On ATS Systems and Corporate Laziness
Or the Ideal Candidate Fallacy
There’s a new old adage among job seekers: “when applying, beat the machine”. When applying to corporate positions, there is rising speculation that it is a machine, rather than a human, to judge your CV passable and worthy of the interview stage. As I have been (more than once) a jobseeker myself, I made it a mission to understand whether this is indeed true or an exaggerated version of reality. As usual, truth lies in the middle, although the use of AI systems in hiring decisions gave rise to a rather disturbing vicious cycle in the development of labour dynamics.
“I’ll get you hired, just pay me first”
During my idle research of the existence of such systems and how they operate, I stumbled upon a video of a former Google, former TikTok, former-all-the-big-names-you-can-think-about recruiter who stated how ATS programmes are nothing more than glorified databases.
The claim here is that the “myth” that ATS systems bear the responsibility of your CV being thrown in the trash is the fault of companies that sell anti-ATS measures, other pieces of software that scan your resume and provide you with a machine-proof version (against payment of course). This is not false per se: there are companies, such as Jobscan or TopCV, that promise ATS-proof resume rewriting after a scan they perform with their blackboxes to yield some sort of potentially true results.
To reinforce her case, we also see how one of these ATSs, Lever, works, and how it only stores applicant information to be subsequently handled by a human. She then proceeded to include in the description that her services will help you, poor job seeker, to find a new job against payment of your already strained resources.
However, although Lever does indeed look like a glorified database, it does not represent the entirety of ATS systems out there. Indeed, this is a case of traditional ATS systems versus newer generations of AI-based ATS systems which do include an AI component that discriminate based on specific keywords and training that the AI received.
When talking of discrimination from machines deciding on your future position, you can bet the ugly side of systemic race- and gender-based prejudices play an important role too. There is a vast literature out there that tackle this issue. For instance, juridical cases such as Mobley v. Workday and Harper v. Sirius XM or a this Cambridge study on AI biases in CV scanning are a good starting point.
In this article, however, I want to instead focus on a different, if not foreshadowing, issue brought by the employment of AI-based judgement in hiring: that of corporate laziness and the self-imposed, idealistic misconception that what comes out of a robot is absolute truth.
“Woe is you, poor jobseeker!”
So, to recap: in front of us we have an army of qualified candidates for a handful of positions where new artificial gatekeepers deem the worth of your career history or degree. But how do these systems actually compute their judgements and why are companies and recruiters so eager to use them?
I’d say the answer to the latter question is quite easy, it’s the same reason why most of us now ask ChatGPT or Gemini to write emails for us: it’s just easier. It absolves us from the daunting task of having to think. The former is a mix of technicalities and the trajectory that labour dynamics between candidates and private sector jobs is unfortunately taking.
First off, to prove the existence of such systems, I went slightly deeper into the belly of the beast, where ATS systems live and develop. My research started from a newsletter piece I received from PintpointHQ, a quite heartfelt and heart warming piece on skills-based recruiting, the importance of recruiting based on skills rather than titles and degrees.
Indeed, a snappy and well-intentioned piece by their VP of Talent does mention this absolutely innovative idea to quit hiring based on degrees and the disruptive notion that a candidate who studied a specific subject could develop skills in other fields. Something that, I have to say, is almost genius, how come nobody ever thought of it before?
I was almost moved to tears at the realisation that people are multifaceted and multi-skilled: the mention of a McKinsey research on this incredibly inspired notion cannot but further strengthen the utility of such original minds and the gems they birth in their ingenious efforts. It does not matter that, apparently, 80% of companies are categorised as “In Name Only” or “Backsliders” when it comes to hiring based on skills: this sort of Enlightenment transcends practice. Woe are the poor jobseekers in the depressing labour market… But at least you have great minds thinking of you somewhere in a glass office!
Indeed, PinpointHQ does provide its own generative AI ATS system the possibility to focus on a candidate skills: their AI does provide scorecards and it does deem the CV of a person worthy based on a specific skill set the machine scans and reports back. Who decides these skills and how the machine is gauged is a mystery, however.
In the words of their VP of Talent:
The irony is stunning. At the exact moment when demographic shifts, AI disruption, and talent shortages demand that we expand our aperture for talent, critics want to shrink it back to these failed credential filters.
I agree. However, a couple of questions still remain unanswered: what are the filters PinpointHQ is giving their clients as part of the AI? Is the Education they so battle one of these filters? Are they really going to rescind contracts and lose revenues if their clients use PintpointHQ not as their VP of Talent intends to?
“Are you sure you hit the bullseye?”
PinpointHQ is, of course, not the only one: in the spirit of understanding whether there was indeed some sort of AI decision-making effort behind the choice of a CV, I also moved on trying to understand how ATS companies branded their products and what sort of technology they aimed at providing their customers with.
What I found is a mix of both traditional, non-generative AI ATS systems and aggressive AI decision-making tools — although the more traditional are apparently willing to get in the game to avoid losing market share.
Get Marco Rossi’s stories in your inbox
Join Medium for free to get updates from this writer.
I must also, at this point, apologise to the poor sales reps that took the calls to demo me their products thinking I was an interested customer that really, really needed AI to make my bed in the morning. They were super professional and did their job amazingly. I am only sorry I won’t be part of their winning streak.
In any case, their invaluable (albeit involuntary) help led me to fill in some gaps candidates won’t find in these companies’ corporate articles:
- BambooHR does not, at the moment, provide generative AI to their customers. Theirs is an HR platform where hiring is the initial part of the employee shelf life with little use for an AI tool that wouldn’t resemble a sort of chatbot, just a little cleverer. However, plans to include an an AI decision maker are in motion in BambooHR, apparently. Not sure whether I was told this to incite me to buy the product with my fake company or not, though.
- Greenhouse, instead, goes in the other direction and even advertises itself as a “Leading AI Hiring Software” (Author’s Note: ad on google accessed as 04/12/2025). They do provide an AI adviser on which CVs should move forward according to scorecard attributes and keyword suggestion. They don’t hide any of this too: you can find a full list of the Greenhouse AI ATS tools in the support articles of their Talent Filtering section.
- Workable, on the other hand, is a bit sneakier. A list of their AI tools can be received only when requesting a demo, as part of a larger marketing brochure. In the demo I received from an incredibly professional sales rep I managed to see how the information scanned from CVs is given a percentage score to gauge suitability through semantic comprehension. Scores are based on millions of other CVs the AI is trained on. Recruiters and HR personnel can also filter by scored points, effectively discriminating CVs the AI did not deem suitable based on its own comprehension of the keywords and historical training. Moreover, users can also let the Workable AI write job descriptions, further empowering a vicious cycle where the AI idealises the perfect candidate and looks for it an a CV tailored to cater its expectations.
- Avature is another ATS SaaS that advertises AI as part of their portfolio of tools. Although I did not manage to get a demo from them, corporate literature on their agentic AI is plenty both on their website and their LinkedIn newsletter, The Talent Shift. Avature’s approach to AI is branded as humane and both American and European certifications are in place for their systems, even going as far as to say that “just because you can use agentic AI doesn’t mean you should”. However, how much this is because they really have to, since they work with governments (e.g. UK’s GovRecruit), cannot be ascertained.
- Oracle and its standalone Taleo ATS system, with the tech giant trying hard to integrate it its Oracle HCM Cloud wider service, used to be an absolute gold standard of ATS systems, with large companies relying on Oracle’s branded technology to make relevant decisions such as hiring someone to fill in a position. Apparently, it is also dreadful to use, as is the case with many incumbents in the market.
What job seekers are facing, hence, is a mix of tools that effectively flattens their experiences as they need to resemble as accurately as possible the experiences of millions of other people, in a sick standardisation of their profiles.
Nevertheless, my intentions are far from those of a luddite rant: however pervasive the degree of the decisions made by the AIs themselves, the vicious cycle job seekers are caught within is not entirely due to the use of AI as much as it is from this new utopian research that recruiters and HR staff are given (or ordered) to find a candidate that would satisfy only and no more than what is required by the role as written down in the great book of theory, but that doesn’t necessarily apply to practice.
The very same humans who could not do their job when they started are now requiring new levees to not only already know the ins and outs of the role they’ll embark on, but not to be nothing more than that too.
Keyword: Risk Aversion
“But why, dear Author? Why can’t we even get an interview or the chance to discuss our experiences?” you might ask. Well, dear reader, there is no simple answer. But I believe it has to do with one big issue we face nowadays with institutions and private companies alike: reduced tolerance to risk.
Or I should probably say reduced tolerance to incur risk. If we think about it, retail investment portfolios are on the rise. When provided with a safety net (such as a regular salary), we individually want to take some risk. But the risk of losing that safety net is now too harsh than it was when another job could be found in matter of weeks.
Combine it with a dedication to AI such as that the Greeks had for the Oracle and we are presented with a mix of the search for an idealised profile and a sprinkle of good ol’ laziness. I mean, why do your job when a machine can, am I right?
In this, we are witnessing the death of transferrable skills. A machine would not have the sensibility (yet) to understand how a set of skills acquired in another industry or role can apply to another. Rather, it keeps on the lookout for the most accurate match in terms of role title, industry name, and coded skill. For instance, some systems use the codification of the ESCO Classification to help the AI scan for specific tags.
It all boils down to the fact that if you didn’t have the foresight to do that job already, you can’t anymore, as recruiters and in-house HR will never take the risk to make a wrong decision and push for a candidate they are not 100% secure they will do that one crucial, hyper-specific, alienating task candidates should have already done if they really wanted to work in the breath-taking company they applied at.
So… Are we baked?
When talking ATS systems, I do intend to challenge their accuracy. I am sure modern AI systems are accurate in their blackboxes and the results they yield fit the request — for instance, let me refer to SMART ATS: Intelligent Resume Screening Using Machine Learning by Vivek M., Dhivya V., Rahul B. Asangi, and Sanketh L. Reddy for a short read on the effectiveness of ATS systems.
Moreover, I have to specify that this article lies within the experience of many job seekers I had the pleasure to speak with and my findings in the tools ATS companies provide within the European and American market. Whether this experience is shared on a global scale, I truly hope not, although the sad state of affairs we are all going through.
In any case, what presented above is very disheartening, there’s no doubt about that. It is as disheartening as it is not talked about enough, in my humble opinion.
We are witnessing the development of a system where jobseekers have to find new ways to be given a chance and to be accepted for their professional choices, for having played the game and desperately seeking some kind of closure to their efforts.
A short-term solution may be that of legally requiring ATS companies to disclose how their AI systems assign scores and advise on CVs. That is, XAI: explainable AI. However, not only this would incur friction against the companies’ “trade secrets” on how they programme their software tools, but would not necessarily monitor the use of these systems in the hiring process too. After all, we already know the results of their systemic use.
At the risk of sounding too simplistic or naive in my vision of the future, I believe this vicious cycle of recruiting idealism and oracle-like AI decision-making to be an extreme application of the Peter Principle. The majority of employees who can prove they already carried out the pre-fabricated list of duties the AI-written job description asks for will be great at doing exactly that. Once career progression is required, we will in time see the shortcomings of hiring based on mono-dimensional skill sets and not on diversity of talent, especially on operational performativity.
And this won’t come without a blood bath in terms of loss of revenues for the single companies and its further impact on the larger macroeconomic ecosystems.
