You may not be familiar with the acronym ATS, but you’ve certainly been confronted with them, and if you were, you might have a grudge against them.
In recruitment, ATSs (Applicant Tracking Systems) have become indispensable tools, first for businesses managing a large volume of applications, then for all sizes of business.
Unfortunately, there’s a joke that won’t work here since in French ATS could stand for “Very Stupid Assistant”. That ruined the original title of the post.
Well, these software programs sort and analyze CVs and cover letters received as part of a recruitment process.
So, if you’re not yet aware of this, we’d better say it in black and white: if one day you speak to a human in a recruitment process, it’s because a machine has found your CV worthy of interest and passed it on to them. And if you receive a rejection, especially a fast one, there’s a good chance that a human will never have read it.
But ATSs are not without their critics, and there’s a certain hypocrisy about them. Many people who promote them on social networks, often to display a certain modernism, are more mixed in private.
Indeed, their ability to identify transferable skills– those skills that enable a candidate to move successfully from one sector or role to another – leaves much to be desired.
Ultimately, off the record, many HR practitioners are in fact divided on the subject: should ATS be seen as a valuable aid or a hindrance to talent diversification?
ATS are not good at transferable skills
Despite growing sophistication, ATSs still struggle to produce qualitative textual analysis.
ATS use keywords too much
They use keyword-based algorithms to filter CVs. Unlike a (good) recruiter, who will use his brain and deduce things from reading your CV, use his experience, even deduce things from the unsaid, the ATS simply looks to see whether or not keywords are present in the document.
Yes, the ATS is stupid. It even says so in its name.
A bit like SEO for website or article referencing, which strips them of all authenticity (How the web impoverishes its authors and readers), an ATS-optimized CV is cold, dehumanized and unpleasant to read. But what does that matter? Just as a website is designed to be read by robots before they bring in humans, a CV is not designed to be read by humans. Too bad for what is perhaps the most human part of the HR function.
All this would be perfect if it worked. But as it turns out, this overly keyword-centric approach often excludes talent with atypical backgrounds.
For example, a project manager in the cultural sector may be passed over for a similar position in tech simply because he or she doesn’t use the same technical terms, even though he or she possesses many transferable skills. If you insist on your project management skills when the ATS wants a project manager, you may also be overlooked.
It is said that ATS exclude up to 50% of relevant applications for multi-skilled positions.
The ATS doesn’t understand context
Unlike a recruiter, an ATS doesn’t interpret a skill in terms of its professional context.
Managing 5 people in a start-up has nothing to do with managing cross-functional projects in a multinational, but the ATS won’t see the difference.
Here again, only the human brain, which doesn’t function solely by keywords, is capable of making the difference, but for the CV to reach it, it would have to have passed the ATS test.
The ATS is uncomfortable with soft skills
Behavioral skills are essential in many positions, but they often elude the ATS.
The reason is simple: these skills are often expressed implicitly. The human brain can deduce them, not the machine.
Bad luck for candidates: if 85% of the skills needed for career advancement are softskills and not hard skills, they will remain invisible to the ATS.
The ATS loves algorithmic bias
The ATS is just a machine that learns from historical data and foolishly perpetuates traditions, which can be detrimental to atypical candidates, such as those who have changed careers or taken extended leave.
They may reject women returning from parental leave because they don’t fit the linear criteria, or exclude them from technical jobs, favor people who have had a “vertical” career path rather than a transversal one, and so on.
On this point, the recent addition of AI capability in the ATS has not solved anything but just added new problems to existing ones.
ATS do not use skills frameworks
Standards do exist for linking skills to contexts (ESCO (Europe) or O*NET (USA)), making it easier to identify transferable skills , but they are rarely used by ATSs.
As a result, relationships between similar or transferable skills remain invisible to the algorithm.
Recruiters are divided
As I said earlier, ATSs make recruiters schizophrenic. In public, they can only say how much this technology allows them to do a better job by spending, as the saying goes, more time on value-added tasks.
Added value for them or for the candidate?
In private, the discourse is more nuanced.
While they are unanimous in saying that ATSs are useful for managing the volume of applications, they often regret their inability to detect transferable skills and fear that this is a brake on the diversity of talent.
ATSs are excellent for fast filtering, but they don’t replace human intuition. It’s good to recognize this, and it confirms the well-known fact that the best is neither human nor machine, but AI, which makes good use of the machine. But what’s the point of ATS if the recruiter can only exercise his intuition on those CVs that he has filtered?
There’s another criticism I found pertinent and full of common sense: “ATS favors eligible candidates over high-potential ones”.
Can we save Private ATS?
Please don’t think I’m on an anti-ATS crusade. I recognize the tool’s potential and indispensability, but I can’t ignore its perfectibility.
And ATS editors are well aware of this, and are working to improve their product.
The first possible improvement is, of course, the integration of artificial intelligence. Natural language processing and machine learning tools capable of analyzing the context and relevance of skills.
For example, a candidate who has “managed projects” in the humanitarian field could be identified as relevant for logistics management positions.
But beware, as mentioned above, AI is a slave to the data with which it has been trained. If you’ve always recruited white males over 50 who graduated from a limited number of schools, your AI will recruit white males over 50 who graduated from these schools and, worse, will restrict the chances of other profiles being selected.
So you’ll need to train the AI with “fictitious” cases that won’t be representative of the past, but of the future you’re looking for.
ATS can then be coupled with standard frameworks to improve recognition of versatile talents.
You can then work on filters to search for general aptitudes rather than very specific technical skills.
Last but not least, we can also improve the training of recruiters, not all of whom are very good at fine-tuning ATS parameters.
A final way of debiasing ATS is the use of behavioral assessment tools as a complement to ATS (Recruitment: stop calling personality tests tests). But these tests are generally used on candidates who have already been selected, once the ATS has already committed its misdeeds.
Bottom line
To illustrate my point, I’m going to tell you about someone you probably don’t know: Pierre Labasse. In his time, he was one of the best internal communications directors in France, if not the best. He was head of internal communications at Danone during the Antoine Riboud era, and is the author of many books, including the excellent “Un patron dans la cité”.
I say this because Pierre Labasse was a historian by training. Would an ATS have given him the chance to become the internal communications director he was?
That’s what worries me.
Despite their real potential and the fact that they’ve become indispensable, ATSs are still immature when it comes to recognizing transferable skills.
It’s a pity, because if there’s one single benefit I can see in using technology in recruitment, it’s that recruiters are limited by their own experience.
Recruiters should therefore not trust them blindly, but use them as assistants, keeping an eye over their shoulder to prevent a nugget from slipping through the net.
Image: recruiter robot by Andrey_Popov via Shutterstock.
