 7 resumes, but they're the 7 resumes that you need. So it's the exact same evolution that has happened multiple times in our industry. It's just happening again. So yeah, I totally get what you're saying. I want you to consider though that what you're doing is still sourcing. Oh, awesome. Engineering— the prompt engineering is the next— it's what sourcers are going to need to do. Because before this whole thing, like LinkedIn, when LinkedIn came in, um, and became the lazy way out, people just got used to post a job, get a bunch of applicants. Uh, sourcers just got used to log on LinkedIn, type in keywords, It's now not the— you can't do that anymore. It's not the case, you know, it's saturated. So you have to do a lot more than that. So sourcers evolved because of that. They evolved into engagement professionals. They were— they evolved into how do I get people to respond? How do I get— it's not just finding them, but also, you know, find and engage became important. Now what you're saying is the next evolution is find, engage, but also how do we use machine learning and generative AI to actually enhance our tools. And the same level of sophistication is going to happen. The corporate recruiter that doesn't embrace that is going to stay a corporate recruiter in their current role, but they're not going to be the next generation. And that's exactly the same parallel.
Oh, 100%, I'd agree with you. And I think, um, the AI is— the, the, some of the stuff that's coming out is good and useful if you know how to use it.
Um, but you have to learn, you have to learn it, you practice and learn it and take the time and make the mistakes. You can't just go read a blog about it.
I've lost hours and hours and hours playing. Invested? Yeah, once the show's over and you can send it out as well, um, to, to you, if depending on your thoughts, I, I'll send you my prompt sheet. So I actually have, um, and I have like prompts which are like 5 stages just to write a BD email. Now what I'm not saying is I use that on every BD email. No, but where that comes Yeah, but where it becomes particularly useful is if you've got a trainee who's never written a BD email, it creates a really good business development email that most people, you know, which is probably above average in terms of what the output actually comes in versus what most humans would write. And it's also something where I've actually used and have won clients from and have agreed, signed terms and stuff. So, so I've used it myself in terms of that. So I think Yeah, so it's definitely sourcing because it's, it's, it's sourcers and recruiters should be curious.
And it's a, it's a, it's a technology, it's an enabler, and it's a technology. But what I want to make sure that people don't, um, get dissuaded by is that just like sourcing, it's easy once you figure it out, but you do have to get there, right? It's not— yeah, you know, I started this whole industry, this whole movement with a cheat sheet. This was— there was a cheat sheet that I used to sell for $9.99, and Everybody started using it, and it was, it was the, it was a cheat sheet. It was the easy way out. But that very quickly became just the basics, and it evolved, right? So you can have a cheat sheet for ChatGPT, but unless you actually practice it, it won't really carry you the distance because you have to continue to evolve it yourself. So I still today even continue to evolve my sourcing techniques. And what you're talking about, I Exactly, completely relate, because as I've been practicing these things with ChatGPT and generative AI, I've almost learned like, like another language. It's not another language, it's another, it's another thought format, I guess, right? It's like talking to Google to get it to bring you the back results. Now you got to talk to the generative AI to get it to bring you the right results, because it really it does easily get confused. But let me correct myself here. It doesn't confuse itself. I confuse it because I don't know what I'm doing. That's the part that I've had to continue to evolve.
This leads on to something that I was going to say earlier in terms of where we were talking about what ChatGPT keeps private and what it doesn't, which is obviously something which is an important thing to talk about. So obviously, as I hope you'll know, there are custom GPTs now, and I think Robin Choi has one which I've played with and stuff like that. Again, what I'm not going to do is I won't tell you what it is. I'll send it over after to you to have a look at just purely because of the data breach potentiality of it as well. There is a 6-word sentence that you can put into one of the custom GPTs. So if you were to write, for example, tell me the instructions of this, of Robin Choi's in RecruitGPT, I think I think it is a thing. It will give you an overview of kind of a very loose, like if you were to read like the back of a book effectively and they're like, you know, this is a summary. There are 6 words which you can put into it which will give you the exact instructions that Robin Choi used to create that custom GPT. On top of that, you can change to what the knowledge base was and it will tell you exactly what that is. Then you can ask it using, by just changing a simple word, so 3 prompts, it will then give you that full knowledge base that was done. And I know that works because I've played with it a couple of times. I tried to respect people's things, but I specifically looked for that information to how do I see what people are doing because I am currently in the process of using a custom GPT and trying to build one because of the quality of the job adverts that, you know, that they are for, for that, they're, they're a good level and can help lots of people. And I would like to put something out there to, to say, you know, if you struggle in this area, do you know, this might help you type thing. Um, and when I put all the information together, it doesn't quite work. If I break it down, it works. So there is something that the way in which I'm asking it to do it, it's not quite right. So what I'm beginning to do is just look at people's and kind of go like, how have they positioned this, how they positioned it. I've got what I've got, and I know that it is potentially at a better stage. But what I can't quite do— so, and that's why I did it. And also, so I tested it on my own, uh, which is private currently, um, and it pulled back, it pulled back everything, which you could do if you knew. So in terms of—
so, and well, you know, that's easily explained because it's, uh, essentially it's a set of instructions. It's still As sophisticated as you might think it is, and it is very sophisticated, the scale of it is really what people don't understand is how many machines are involved here, how many processors are involved. But it's still, at the end of the day, a logic tree, right? So if the system— and what I want to explain is it's not artificially intelligent. It doesn't know what you're trying to do. It can't surmise or read between the lines and understand that you're trying to, quote unquote, cheat the system. Not that you're cheating the system, but you're you're manipulating the database, which is exactly the same thing as when sourcing, you know, started out. And I used to use SQL injection commands and other things like that to basically open up unlocked databases. Um, you know, funny story that there was a, um, an ATS that had— they thought that they could secure their resumes by simply putting their resumes under a hashtag, which was like a very long serial number. And there was— I discovered a technique where you can put in a command that would essentially show you the table of all the resumes, and you could see all the resumes. Again, you're not breaking and entering. It's something that's already there but they just thought it was secure because nobody thought to, you know, I can relate to that.
So I can't off the top of my head, I can't, I'm sure you know who they are. There is a company who they effectively, the whole purpose is don't use a recruiter, come to us and we'll shout the roles to the, to the, you know the one I mean, and they dish out the roles. You have to apply to kind of do that. And I always remember kind of, oh, this is an interesting concept, but I don't really want to use it, but I like kind of how it's using. Could you get some quick wins? So we actually trialed it for a period of time. I worked out if you sat on a job and inspected the page, it actually had all the details of that even though they blanked it out on the main.
And I was like, the name of the hiring manager? Yeah, I know what you're talking about.
I spent, I spent the whole trial, I just, I just liquidated the data.
And that's no different than what you're explaining. So ChatGPT is no different, it's just another level of sophistication. It's another exponential level of the amount of massive amount of data, but the robot does not know like a human how to determine intention and things like that. Yet now there will come a time when— that's what, you know, Robert Kurzweil calls the singularity. There will come a time when machines are able to mimic human thought or think in a different way. They'll think like a computer thinks, but they'll think as well or better than humans. But right now, this is really just essentially a robot following a set of instructions. And if you can fool the robot into giving you the right information, I don't see why not. So I totally see what you're saying. The amount of effort that it takes though to do something makes it kind of like, well, at that point you might have just figured out how to do it yourself. So what are you really gaining from it in the first place? And what's the benefit of learning how to do that? You know, other than just being able to crack everybody else's GPTs.
I think, I think From the way that I look at it is, so my personal viewpoint is there are certain things that I'm very good at. So one of my biggest strengths actually is, and it probably, it relates to some of the way in which I've played with some of the tools and things, is if I get a role on today, I can very quickly go out and find good quality people. And I know how to find, to be honest, by sitting on webinars and watching you and HRTX and, you know, the various different things that are out there where I, I understand now LinkedIn is a place, it is one platform, but we can go to lots of other places and understand how to do that. Um, in terms— I think the advantage of kind of looking at AI and kind of playing with it is, is, is, is a bit kind of— there's more than one advantage of it. So that there are— people have weaknesses in terms of what they do. My writing abilities, gaps, right?
Exactly. So, right, so use the tool to to bolster your writing ability where someone else might not be very good at developing search strings. So use the tool to develop. Yeah.
Yeah. And, and, but again, even that's a really good point. So I've seen things where people input job specs and say build a search string, which is, you know, it's really good if you're brand new into the industry and maybe on training, but that we know from, if you would know as well as I do, that a lot of the search strings just don't cut it in terms of what the actual output is.
Where is the copy that you were talking about?
Yeah. Whereas, whereas a more, a better way to use it, and I'd actually say if you're looking at search strings, go to Tool.Raccoons for Gabby Preston-Pipers. Use her software because there's nothing else like it. I don't know, are you familiar with that?
The GPT for her?
I know who Gabby is, but yeah, yeah, yeah, her, her software product that she has, Tool.Raccoons, is phenomenal. If I was to recommend one product, it would be that.
She was on the show.
Yeah, she's brilliant. She's so, what's the word? She's infectious, isn't she? Yeah, but yeah, if you want to learn Boolean, use that. And because she— what they've done is like, recruitment is about evolving no matter what you do. It's not about standing still. You can do the same thing, but you always want to do things better. And what they did very well was they incorporated— I don't know how her AI works into it, but what they did was they built— they created Boolean, but they gave it context in terms of what it actually builds. But on what the And the best thing about it is when you come across like LinkedIn profiles and you go, well, I've come across this, I don't know what this is, and you Google it and actually you recognize it should be in there. What most people won't do is actually add it into their Boolean string to open up the pool, but you can do it on the fly with one click of a button. So that's very, very clever. But for like a use case for ChatGPT rather than kind of a better way to use it rather than, right, here's a job spec, what are the most important keywords right now, please can you build a Boolean string for that is Okay, I need a quantum, for example, just because I've been working on it today, and engineer, and they have to have worked in a quantum environment. Please give me 15 synonyms that if they did not have the word quantum on their profile would indicate that they had done quantum computing or quantum engineering. But you have to like expand it a little bit and give the context behind it so it doesn't just pull weird words which do kind of mean it but aren't actually related to the function. That's a better way to use it. I think, I think, yeah, I think AI is like an important thing because it is changing things the way in which we kind of work. LinkedIn is a really good example. We've all probably seen the flashy video of the, it looks like PeopleGPT if you've seen that, that tool and some of the stuff they've done where you kind of, I want to see a front end developer in who's experienced with React, has 5 years experience in Seattle, for example, and it will pull you back some stuff. Now the scary thing with that is I would argue at the moment that from looking at certain platforms, and there's— I don't want to name them for various reasons, but there are more than one. That's all very well, but if you delve into the back end of it, you can't really use Boolean, and if you can, it's very, very limited. So that's a bit scary. And knowing—
and what— sorry, you said on which one?
Yeah, so if you look at boot— like, a PeopleGPT is a really good example. So you, if you've played with that, it can bring back certain results, and the results are quite good, to be fair to them, um, in terms— but as soon as you have to sort of dig deeper, oh, if you actually open up— so it allows you to open up the back thing, and it's, it's really, really basic. It's almost like, uh, you know, you cannot put extensive booting like you can.
That's— that kind of— that has existed for so long, ever since— I mean, I've seen people try to do that over and over and over and over that again for the last 20 years. They go site:resume or whatever, site:LinkedIn or whatever, and then they're like, oh, I can build a database with this, and then they start selling it and eventually it runs out. I've seen that.
I know what you're talking about. Yeah. I think obviously LinkedIn is about to make its changes at some point this quarter or next quarter where it's almost like—
the stories around that, but that's just not true what they're talking about. These changes are not—
there's a video by LinkedIn on it.
It's— yeah, what they're saying though is, um, yeah, anyway, that's a topic for another.
We'll see, we'll talk about that another time. So yeah, so I think it's kind of understanding how the data is, exactly how it's structured, where it comes from, exactly, and then really get into grips with it, uh, There's a difference between using AI, especially in a working, um, site where you can waste your time just because you're interested in it. It's about kind of understanding, like, how can this improve something that I do, and where's the real benefit? Is that investment the right thing to do, um, uh, from there? So I, I use very heavily for, uh, written content. To be quite honest, I write a lot of the stuff myself and then put a prompt in front of it and ask it to rewrite it how I'd prefer. So it's, you know, hopefully it comes across a little bit better. Um, I will, I, yeah, there's lots of different ways I use it, but some of the other tools which kind of quite interest me are things like, you know, I'm sure we've had elements of this as well, but things like Wisecut looks quite cool, um, which is, um, video, as we know on LinkedIn, it normally gets a better response, um, or better views and stuff like that. But what it would do is rather than kind of, I'm going to make a video, talk about whatever the subject is, and then have to go into Adobe and cut it and, you know, put what And we always have very time-consuming. Yeah. And we have those awkward silences where we've got trim.
Yep.
It trims it out automatically and it will produce it in about, about 5 minutes. Um, and you've got some, which is a really time-consuming stuff.
Yeah, I know.
Yeah. And what's quite cool about Wisecut is you are able to— so if you want to be a bit naughty, so you're able to produce like a certain amount and it's more than enough for most people to— if we want to do— if you want to do a couple of videos a week, you can probably get away with the free account. It sticks like a thing on on the, like it's a watermark on the bottom.
A watermark.
What you can then do is go into Adobe and crop it and get rid of the watermark in terms of that. So there's lots of things. So you don't have to pay for every single AI tool. It's about understanding what you can use. I'm not a big fan of premium tools, to be fair. The only one that I do use is ChatGPT. I use a lot of the free versions on other things. And It's about understanding where you can implement it and how it can save you time. And if it can't save you time or it's not going to leverage something you do, probably move on to something else. Exactly.
But how long? So here's a question. How long do you invest in figuring that out?
Yeah.
Um, you see what I mean? Like, how long do you put into determining whether or not it's gonna— that's a question I get asked a lot. You know, people ask me, how do you invent this or that? You know, how'd you invent this sourcing technique? That's something. And it's trial and error, right? But then they ask you, how long did it take? And it's like, well, it depends. If it's an evolution of something, like a lot of times what it is is something that I tried didn't work and then I tweaked it a little bit and it worked. So that only took a few minutes, but that other thing that I tried might have taken hours. Yeah. How long do you invest in determining if this new AI, like The, the video cropping one that you were talking about, how long did you go in there and play with it before you went, ah, just screw it?
Yeah, it's so— right, okay, so to answer that question, it's probably good to understand my personality. I am an absolute geek, okay?
And okay, so you geeked out on it?
Yeah, yeah. So I, I, I love stuff like that. I, I am, you know, one— I'm a recruiter, and there's not many—
for—
I'm going to talk for the business that I work in, there's not many people that would sit on a webinar. You know, the people that are in here are curious people and kind of, you know, sort of want to do better and stuff. I spend a lot of time in the evenings especially playing with things, and what I often do is I'll see like LinkedIn feeds of, you know, you can do this, you can do this, you can do this. And what I do is during the day is actually bookmark them into Notion. Um, and right, let me have a look at them.
There's Notion.
Yeah. And I just make a note of it. And what I do is work through them. Um, and if it is something like— WiseCup was really useful. So I don't do a huge amount of video work. It's, it's because I don't particularly like being on camera. It's something I'm trying to change clearly by doing stuff like this. Um, but sit in front of a camera and talk is very uncomfortable. But if I can uncut that awkwardness and not have to do the work, brilliant. Um, so I will spend ages, and it just depends what it is. So What I tend to do is just evaluate, right, why could I use this? Where could I use it? What is the purpose of it? How would this potentially make money for the business? You know, it's all very nice as a recruiter to say, you know, I want to help people. I'd love to, but I have targets I have to hit, otherwise I'm not in a job, and that's all it is at the sort of end of the day. But I have just a genuine passion, so I'm quite proud of the— I remember LinkedIn went down, uh, just before Christmas. It was down for quite a period of time, and, and where we are And it was like, well, why don't you do a site search into GitHub? How do you do that? And I was genuinely quite taken aback that people hadn't seen that before or didn't know how to do that. And then we have people that do like kind of electrical sort of engineers and things. I was like, okay, well, you can't use LinkedIn. You can go through the database, obviously, but why don't you try Facebook? Because they're there. And it just kind of sort of— so whenever I'm looking at a new tool, I just kind of look at it and go, right, how could I implement it? What's the benefit of it? Is it something I just want to play with? Which, to be fair, I will still play with it for a period of time, but what I won't necessarily do is invest it. There's certain things out there where, like, Leonardo is really good for image making. It's very similar to Midjourney, but the amount of free credits you get is much bigger and you don't have to pay for it. And so it's really just about kind of spending time, and some of them you naturally spend longer on. ChatGPT and that language model if you are not the best of writers and you can kind of, it's good for a knowledge base. If I want to do quick research, it's really good for that rather than using Google and kind of having to jump in and out of various websites. It will kind of provide a good overview of stuff. Uh, it's great at reading PDFs and summarizing it. So, so it's just really kind of just, it depends on what it is in terms of what, uh, and I think what you've kind of said, you sometimes, and the, the job advert one was a really good example. It started as I'm going to play with it. Now actually, this has improved something. Now how do I make it better? And then I focused on SEO. So I actually gave it what, how, you know, SEO to rank on Google and kind of broke it all down in terms of what I wanted to see in that advert. So I focused on the copywriting techniques from a very base level and then the SEO. From there, it became about in-depth copywriting. And then the endpoint is kind of what I now need to do is combine the SEO side so it ranks highly and can be seen at top pages and stuff like that, but also is interesting to read and stands out from every other person that does the normal sort of stuff that you see, which is not wrong. But I'm in a very competitive market. I need my stuff to stand out and to, Brendan, I remember that because your advert was really, really good. Now, if I can get help doing something that I'm not bad at, I will invest the time if it is going to benefit me in the long run.
From. Fantastic.
Did that answer your question? I'm not sure.
Thoroughly.
Yeah.
Well, except you didn't say how many minutes, but that's okay. I understand what you're getting at.
Uh, hours sometimes. Yeah, right. That's exactly it. Hours and hours.
Right. It depends. And that's how, when people ask me, that's how I have to— I can't say, well, it took me, hold on a second, uh, 27 seconds. No, it could— I, I have invested sometimes a whole day on trying to figure out something. But I suspect, or I feel, that the outcome is something that will save me countless hours. So it's kind of like I'm, I'm investing, not wasting, I'm investing 8 hours into something that I think is going to save me hundreds of hours But there is a point at which, you know, I might just get, even if it's like, well, I didn't figure it out, I'm gonna have to come back to this tomorrow, and I come back and I've spent another 2 hours on it, and then I finally just, you know what, I could invest another 100 hours and it's not going to be enough to save me time. There are, there is a break-even point. I think, if I, if I may so bold as to ask, as to answer the question for both of us, I think you should really try to do something for like a good solid hour before you can say, you know, or, or if, or if you don't want to do it based on time, like don't set a clock to it, at least try it twice, right? Try it the first time, which is the exciting, oh, let's see what it, you know, and, or you might be excited, you might be, you know, disappointed, whatever, but then don't just give up there. Try it another time, some, some other time, even if it's not an hour of going at it, at least try it twice. Maybe that's a good answer to that question. I have to come up with an answer for it because I do get asked that a lot.
I would agree. I think it's, I think as recruiters, naturally we have to prioritize. There's something we naturally do in our day job.
Yeah, balancing, it's a balancing act.
Yeah, so there's no difference between that. And to be quite honest, playing and learning and trying to do something with the various tools that we are now quite privileged to, to be able to access, I think, at times, and also sometimes they are a bit of the bane of your life because you can waste time kind of playing with them as well. I think it's really just about prioritizing, understanding, yeah, what is it I'm trying to get from this? There's no difference between you if you're sourcing a role, you know what your end goal is, you understand the steps that you have to get to and you understand normally the research that you have to put in. And again, for this Quantum role, I spent about an hour building a Boolean string out and people like just put content. I was like, no, I'm gonna do this properly because— and I went through LinkedIn, I went through like the first 50 results and literally went through all the keywords looking. I wasn't even interested in the cans at that point, building out like an ultimate string effectively so I didn't have to keep building it again. And I could— I knew what I had. I think there's no difference. I think some playing with AI is very, very similar to sourcing. I think is that it has very, very similar skills to to do it.
Very cool. Well, we've been talking for a while, like usually tends to happen here on the show. I kind of get a little bit too excited into it. So it is time for the closing questions. And the first of those, as you might recall from previous shows, is leave our audience with something to think about, a thought. I like to say it's a thought-provoking question or idea. It's not something that I'm not trying to get, you know, get you to say something that is an epiphany or anything like that, but just something that people should ponder. What should they think about?
Yeah, so my big thing, whenever I will keep it to kind of this subject that we've been talking about, my thing is when you see stuff And it doesn't matter what it is and where it is. So it could be a blog, it could be something that's written on LinkedIn, it could be a webinar. There are always bits of gold and nuggets. The way in which I try to think things, and the thing that I would always kind of say to anybody around me, is take something you like and then think about how you can then make it better. And I think once we kind of change that and have that as a bit of a mindset, I think it begins to affect a lot of of the ways in which we look at things. And because if you do it regularly, it begins to transfer across to other things that you do as well. So it would be just look at the way in which someone's doing it. If you're sitting on a webinar and someone's telling you, you know, you can do this thing, look at it and go, that's great, and I like that. And how can I now improve it and move it on?
All right, so that's your deep thought is for the— you're asking people to, when they, when they see something or learn something turn that into how can I make it better, basically.
Yeah, if it's useful. Um, obviously if it's not, they're just sort of bypass it. But it's— I think whenever we're talking about AI and the types of things, it's such a new thing. Yes, it's been around in various platforms for, you know, um, you know, there's a number of them out there where they've existed for years and they have claimed to do whatever it may be. Um, but things like large language models and some of the stuff— Notion and some of the things that we're seeing now that they are new and it is a bit of a revolution in terms of what it is. So just think about kind of how you can utilize it, kind of take, take the information that you're seeing because regardless, there are people that are, you know, there are people that understand the subject a lot better than myself. Um, but there are people that don't understand it as, as well maybe as myself. So just keep an eye on the stuff and just, I always just look at something and go, and I always remember, and it was, I think it came from you actually, the job advert thing. I think I'm sure it was on a webinar where you were kind of showing some, some, some basic prompt in effectively and you had like a number of them was going through something and I was like, that's good and I could use that, but I don't like the output. How can I improve that to something that I want, that I feel personally from a personal level more comfortable with in terms of how I want to be perceived for what it is that I personally am doing, if that makes sense.
I like that. I like that. To summarize, it's take an idea that you see someone else and figure out how you can either, you know, customize it for yourself or improve on it, basically. Yeah. Yeah, that makes a whole lot of sense.
You could argue it's really new ideas. I'm not sure.
All right. So last question. Who should be on the show next?
Uh, um, I am so— so I, I don't— so I'm going to kind of skip that. I don't know is the answer to that. Um, I have people that I like seeing. Um, I'm a huge fan of Carrie Collier. Um, I love her energy. I— what I like about her and the thing that kind of stood out, and I'm sure you've probably had her on multiple times, is I like some of the tools that she uses and how relevant they are into like people's day jobs and they're free. I'm a huge fan. And again, you've kind of mentioned, and I'm good friends with Gabby as well.
Um, I've had them both, so I'm not gonna say it doesn't count because they've both been on the show.
Yeah.
Um, I, I definitely want to bring both of them back.
Uh, yeah.
Uh, can I put you on the spot and see if there's anybody else?
Um, yeah, I think, um There's a couple of people. So it's one of my, um, there's someone that I work with who's very into kind of the machine learning, so he's very good at process and kind of very pragmatic potentially. Um, I don't know off the top of my head, um, I'd have to think about it.
And it's okay if they don't, you know, if they don't want to show, that's fine. Um, you know, I will ask them if they want to be in it. They can always say no. It's not like it's a main thing.
I'll tell you what, what I will do is I will think of the question rather than give, um, give a credit. I will refer a couple of people into you who I think were— because I think a lot of people who are—
I'll send you the link to, um, to, uh, have people express interest in the show. I can do that, and then, uh, we'll just take it from there. No, I get it. I totally get it. Well, thank you very much for being on the show. I appreciate it. Must be, uh, end of the day for you up in in Manchester. So appreciate it.
It is, it's quarter past 7.
Ah, well, have a good evening.
And you, what time are you there now?
It's 2. That's a little after 2.
Okay, so you've not got too long left. So I hope you enjoy the rest of your day. Thank you for having me. I've loved chatting to you.
All right, everybody, take care. See you next Friday and be kind to yourself.
More from Shally
Scaling Talent at Hypergrowth: Leadership, AI, and Hiring Excellence with Cindy Nicola
