cube19 CEO, Dan McGuire discusses the future of Artifical Intelligence in the staffing sector and it’s possible impacts.

Artificial Intelligence in Recruitment… It’s coming up a lot nowadays. The reason why I’ve called this artificial intelligence and all that malarkey is that I think there’s still a bit of a debate going on about what it is, what it means and how we can take advantage of it.

Actually, a question, is there even any artificial intelligence in recruitment at the moment as it is? What was interesting about this is that 5 or 6 years ago, when I was first getting cube19 off the ground, people would ask me to come and give talks about big data, they assumed that I would know about that because I was in the analytic space. 

The point I was making back then was that our industry has an issue with small data, the data that’s being created in our own companies. So don’t worry about big data for now. Just think about the value you can get from your own data and think about trying to fix that. Here we are five years later, now everybody wants to talk about artificial intelligence in recruitment. Big data? That’s so last year.

No one’s talking about big data anymore. Yet we still have the same problem. We still have this underlying data issue in recruitment.

The first thing I wanted to point out is that I’m not an artificial intelligence in recruitment expert. If you were coming to see an expert, you’re definitely in the wrong place. But it’s my 20th year in recruitment. I actually started at S three.

So I started in recruitment. I then co-founded Broadbean Technology in 2002 and I’ve been doing cube19 for the last few years. So I’ve been in recruitment for a long time. I’ve been much more on the tech side, but of course, I’m not a technical person.

So whenever I’m looking at technology or talking about it, I’m always coming at it from the commercial aspect. So looking at the commercial realities of technology and how it can be applied in businesses. So I thought it would be kind of remiss to not mention that last month was actually the 30th anniversary of Tim Berners-Lee inventing the worldwide web. What I thought was great about that was that Tim was an IT contractor.

He was working at The Sun and was trying to work out some stuff about information management and just accidentally came up with what effectively became the World Wide Web.

He gave it to his boss and was obviously quite excited about it. And his boss wrote on the top of it like a teacher marking work at school. ‘Vague but exciting.’ How right he was!

Artificial intelligence in recruitment

And actually, what was great about it is he then gave Tim Berners Lee a bit of time to kind of work on this project, and they turned it into the Web and came up with all of the precursors for the Internet. And they made what was probably the greatest decision for humanity but the worst commercial decision ever… They said we’re going to make the underlying code of the web free forever. So that’s kind of led us to where we are now, although you can’t help but feel a little bit for Tim Berners-Lee, the inventor of the Internet who just kind of gave away the keys on day one. You could have owned the whole Internet.

So everyone is talking about artificial intelligence in recruitment. What I thought I’d do is pick up a few of the quotes that we saw from these people that seem to know what they are talking about.

Elon Musk said,

“It’s more dangerous than North Korea.”

Now. I will say this quote was from a couple of years ago. In the last year or so, Kim seems to have chilled the fuck out, and that might not be quite as relevant.

Then you’ve got Google doing what Google does – massively going over the top,

“It’s more important than fire and electricity.”

Sundar Pichai, CEO Google

Seeming to completely ignore the fact that without fire and electricity there would be no AI, but he’s the CEO of Google so he gets to say what he wants. 

Stephen Hawkings, there’s a man who does know what he’s talking about. He actually said,

“AI could be the worst event in the history of our civilization.”

And this was actually part of a much wider interview where he was kind of painting the picture of a well, where at the moment we’ve created AI, you’ve got Syrian Elector and that sort of stuff, we’re in control of it, but it will eventually overtake us. Humans can’t evolve fast enough to keep up. The AI controls us. There’s all these AI-powered weapons and civilization ends, which Mark Cuban more succinctly put as, ‘If you don’t think Terminators are coming, you’re crazy.’

I did hear a good one, actually. They said ‘artificial intelligence doesn’t stand a chance against natural stupidity’ – So at least Gary Goldsmith’s got nothing to worry about!

Now onto the serious stuff. So one of the biggest things is that when we talk about artificial intelligence in recruitment, AI and recruitment automation are getting used kind of interchangeably regularly, and they shouldn’t be. They’re very different. This isn’t an official definition, it’s kind of like the world of AI according to me…

“Automation is software that follows pre-program rules, and AI is software that’s designed to learn.”

Dan McGuire, CEO, cube19

I think that’s the fundamental difference between the two. So there’s a lot of companies out there now that are saying that they have some sort of AI solution, but really it’s just automation. So whenever anyone’s talking about that, that’s the question. How does it learn? How is it designed to learn? What is it learning? Can it change? And if you break down intelligence into its rawest form, it’s really the ability to solve problems, to learn, to be able to change your mind based on new information.

You can think of it in real simple terms as a baby or toddler learning to walk. Although most people seem to do a pretty good job of getting their children up and toddling around, you couldn’t necessarily say what was going on in a baby’s head or toddlers head as it’s learning to do those things. It just pulls itself up, walks around, falls on its arse a lot, has a lot of rinse and repeat, and eventually, it solves that problem, and it can walk. That is actually quite a good metaphor for the way that a machine learns. It will just keep going over and over and processing and trying to reach its outcome, failing many times and literally falling on its arse until it solves that problem.

That’s the difference between AI and automation.

What I’d say on this as well is that there’s an awful lot of the benefits that people believe that they will get from AI, things like removing laborious, manual jobs for humans, making people faster, better, more efficient. Those things can all be done already today with automation. So it’s not like automation is the shit version of artificial intelligence in recruitment. There are business benefits that could be had now from automation. That’s really what I think people should be spending a lot more time focusing on is how can they automate in their businesses?

Don’t worry so much about AI and whatever else. And I’ll explain why I say that – there was a survey that came out last year, ‘top three adoption barriers to AI working out’. The main one was a lack of data and skills.

AI needs data to learn. There was a good quote from Michael Dell, who says, ‘If AI is the rocket ship, data is the fuel’. I love that quote. AI has to learn something from us as an industry. If we’re even talking about AI in recruitment, we still have these challenges. If we want AI to learn from the data that we currently have, well, I mean, you lot will know what your data is like in your business. (Some will be better than others.) But if we’re realistic about it, is it a good enough data set for any AI to learn from? 

I want to give an example of real AI in play. It’s a story about how Google DeepMind created an AI to compete with the game of AlphaGo.

So for those who don’t know, it is a Chinese game. It’s 3,000 odd years old, similar to chess, but far more complex. The most complex board game in the world. To give you an idea of what that means, after two moves in chess, there’s something like 400 possible moves, whereas in Go, after two moves, there are 130,000 moves.

They made a documentary on this. I know at least one person in the room has seen this. Has anyone seen AlphaGo on Netflix? Mike Smith, it’s just the three of us, Mike, Francis and I get together on Fridays and watch documentaries.

It’s really interesting. Because what it does is they’re saying that they thought this is like the gold standard. If AI could be a human at Go, then they’d really cracked it because of the sheer amount of human intuition that goes in.

So they got this guy Lee Sedol, he’s untouchable. He’s got like 18 world Championships. It couldn’t be beaten. They built this program and they gave it thousands and thousands of games that have been played. Let it learn from that. They effectively taught it hundreds of years, thousands of years worth of playing in 40 days.

Then they let the AI start to play itself, and kept playing itself, and it learned how to play the game. Even after that, the Asian media said there’s absolutely no way the machine can beat Lee Sedol in this. So they got the game together. They put a million dollar prize money up, and they let him play the machine. Sure enough, the machine beats him.

So he gets beat and then they play again. They’re like that’s surely a one-off. So then the machine beats him the second time, and you can see this guy is starting to break because everything that you thought you knew is just being crushed. The media can’t believe it. Third, the game wins again. Then he got one back to get it to 3-1. Just at that point, it had flipped completely from the machine can never beat the human, to this human is pretty amazing that he even got one against the machine. 

It finished up 4-1, I think they donated their prize money to charity because it’d be a bit weird giving it to the computer. I’m sure you’re wondering what the hell has this got to do with recruitment. Well, there was this line in it, and they said everything AlphaGo does, it does because a human created the data which it learned from.

So when I think about that, in the recruitment world at any point, if we really want to be able to embrace AI in recruitment, we need to be able to have the data to reflect what happens in the real world, and it just doesn’t. I wish I wasn’t still having to say this. I was saying this 15 years ago at Broadbean Technology. I will keep saying this until it catches on. But we end up with a situation where we have data that’s incomplete, inaccurate.

We haven’t got the right fuel. AlphaGo is just a good example of if you have every move, the good moves, the bad moves, the anomalies, the things that you thought were good that ended up being bad. That’s something everyone needs to know in their business anyway. But actually, in your business, it’s all well and good knowing what’s good. But you also need to know what bad looks like.

You just have to be able to identify when things aren’t going according to plan as soon as possible. We still seem to be obviously heavily reliant on recruitment consultants to be better at admin. But this is really a cultural thing. And one of the messages I hope comes across today is about the importance of actually not just collecting data, but having a data strategy, what data is needed and not just for the future, (which is very exciting) but for now and today in business are people collecting the good and the bad?

Does the data reflect what’s happening in the real world now? That’s obviously kind of like the gold standard. There was a different one where I thought, which was a bit more relevant, which is top scientists completely getting it wrong. This story is only a couple of months old, and it was about machine learning in biomedical research. So I got thousands of scientists running thousands of medical tests doing their research and kind of running all these scenarios, and they’ve got this thing called the reproducibility crisis.

The reproducibility crisis is taking one set of data, running a whole bunch of tests on it, then letting a different bunch of people run the same tests and not getting the same result. They found out that they’ve been kind of getting really carried away with this, that their underlying data was fundamentally flawed. I couldn’t reproduce the results of the first test or the second test. The point they were making, which is probably most relevant to this, is that the software can only identify patterns that exist in the data.

I can’t identify what happens in the real world unless the data is a replication of the real world. I think that’s probably the point that really stuck to me, we are sitting on really poor quality data as an industry. That problem needs to be solved. It may not be the most fun thing to always keep talking about, but without that, not just today in terms of what people are doing in their businesses, the decisions that people are making off of data, everyone seems to be converted now, we need data to make decisions, but people are still willing to make important big business strategic decisions off of flawed data sets. That’s mad. I guess I wanted to put it in the AI context for the future, but that’s really the point that I think is most relevant about this today. So where is AI working in recruitment? And I thought that we’ll move on to that. The lecture is over. Not very many places is the short answer.

There’s a lot of people that are starting companies now. On the note, because you can go and register the domain name .AI and just put a company name before it. So there’s a number of people that are now in the artificial intelligence in recruitment space when they kind of call you up. That’s assuming it’s a human calling up, could be a chatbot. 

Bullhorn announced last year at their big event that they’re starting a shift scheduling piece of software that is designed to kind of take out all of the manual work that goes into effectively blue-collar shift working.

It’s supposed to be a bit more clever than can you do this shift? Yes or no. But it should actually be able to engage in a conversation with an individual about when they’re available, what they’re available for, and just do the whole organization. It looks good. I think they’re still kind of building that chat bot to becoming much more relevant, much more kind of widely spread. So seeing those a lot more now when people are going to support desks, certainly. I think this is going to be very big in recruitment.

One of the biggest things that people talk about trying to save time with AI is candidate screening. A lot of people are using either natural language for that, but a lot of people are using chatbots to try and do it. Has anyone got involved in trying to game a chatbot? Is that just me again? Is it Francis, help me out here. But there’s a lot that’s going on now where people think they’re talking to chatbot. We had a guy in our office the other day who was kind of on a support desk, and he was adamant it was a chatbot.

So he starts trying to game the chatbot, and it turns out it’s just a really dumb human who was then really insulted. But there’s a lot of this now where it kind of goes, ‘What salary are you looking for?’ And you say ‘I’m looking for a million pounds a minute’, and it goes, ‘Wow, that’s great. Yeah, we’ve definitely got a few jobs for you.’

We saw this stat from Mya, who does chatbots for recruitment in the states. They did a survey is 72% of their applicants thought they were chatting to a human when they were chatting to a bot. I don’t think that’s going to last very long. So it’s all right if people feel like it now. I mean, take aside for a second, the ethical, moral views on having individual humans talking to bots, these are starting to become widely spread and they won’t just be in the recruitment industry on our side of it.

They’ll be on the other side of the fence in the quarter. So there’s a chatbox. Now I saw this – This is a 60-second clip of an AI assistant booking an appointment:

So how can I help you?

Hi. I’m calling to book a women’s haircut for a client. I’m looking for something on May the 3rd.

Sure. Give me one second.

Sure.

What time are you looking for? 

Around 12 P. M. 

We do not have a 12 PM available. The closest we have to that is 1:15.

Do you have anything between 10 a. M and 12 P. M.

Depending on what service she would like. What service is she looking for?

Just a woman’s haircut for now.

Okay. We have a 10 o’clock.

10 am is fine.

Okay. What’s her first name?

The first name is Lisa.

Okay, perfect. So I will see Lisa at 10 o’clock on May 3rd.

Okay. Great.

Thanks. Great. Have a great day. Bye.

When I look at that, a lot of people are now thinking that that’s the sort of stuff where that has the biggest impact, certainly, in the market, we’re in because there are lots of calls and lots of communication coming. People are seen in a way that bots can take over. That’s all right, to begin with, but at some point, everyone’s going to have it. As soon as you start to remember when marketing automation started, all of a sudden, emails can send themselves and everyone would get bombarded with emails and they’d have to legislate for it and regulate for that kind of stuff. Well, that has to happen in the AI space, but it’s a way off. I mean, you do look at what’s going on in Parliament now and you ask them to start legislating for AI and appropriate conduct in the AI world. The thing I think that’s most important of all of this is how that relates to what we do as an industry, which is the human relationships that we have.

The more chat bots that are out there, the more calls, these things are like the Terminator. If you just go, you just call through these lists, talk to people, find out their availability, check if they got a UK driver license, see if they’re looking for work, whatever it may be to set it up and having things messaging out, they know no ends. They just do what they’re told. And if every agency has it and every employer has it and people are getting bombarded with it for the rest of their lives, they’re going to crave human relationships.

Candidates are going to want to talk to a person. Clients are going to want to talk to a person. And so I feel like as I was getting through this presentation, it just came back to me. It might just be my S Three training kicking in and sort of banging the jump for recruitment and human relationships. But Let’s face it, everyone in this room has got to where they’ve got to, largely by their ability to build relationships with other humans.

I don’t think that goes away from AI people talking about taking jobs away. I don’t think it goes away. I think it becomes more important. So whilst I think people need to be really thinking about what they do with the data and focus on it, never, ever losing a bit that we, as an industry, have always been brilliant at, which is building and developing those relationships. 

So to summarize, 

  • AI is only as good as the agent it learns from. I’m pretty sure I drummed that in earlier.
  • Automation is often proclaimed to be artificial intelligence in recruitment. So don’t forget that right now the benefits you think you can get from AI exist today with automation. 
  • And, of course, your data strategy is more important than ever. 

When artificial intelligence has really arrived in recruitment, data and human relationships are going to be paramount. And that’s it for me. Thank you very much.