Read in: Nederlands
If you’re looking for anything recruitment tech right now, you’ll stumble across many promises of artificial intelligence. But is there really AI in recruitment?Well, that’s debatable.
Lots of recruitment tech companies claim to have an ‘AI based’ product. While often technically true, it is a little bit like the ‘renewed recipe’ sticker on a fruit salad. In the strictest sense, it's true, but often it doesn’t tell you anything.
Let’s look at some of these acclaimed applications of AI in recruitment.
- To uncover new sources of talent: to identify where talent is ‘hiding’ now.
- To decipher Frankenstein job specifications to match cv's.
To give you a sense, something like this:
- To decipher cv’s to find drivers, career paths and and other essential patterns of talent.
- To smoothen the candidate experience by answering candidate questions via chatbots.
Sounds promising, right?
Some of these applications are actually quite good. Look at automated conversations in chatbots - although we all know good and bad examples of these. Others... not so much. Yet.
So what exactly is the promise of AI?
In the most general sense it would be any form of ‘non-human intelligence’. AI will advise you to make a certain decision, or makes that decision for you.
So, we have the techniques, we have the algorithms. The problem is data.
Why? This was beautifully phrased by Spyros Magiatis (CIO at Workable) in his session at Recfest London past july. ‘AI is something you’re aiming for, it is the holy grail, algorithms working together to solve your complex problems.’
‘Keep the AI discussion grounded’
But we’re not there yet. ‘That’s why we should keep the AI discussion grounded’, he warned.
AI is real and promising, but overhyped. Because the use of artificial intelligence is currently anything but accurate. Something like 50 to 60% accurate, in the case of many early tools. And when it comes to business decisions like hiring someone, half is not enough.
And there comes the data problem again. The techniques are ready. But the data of clients which are being used, are not.
‘Barry is a superstar, I want more Barry’s’
Let’s explain this by a common, but complex recruitment case. A situation in which ahiring manager says: ‘Barry is a superstar, I want more Barry’s in my team’.
So what is holding us back to get to let’s say, 80 or 90% accuracy that our systems can actually give us ‘more Barry’s’? Magiatis gives 4 reasons.
1. The quantity of data
Our ATS and HR systems are right now the most powerful databases we have. But right now we don’t have enough data to ‘feed’ our algorithms. For this kind of sourcing we need millions of cv’s and other datasets that showcase Barry’s performance.
2. The quality of data
Recruiters don’t update their candidate data often - or at all.
That means the data is not ‘clean’ enough to predict who the next Barry’s are going to be.
Not even taken into account that cv’s not always represent the truth. Magiatis: ‘In most cases you don’t really know anything about your candidates.’
That makes it difficult to get accurate results - fake data give fake results.
3. Subjectivity about 'The Perfect Barry'
Problem: the ideal candidate for me is not the ideal candidate for you.
You set a specific problem for AI to solve, and you aim at a specific outcome. You need to first understand what ‘perfect’ looks like in terms of skills and experiences. What makes Barry’s success? That's different to any hiring manager.
4. Fragmentation of data
Data currently come from many different HR tools. Some data from your ATS, some from your performance management system, some candidate data from social media, etc. This impacts the quality of data - which means, less accurate results.
Many talent acquisition departments are currently still working to get their recruitment data right.
When are we ready for AI then?
In other words: when the data are not ready yet, it doesn’t matter how advanced your AI- techniques and tools are, the results will give a low accuracy rate.
50 to 60% is quite common now. A rate like this is fine when you are Amazon: a 50 to 60% chance that the recommendation for buying other products, is exactly to your liking.
A bad buy on Amazon is surmountable. But when you’re talking Barrys, high costs per hire for a false recommendation are the price to pay.
But wait, is there AI in recruitment or not?
Yes, there is.
But to accurately predict patterns of talent? We still need a couple of years before our data - and we ourselves! - are ready.
Despite some of the overblown promises by tech companies, there are definitely some really promising technologies being used in recruitment as we speak. Simple forms of chatbots that automate frequently asked questions from candidates about the company or the application process.
Is that it? No.
The next steps are real and promising as well.
Complex chatbots are tackling the problem of candidates wanting more specific information about their own application. Higher intelligence robots are capable of looking up such personal information. From more general to more personal answers.
At Endouble we are working to integrate AI and machine learning into our product offering to help recruitment professionals become more efficient. Taking into account the user experience of these bots of course, and really addressing the challenges of your company. So that ‘AI tools’ don’t become your newest gadget, but a real solution to your company’s recruitment problem.