Automated hiring software is mistakenly rejecting millions of viable job candidates

Illustration by Alex Castro / The Verge

Automated resume-scanning software is contributing to a “broken” hiring system in the US, says a new report from Harvard Business School. Such software is used by employers to filter job applicants, but is mistakenly rejecting millions of viable candidates, say the study’s authors. It’s contributing to the problem of “hidden workers” — individuals who are able and willing to work, but remain locked out of jobs by structural problems in the labor market.

The study’s authors identify a number of factors blocking people from employment, but say automated hiring software is one of the biggest. These programs are used by 75 percent of US employers (rising to 99 percent of Fortune 500 companies), and were adopted in response to a rise in digital job applications from the ‘90s onwards. Technology has made it easier for people to apply for jobs, but also easier for companies to reject them.

The exact mechanics of how automated software mistakenly reject candidates are varied, but generally stem from the use of overly-simplistic criteria to divide “good” and “bad” applicants.

For example, some systems automatically reject candidates with gaps of longer than six months in their employment history, without ever asking the cause of this absence. It might be due to a pregnancy, because they were caring for an ill family member, or simply because of difficulty finding a job in a recession. More specific examples cited by one of the study’s author, Joseph Fuller, in an interview with The Wall Street Journal include hospitals who only accepted candidates with experience in “computer programming” on their CV, when all they needed were workers to enter patient data into a computer. Or, a company that rejected applicants for a retail clerk position if they didn’t list “floor-buffing” as one of their skills, even when candidates’ resumes matched every other desired criteria.

Over-reliance on software in the hiring world seems to have created a vicious cycle. Digital technology was supposed to make it easier for companies to find suitable job candidates, but instead it’s contributed to a surfeit of applicants. In the early 2010s, the average corporate job posting attracted 120 applicants, says the study, but by the end of the decade this figure had risen to 250 applicants per job. Companies have responded to this deluge by deploying brutally rigid filters in their automated filtering software. This has had the effect of rejecting viable candidates, contributing to the large pool of job-seekers.

The use of this software has become a huge business in itself. As the report notes: “Over the intervening years, automation has come to pervade almost every step in the recruiting process: applicant tracking systems, candidate relationship management, scheduling, background checks, sourcing candidates, and assessments. The global recruitment technology market had grown to $1.75 billion by 2017 and is expected to nearly double, to $3.1 billion, by 2025.”

Despite this, companies seem well aware of these problems. Nearly nine out of 10 executives surveyed for the report said they knew automated software was mistakenly filtering out viable candidates, with some saying they were exploring alternate ways to hire candidates. But, as the study’s authors note, fixing these problems will require “overhauling many aspects of the hiring system,” from where companies look for candidates in the first place to how they deploy software in the process.

Correction, Wednesday September 8th, 10:42AM ET: A previous version of this article incorrectly referred to one of the authors of the study as Joseph Miller. The correct name is Joseph Fuller. We regret the error.


The ‘floor-buffing’ criteria seems to suggest this is a crass human error in listing the qualities necessary for a low-skill job. The first rule of data: Garbage In, Garbage Out. There may be all sorts of problems with the software, but adequately listing the right criteria for a position is a basic human error, no?

Saddens me that everyone is aware how imperfect this system is, meanwhile rents don’t get paid and food needs putting on the table…

yeah, I mostly agree with this take, but I will say, there’s a chance that the system is designed in a way that encourages Garbage in. For example, if it asks you to enter tasks that the job entails and then subsequently uses those as a hard "necessary criteria in order to be considered for the job", for example.

It is some of both. A poorly designed interface encourages people with little domain experience to choose the wrong criteria or too stringent criteria, and a software engineer who does not understand how hiring works is likely to make a poorly designed piece of software.

For anything except the most menial jobs, some understanding of the job itself is necessary to make a hiring judgment. A sort of generic sorter that applies to all jobs based on something like production rules gives the impression of doing something that it doesn’t do. And as an HR person with a social science or business degree, you probably would not understand that.

The issue is, in the pre-internet era if a company wanted to make a job posting, they would need to examine each applicant’s CV by hand, in paper form. The posting may have still listed "floor-buffing" as a requirement but the fact that a human is actually reviewing these CVs means that one missing keyword probably won’t be a deal breaker, because a person decides whether an applicant is promising based on their job history and experience, not whether their CV passes some keyword validation check.

so, basically, the issue is not that automated hiring software is mistakenly rejecting candidates but more that the HR people using the software are either asking for needlessly high requirements for low-skill jobs or simply bad at their job?

the HR people using the software are either asking for needlessly high requirements for low-skill jobs

Now hiring! Looking for a junior/entry level software developer! Come work in a fun and challenging environment where you can grow your skills and work with a great team!


  • MS or PhD in Computer Science or Computer Engineering
  • 10 years work experience designing, developing, and shipping high quality software.
  • Proficiency of Java, C++, C, Javascript, HTML, CSS, Angular, Ruby, Python, Django, PHP, Coffescript, Go, Scala, REST, JSON, Swift, Kotlin, and C#.
  • Experience in network programming including RPC, Client/Server, and Web Technologies preferred (HTTP, Web Services)
  • Strong math skills including Advanced Linear Algebra, Discrete Mathematics, Trigonometry, and Calculus
  • Mastery of version control systems (Git preferred)
  • Proficiency with full stack .NET Core Development for web services
  • Proficiency with Relational DBs and/or NoSQL (e.g. MongoDB, MSSQL, MYSQL)
  • Experience with cloud technology stacks from Microsoft Azure, Amazon AWS, or Google

Starting salary $40,000! Apply now!

I wish I could say this is an exaggeration, but it is not…

most import one: paid by exposure.

Having been seeking a career change for a while, I can attest to the accuracy of this and it’s really discouraging. It’s pretty spot on, and it’s why I pretty much gave up on applying anywhere. They make their requirements so high that despite having worked practically my whole life I’ve got no hope in hell of ever getting hired anywhere.

Hey, the free market is short of workers. Should we offer higher salaries, which is apparently what should be happening, according to the textbooks? Na, there’s still a chance we won’t have to, so let’s first plough billions into better AI to solve the edge cases.

If these positions had few workers applying for them, they would not be using AI to screen applicants. These software packages are typically used for positions where a company may have hundreds or even thousands of applicants for a position. A situation like that legitimately places downward pressure on salaries, not the opposite.

That’s charmingly naive. For HR people, it is a case of "if the only tool you have is a hammer, every problem looks like a nail."

I would counter that you have clearly never had any experience in management. HR people are people, just like you and I. Further, even if, as you suggest, this is all the fault of poor HR staff, it does not change the absolute fact that these programs were created and instituted to deal with large volumes of job applications (which is the topic I was actually responding to).

No company receiving 1-2 applications for an open position (as the OP was suggesting, and claiming this should be causing salaries to rise) would spend money on AI application filtering. That is just the way it is. The original assertion is a false one.

It’s not. The article itself lists two examples – a hospital looking for people to simply input patient data into computer databases, and a retail clerk position. Those are pretty much menial jobs that only require a bit of common sense and you’re good to go. Instead, needlessly specific computer algorithms were used. So what if there’s 200 applications – if you haven’t got time to read them all, either skim them to filter out stupid ones, or just grab 50 random ones and go with that.
The increasing over reliance on computer algorithms for things humans do far better will be the death of our civilisation. I’m sorry, but it will be.

I think we need a recap.

Chiba, the original poster, stated that instead of using AI algorithms companies should be increasing their wages. I countered that a company receiving so many applications for a position that they need to use AI algorithms to filter them is unlikely to face any pressure to raise wages, because obviously that position is attractive to many people.

Now MHA71 is saying that is incorrect, because…some of the jobs do not require a lot of expertise?

I do not actually disagree that over reliance on AI algorithms can be to our detriment, but honestly that has very little to do with the original discussion you are replying to.

Another recap: the headline in one of the source articles to which this article linked began with "Companies need more workers", and the body began with "Companies are desperate to hire".

if there’s 200 applications – if you haven’t got time to read them all, either skim them to filter out stupid ones, or just grab 50 random ones and go with that.

Two issues with this: scale and compliance.

1. Reviewing 200 applications for one role is just about doable. But take the example in the report of a healthcare provider. They might be hiring for 200 positions. If each one receives 200 applications, that’s 40,000 applications to review. Try doing that without software.

2. Most employers cannot discard job applications or choose a shortlist or final candidate at random. Employment regulations in many countries require employers to review all applications and to base shortlisting and hiring decisions on clear criteria, to avoid potential bias against protected classes.

If a company is hiring for 200 positions, it’s unlikely to be one or two people doing the actual hiring process. And again, sift through them quickly to filter out the ones that are not qualified – and if no qualifications are needed, then the sifting through and picking out some is even easier and quicker.
It may sound like I’m telling HR people to do their job – but my point really is that using computer algorithms to do this is a horrendously bad and almost evil-level wrong way of doing it.
I’m sorry – but there is no way that a computer can do the hiring process or the "reading applications process" anywhere near as well as humans can. The only reason companies are doing it is that it’s cheaper. It’s wrong, it leads to bad results – but not saves the company some money and there are plenty of applicants out there so they don’t care. But we normal people should care. And we shouldn’t accept it.

oh noes we are so poor as a country we can only do one thing and not both

Professional resume creators are probably the most qualified for any position.

Over-reliance on software in the hiring world seems to have created a vicious cycle.

Emphasis mine. The problem is not (necessarily) the software. It’s its users who just use it as a crutch to not do their job properly.

It’s like drivers using Tesla Autopilot to play games or watch movies… Keep your eyes on the road. Do your fucking job. These tools are only that, TOOLS. You still have to put in some actual work for them to function as intended…

We are surrounded by technology and algorithms, and we often barely understand half of it. This isn’t good.

I experienced this out of college 15 years ago, when multiple online application processes required entering my GPA. The automated cutoff appeared to be a 3.0 GPA. I had a 2.9, but in one of the harder majors (EE) at one of the most difficult engineering colleges where the average graduating GPA in my major was 2.7.

So yeah, this has been a problem for a while and gets even worse when lumping in other filters without actually talking to a candidate.

That’s always a very real problem. In the real world the school you go to impacts the value of the grade, a low grade from a good school can often be worth the highest marks at a so so school. It also doesn’t account for what else a person does, someone that gets perfect grades with little practical experience or social skills is often less capable than someone that just did okay, but packed their days with extra curriculars and outside activities.

I’ve had similar issues with my original degree to the point I went to get a Masters just because generally that means your first degree doesn’t get asked about. And it only ever was an issue on those big multinational employers anyway or people that use off the shelf HR products. In the real world though I’ve actually done a lot better and my career has moved much faster than the majority of the people I’ve known that have gone that route so its not the end of the world to be excluded from the corporate grind and get in with smaller employers and direct entry to better scoped roles with an actual hiring manager thats a person not an AI.

As a total aside though GPA is such a mad metric, the UK/Commonwealth common marking system for university level education is so much better as it actively bell curves at the average and so only the truly exceptional get the equivalent of a GPA >3. The drive to to have a GPA of 3-4 be achieveable by more people only serves to dumb down teaching standards, I remember on my MSc in the UK some US students actually breaking down in tears when an assignment that was by their standard an A grade only coming back with a 60-65%, which was all it was worth under common marking scheme and the same thing as an A on a US marking system

Mine wasn’t the literal best ones but top 50 and I experienced getting auto denied from companies when I had a fucking 3.7 GPA

And here I am with no degree or anything; my chances of getting hired anywhere are probably slim and none.

View All Comments
Back to top ↑