#426 Recruiting and Artificial Intelligence (AI)

Recruiting and Artificial Intelligence (AI)

Article written by Niels Brabandt

Recruiting is under pressure.

Artificial intelligence can increase efficiency and effectiveness. The question, however, is how this can be achieved. Recent studies by the Wharton Business School and the University of Pennsylvania have shown that the majority of investments in AI fall short of expectations, in some cases significantly so. In recruiting, in particular, there is a desire for quick, tangible, and implementable solutions. What might these look like? What limitations exist, and how can you optimally position your organisation in this regard?

Recruiting

The pressure on recruiting is constantly growing. Regardless of whether there are more or fewer applications, more resources are rarely or never made available. Often, expectation management does not take place or only to a minimal extent. Expectations quickly arise that the best results must be achieved within a very short time with a small and inadequate budget. Artificial intelligence (AI) can certainly help here, provided that the framework conditions have been clarified correctly in advance.

The Pros and Cons

There are clear reasons in favour of AI. AI can check pre-defined factors faster than humans ever could. The time saved leads to cost savings. Only if the data basis, the model, and the structure of the AI have been excellently designed can AI be more objective than a human being. People who have never been professionally trained to recognise unconscious biases are particularly unsuitable for making decisions in recruitment. In addition, AI is significantly more scalable than human resources will ever be. Finding the right candidates is easier and is well-supported by AI-powered chatbots.

However, there is another side to AI as well. If the database or model is neither well-structured nor well-known, discrimination can occur to an even greater extent than it already does. The lack of transparency of AI can become a legal risk. Dependence on AI can also lead to a standstill or even incapacity in the event of a failure. Additionally, applicants often decide whether to work for an organisation based on the personalities of its people. The people they talk with are a decisive factor in the process. This aspect is especially true for highly competitive positions. AI can only reflect this factor to a limited extent, if at all. Furthermore, the initial investment in AI and the necessary training often involve considerable costs and effort.

Regulation and compliance also play a role. The class action lawsuit against Workday, alleging discrimination, shook the world of AI on a global scale. It is therefore still important to carefully consider which path you want to take here.

You can find more details on these aspects in this week's podcast or video cast; see the links below.

Implementation

If you are now wondering how to determine whether your recruiting will require more or less effort, certain factors need to be examined at the beginning. If you need to handle a high volume of applications, you will likely need to review numerous standard factors. These factors can be efficiently checked with AI. In general, the higher the degree of standardisation that can be defined in recruiting, the faster AI can be helpful to you.

Ultimately, however, the decision is always made by a human being. Regulatory and compliance requirements often demand this, in some cases, very clearly regulated and prescribed. To date, there is no known case in which an organisation has been held liable for an AI error. The outcome of the class-action lawsuit against Workday remains unclear. Also, always remember that when it comes to comparable offers, the human factor often determines whether a person decides for or against an offer. Thus, AI is certainly helpful and useful for you. Still, it does not replace the human factor, which remains the most critical aspect in terms of talent acquisition, development and retaining talent within the organisation.

---

More on this topic in this week's podcast: Videocast / Apple Podcasts / Spotify

For the videocast’s and podcast’s transcript, read below this article.

Is excellent leadership important to you?

Let's have a chat: NB@NB-Networks.com

Contact: Niels Brabandt on LinkedIn

Website: www.NB-Networks.biz

Niels Brabandt is an expert in sustainable leadership with more than 20 years of experience in practice and science.

Niels Brabandt: Professional Training, Speaking, Coaching, Consulting, Mentoring, Project & Interim Management. Event host, MC, moderator.

Podcast Transcript

Niels Brabandt

The application arrives, an AI takes a look and immediately you have the decision. That is the dream of many companies. Many companies say it needs to be possible today that someone simply sends an application to my organisation. The AI just checks things and at the end the AI says, okay, look, we had 150 applicants, these three are the ones you need to hire. Well, is that a dream or is it a reality? We need to look at what does recruiting and AI actually mean? Because not only from a factual but also from a legal side, quite a number of things are going on right now and let's take a deep dive into what is recruiting and AI really about.

So let's get straight into it when it comes to recruiting and AI. And of course some people said, isn't something on the legal side going on there? Yes, there's quite something on the legal side going on there. That is one way, how to put it in. Quite, quite massive, quite, quite a massive aspect is going on there. Very important here. I'm going to of course quote a magazine here, quote a newspaper here, everything of course.

And that's as a fair disclaimer. It's all fair use, it's all covered, I'm allowed to quote that. So I'm always quoting real world material to make this as relevant as possible for you as an organisation recruiting anti AI, what of course is what people are looking at. So first, at the moment, recruiting is under pressure. It is not the case that people say, oh, you have more applicants. Oh, I'm very sorry, dear HR person, we are going to give you more time for that. We are very appreciative.

We know that you have more work to do here, you have more time, we just take a couple of tasks off you. That's not how it works. Usually they say, yeah, there are more applicants, that's more choice for you, so good luck and take it from here. And then you simply have to deal with it. So you need to have the best result. At the same time people say, look, depending on which countries, of course you work and they say might be a bit tough on that. And so when it comes to the budget, they say, not a budget race at the moment, so you need to get better results.

While poly at the moment you might have a couple of more applicants at hand. You can of course simply say, yeah, we do a bit of bingo session and we randomly cancel people and say, not you, not you, not you. Because we didn't have time to look at it. However, the legal consequences of these might be severe and extremely expensive. So you need to be right on that end as well. So that means you need the best budget as well. And of course, when you now say, okay, best budget, that usually means we have to go on a tiny budget, limited budget, and that means that we then need to have more manual work afterwards.

So we need more time. And usually the answer is more time.

What about, what about no? What about not giving you more time? So you also need the best time, you need the best result with the best budget and the best time. And that is where the conflict begins. When people now say it needs to be possible, usually people who have no idea about AI. I got my AI training from the University of Pennsylvania Wharton Business School. These are leading institutions.

The training was extremely hard, but it also was extremely useful. When people say, and usually these quotes always go the same, people say something like, in today's times, it needs to be possible that an AI just tells me who's the best applicant. Isn't AI so smart that that needs to be possible? Right? And especially people who have no idea of how AI works are the ones who, let's just say, speak the most about it without knowing what they're talking about.

It's very important. We need to take a deep look and of course we can talk about this topic for two weeks. This is a, this is a very short version of the pros and cons. Because especially now when I'm invited for keynote speaking, training, coaching on consultancy projects, project interim management or any kind of panel on which I'm sitting, usually people say we need to have quick solutions. And when I ask them what kind of solutions, then they are looking at and what kind of AI they are looking at, they only say, yeah, you know this AI that checks things. And surprise, surprise, we do not have AI check things in science because that doesn't exist. There is no AI checks things in science or in real world practise or anywhere.

This is simply not AI. This is simply not AI.

So probably, surprise, surprise. So when we look at the pros, and of course what people always look at is time. And it is very obvious that, for example, let's say you have people who work for you and they need to be in a courier service, you need people who have a driving licence. And of course Americans will now say, are there people with no driving licence? Yes, we have a rising number of especially young people in large organisations and especially in large cities who say, I don't have a driving licence because I don't need one. When you live in Europe, for example, in Hamburg or in Berlin or you live in London. Many people do not have a driving licence because they never needed one and it's perfectly reasonable not to have one.

But of course, when you say, hey, our job simply means you need to have a driving licence, that means, and that is very obvious here, it's less time consuming to do that with an AI. When you say, look, there are 250 applicants, we have five jobs open. Please say no to absolutely anyone that has no driving licence and that is done within minutes.

And AI can check very much quicker. And of course that also means it's more cost effective. Let's say people apply in your organisation via PDF. So you don't have some website, you don't have a portal, you simply have an email address. People apply by sending your CV via email. And now you have to read through all these open text CVs, 200 of them, sometimes even more. You will take way longer compared to the AI, because the AI can simply say, okay, these are the 50 applicants who mentioned having a driving licence.

And not one for a motorbike, but for a car, the driving licence. So that is time saving, that is also cost saving. And it's also, that's very important when you have an AI which is done, which is designed with proper data. And I mean, when I say unbiased data, hardly any data is unbiased. It's almost impossible to have non biassed data. But when you have the data most close, closest to no bias, the AI in the first step is way closer to objectivity. When the data is designed the right way, always with that limitation, when you have really bad data and a really badly designed AI, everything just gets a turbo button, everything gets way worse from there.

But when you, for example, say, hey, I am, I have no biases towards people who have a different appearance, have a different picture than I expect them to have, or they, they wear something different, not suit and tie as I expected, but they wear something different. And I am not biassed towards names or towards addresses where they live. When you are, when you, when, when you know how local areas work and which is the good part of the city or the bad one or the not so great one, let's put it that way, easily, very easily. When you were not professionally trained on unconscious biases, you will have unconscious biases. And you always spot the people who have unconscious biases by these, because these are the people who tell you, oh, you know, Nils, when I have an interview within two minutes, I know if it's a fit or not. I just know Because I'm in the industry for so long, it's my experience, you know, I just know in two minutes. And you usually don't know in two minutes. Very important.

So objectivity, when the data is done and designed the right way and implemented the right way and the AI is a good one, then the objectivity there is way better than the objectivity any human being can have. And of course it's scalable. I just give you a very simple example. Let's say you put a job ad out there and you think, yeah, let's hope for 20 applicants. We are small company somewhere in rural northern England.

Let's hope for 20 applicants, maybe 25. And for whatever reason your job ad ends somewhere getting shared on social media and suddenly you have 150 applicants. But you still need to check who has a driving licence because you need a salesperson. The salesperson needs to drive around the countryside because that's your product. Your product is sold on the countryside in, in the countryside. And I will simply check for driving licences quicker. So it is scalable and you can always say, okay, I add more server power, I add more it, I just book it on demand for the next couple of weeks.

That is way more and way quicker on a scalable level and implemented in your organisation than saying, okay, so for this one recruiting, we need two more HR experts. Hiring them, finding them, onboarding them, takes way longer compared to just booking more AI or more power with a server to calculate through lots of applicants. So scalability is simply given. Also the matching, when your matching follows very strict criteria, the matching can be done way quicker with an AI and especially when people now have questions. So, for example, let's say you are looking for someone who works in logistics and they need to load and unload ships because you are somewhere at the seaside and your goods arrive by ship and you need to get goods off the ship and you need to get goods on the ship. Import, export business. And you know the usual question that are now coming up, do you have a shift system?

Is it two shifts or three shifts? What about weekend work?

What about Sunday's work? What about night work? Oh, by the way, what about weekend Sundays at night work? Do you pay more for that? If so, how much? These are things you're probably not putting in your job ad because you do not want to see your competitors using your data and then simply adding a pound or a dollar or euro per hour and outperforming you on that account. So chatbots can answer these questions right off the bat.

So when someone says, I looked at my job ad on a Saturday, Saturday night, didn't have anything better to do, I want to go to the pub at 8:30 after I watch Match of the Day, a bit of football now, surfing around for job ads. But about 8:30 or 9, I probably just go to the pub. And between that they find you a job ad and they have questions. When they then have to think, oh, there's no one in the office, so I have to wait till Monday morning. Oh, I'm, I'm at work on Monday. Yeah, maybe Monday afternoon, I just give them a call for my mobile and they most likely will have forgotten about your job ad by then. So when you have chatbots, they can simply answer 24 7.

And they are also designed by AI. And, and when it's well set up so you tell the chatbot, only give answers when you really know, you put the temperature, that's the term you use in AI, you put the temperature low. That means the AI only gives answers. That actually is the answer from, for example, a file you uploaded. So very important is that the chatbots only give answer where they are sure about and they do not hallucinate, which you often have heard about, by the way, when you look into science. Graham Jones just wrote, he's a professor at the University of Buckingham, just wrote a great article about that. When people complain about the 0.7% of hallucination that AIs have, humans have about 7% of hallucination 10 times more than AI, we sometimes say things, oh, I'm very sure you can go to our location by public transport and someone says, no, I think it's under construction.

No, no, we have public transport and you find out, no, it's under construction, it's not working. So you see, hallucinations are a big thing, of course, with AI, but we should take care of our hallucinations first. So chatbots can tremendously help. However, there are of course cons to the whole system first, of course, when you have bias in the data, and that is the main thing you probably heard about the tie, which was a chatbot by Microsoft, had to be taken offline for racist utterances with a chat person they chatted with and they made it public. Or Amazon had to take their recruiting algorithm offline because it was discriminating. By the way, these discriminations always go against women and minorities. You never so far saw any AI where you said, oh, we had this discrimination case against rich white men because, you know, they are so discriminated.

And now I Made it what? No, we just don't have these cases because usually we have discrimination against women and minorities that gets put into data, then the data gets uploaded or implemented into an AI and then you just put a turbo button on discrimination. So bias, of course can be worse when you simply buy a cheap AI from somewhere where you don't even know what the model is and how things work, but you thought oh, 30amonth sounds fair, so let's just buy the cheap thing. And then of course everything gets worse from there. And of course there's a massive issue with intransparency. When you already look at the regulatory we have right now in the European Union. Especially in Europe, the European Union, where people have a right that you give a reason and when you then say oh, why, why did you get a no?

Yeah, because the AI said you get a no. You don't like that? Yeah, bad luck then because the AI said no. This is the definition, this is the definition of remember maybe computer says no.

Remember that computer says no. Little Britain, that's where this happened. So the intransparency here is a massive issue and of course a risk with legal aspects, of course. And also in addition to that, let's say you are in competition, you need developers. And we all know that the, the competition for talent in the development space is absolutely insane. Mid sized business often complain large corporations outperform them because large corporations simply pay more or for more benefits. And then they say I don't get any talent because of these large corporations.

And some mid sized business say no, we actually find people, we find them without job ads because we find them because we have a good reputation because no one ever will say oh, I had these three different corporations or these three different companies, a small one, a mid sized one and a large corporation. And, and I compared them and this chatbot was so charming that I actually decided in your favour because of your nice chatbot personality is often, especially when your offers are comparable, people say okay look, maybe the large corporation pays a bit more. The mid sized business is a bit more family, like a bit more personal. And the small size business also has advantages because you have freedom, you can do whatever you want as long as you deliver on the promise. And this personality aspect is something which usually gets lost in AI. I haven't heard a single person so far saying I like the personality of my AI chatbot so much. I decided in their favour. No. And in addition to that, when you are relying on AI and suddenly the AI gets taken off the market and of course we already have companies who simply became way too large, way too quickly and then they were purchased and someone decides, oh, we just take this product off the market.

And someone says, excuse me, you can't just take our video interview analysis AI of, of the market. And believe me, they can. When you have a dependency on AI, it means as soon as AI is not working, you cannot process a single applicant anymore because you have to go back to base and do everything manually, which can take weeks or months. And we all know what happens when you let applicants wait. You get the rest of the people who simply didn't get a job anywhere else. So the dependency, of course is an issue here. Many of great job situation.

Many of these great job moments where people really say I am very happy to work for this company happen in the context they say it's about the personality. And when they say, look, we had comparable offers, but my recruiter did understand me way better than anyone else. And the leader, the executive I spoke to in the company really understood what I am looking for, I understood what they are looking for. And it was really a match between human beings. That is so important still that many people say the AI is helpful to make things efficient and quick. However, the decision is made between human beings and that of course is the context. And this context can't be captured by AI, at least not as of yet.

And of course, costs while you can do, while you can have cost savings in the first place, there is an upfront cost to get AI into your organisation. You need to train people professionally, train people on the tools and everything. If you think you can get AI through the door by simply saying, oh, let's just invest the bare minimum with a couple of training videos, that is going to make everything worse. So there is an upfront cost, be aware of that. So the first investment needs to be made. And in addition to that you have regulatory and compliance aspect. And when I said at the beginning, isn't there something going on at the moment?

Yes, there is. Workday, one of the largest companies using AI in recruiting. It's a massive tool used globally. Many of my clients use them. And now Workday gets sued. There is one man who said he applied hundreds of times with a company because he got suspicious. So he applied more often and often within a very short time, even during weekends and especially at night when he applied, he, he got a very quick answer.

So he knew no HR person could have looked at that, not at that speed, not at that time. And very important is this lawsuit now doesn't go well, so of course assumption of innocence applies. So we will see how this goes. A workday says they didn't do anything wrong, so let's hope they are right with that because the claim here is there's a structural approach to discrimination against surprise minorities in this case based on disabilities, based on mental health, disabilities, based on race, skin colour, people of colour discriminated here. By the way, all of these cases are often going against women and minorities because often the discrimination is in the data. So here we see a case which at the moment doesn't go too well, at least up to this point with Workday, because Workday so far had to contact every single person who got a no from the software and any person over 40 needed to receive an email to be able to join the class action lawsuit. That is probably one of the worst case scenarios as a software company you can ever face.

Let's see how it goes. It's going on for a while right now. There's a hearing in August of 2025 and let's see how it goes from there. It will not be decided by tomorrow. And no matter how this ends, either with a yes discrimination or no, or maybe they settle for a small amount or for a huge amount, we will see how this goes and then we can see how does AI perform in the light of being looked at from a legal point of view. So when you of course now say, okay, that is quite something and I need to be on the safe side, it can't be sued, can't risk that, how do I implement that? And there are a couple of scenarios where AI is a safe bet and number one is a high volume.

When you, for example, say, look, we need someone working for us in logistics, you need to load things off a ship and load things on a chip. It's nothing more than that. Get things off the chip and then, and then get things on the ship, on, off, on off, roll, on, roll, off the roll, roll ships. So when you have high volume applicants, but you say driver's licence is a must have, or you need to have any kind of Operational Safety Hazard accreditation or any kind of work safety certification because it's part of your role, that can be done way quicker by AI and use AI for that because AI is really good at that. And as soon as you say, hey, it's not just that we have a bit more than that, but you have standardisation in place. For example, you have a three tier application process. Step one is the AI looks at do you have the Operational Safety Hazard accreditation or Any kind of certification for work safety? Yes. Cool. Proceed. Driver's licence? Yes. No. Okay, you have that. Very good.

By the way, we are an international company, so everything is in English. So we need to make sure that you are able to read English papers. And then they have an English test that you, for example, with the European reference frame that we have there, you need to speak English on the level of B2, or maybe even C1 or even C2, which is native level. And then you have an English test which you have to do and these tests are standardised and then you can see. Okay. Did they fulfil the operational safety hazard thing? Yes. No. Do they have a driver's licence?

And I don't mean for a motorbike, but for a car? Yes. No. If you say, oh, we need a lorry, so a truck driving licence, it gets way difficult from there. But also it's simply something which needs to be checked. Or you say, we are willing to train people when they have no lorry or truck driver certification. So far we are willing to pay for the driver's licence, but only if they already have a driver's licence for a car.

So we don't start from the scratch. No one ever drove anything and we take them to truck driving. No, we only pay when you have a driver's licence to get to our location back because there's no public transport. Everything that is standardised and then followed with the language test. Everything that is standardised can be done by AI, way better compared to manual work. However, at the end, especially when people say, could you give me a decision? And I want to have a reasoning for that, the decision is with the human being.

And so far I do not see one case where a court upheld the decision and said, oh, was a bad decision, but the AI is to blame. No, it's not your fault. So far. People said anything in your organisation is your accountability. And when you do something wrong, you have to stand up for it in any kind of option, in any kind of context. And also be aware with platforms such as Kununu in Germany or glassdoor.com in the English speaking world, people can give feedback how they liked your recruiting process and if, if you liked it or not, it's not relevant when people say it was very stiff, it was very formal, it was not very personal, it was all done by AI and I really don't think this is a great company to work for. Then you might dislike this judgement.

But still, people have the right under freedom of speech actors. For example, in the US to put this out there and it multiplies from there. So be aware that when you take anything into account, we have here and you say we have a high volume, we have a lot of tests where I just have to tick boxes. Yes, no, lots of standardizations. And from there we do the interviews and then humans make a decision at the end then AI will be tremendously helpful and make things for you more efficient and more effective. And I wish you all the best implementing that in your organisation. And when you now say, hmm, not sure.

I have two to 47 questions right now, feel free to contact me. So first, of course when you're now listening or better say watching on YouTube, feel free to leave a like here and subscribe to the channel. Also put the little alarm bell on so you don't miss future episodes. Feel free to comment. I'm looking forward to future discussions we have here. When you are now listening on Apple Podcasts or Spotify, feel free to leave a five star review there. Thank you very much for doing so and recommend this podcast and video cast amongst your friends, colleagues, anywhere you like.

Feel free to share it on social media. Thank you very much for doing so as well. One thing which we only have on YouTube, very important are the YouTube shorts. Leadership tips. A leadership tip every single day published via YouTube. Shorts are only available there. So it really pays off to follow YouTube because we are expanding our content there also and it's all free of charge.

So there you go. Feel free to take it from here. And Apple Podcasts and Spotify of course also worth following. That's our podcast. So when you want to listen on the go, that's where you find it. Also feel free to go to my website, NB Networks Biz. That's where you see everything I do.

When you now say I need to discuss something but I do not want to do in the comment section, feel free to drop me an email nb networks.com I answer every single message within 24 hours or less. This episode happened due to questions I received last week. So thank you very much people for doing so and then we just take it from there. When you have something very specific where you say, hey, I need someone for our company, we need a training speaking at our events. Training, speaking, coaching, consulting or mentoring programme. Or maybe could you help us with project or interim management? Of course I can do so feel free to contact me there.

But you can also simply send me an email and then we simply have a chat. Doesn't cost you a penny and I'm looking forward to hearing from you anytime. Thank you very much, by the way, for the great discussions we had in the last couple of weeks. When you now say I'd like to have a live session, that available?

Yes, it is. Expert.nb-networks.com you can put your email address in there and as soon as you do so, you only receive one email every Wednesday morning. It's 100% content ad free guarantee. And as soon as you put your email in there, you receive this email every Wednesday morning. As I said, 100% content ad free guarantee. It's free access to all the articles I published, no paywall or nothing. Free access to all the podcasts I have and the video cast of course as well.

And we only mentioned in the Leadership letter. Only there, nowhere else. Only in the Leadership letter we mentioned when do we have the next live session and where the access is. That's only mentioned via the Leadership letter. So feel free to join.

I'm looking forward to seeing you there. We just had a great live session. Thank you very much for anyone joining there. The last bit. And by the way, of course, feel free to connect with me on social media. I'm on LinkedIn, Instagram, Facebook, the German platform, Singh, by the way as well, and on YouTube, so feel free to connect with me on there as well. At the end, however, the last aspect is always the most important one.

Apply, apply, apply what you heard in this podcast. Because only when you apply what you heard, only then you will see the positive aspects that you obviously want to see in your organisation. I wish you all the best doing so. If you need any kind of help, feel free to contact me.

I'm available 24. 7. I'm answering every single message I receive within 24 hours or less. I'm looking forward to hearing from you there. And at the end of this podcast and also at the end of this video cast, there's only one thing left for me to say. Thank you very much for your time.