Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus convallis sem tellus, vitae egestas felis vestibule ut.

Error message details.

Reuse Permissions

Request permission to republish or redistribute SHRM content and materials.

Is Your Applicant Tracking System Hurting Your Recruiting Efforts?

Some AI-based systems may be increasing the worker shortages they were built to address.

An illustration of a machine with the words system error.

Recent leaps in technology have paid significant dividends in the search for talent, where the use of artificial intelligence has saved recruiters vast amounts of time in screening, evaluating and interacting with job candidates at scale. AI-driven tools that automatically match resumes to job descriptions and chatbots that quickly answer applicants’ frequently asked questions have revolutionized the hiring industry and lifted a heavy administrative burden off of recruiting teams.

But recent studies show that a perpetual quest for new efficiencies may be exacerbating one of the biggest problems organizations face today and preventing many of those same recruiting technologies from achieving their goal: finding candidates to fill open jobs amid severe labor shortages.

In particular, new research has found that the manner in which applicant tracking systems (ATSs) and other automated screening tools are configured often leads them to reject candidates who may be qualified to fill roles but who lack certain pedigrees or don’t possess profiles that exactly match an increasingly inflated wish list of skills that recruiters demand.

Overlooking ‘Hidden’ Talent Pools

A 2020 report from Harvard Business School (HBS) and Accenture titled Hidden Workers: Untapped Talent found that job candidates who had gaps in work histories or who didn’t possess college degrees or other credentials were often disqualified by automated screening tools even though many of these individuals likely possessed the skills or knowledge needed to perform well in the roles. 

Applicants who described their skills or experience on resumes using language that differed from the requirements posted in job descriptions also found themselves on the outside looking in when automated keyword-matching technology identified them as poor fits for the roles, the report found.

Joseph Fuller, an HBS professor of management practice who led the research, says many ATS platforms are still configured to use “proxies” such as a college degree for skills, self-efficacy, work ethic and other attributes. Automated screening tools also use factors such as gaps in full-time employment as a basis for excluding candidates regardless of their other significant qualifications,
he adds.


‘The problem comes in setting up screening software in ways that search only for an exact match to job criteria.’

“The study found organizations continue to use filters in their recruiting technologies that are inferential in nature,” Fuller says. “They infer, for example, that someone who hasn’t been employed for a certain number of months may not have a good work ethic or the needed skills. There’s often a lack of screening filters based on the attributes and skills of workers who’ve proven they’re good at a job rather than those that just imply they may be good at a job.”

The Harvard report found many of these overlooked or “hidden” workers are veterans, immigrants, mothers, caregivers, high school graduates, and relocating spouses
or partners. 

Experts say these groups represent untapped talent pools that, if evaluated using a wider screening aperture, could help mitigate labor shortages across many industries. Veterans, for example, who have all the necessary skills for an open role often neglect to include those skills on their resumes, the Harvard report found. 

Recruiting analysts say that in their desire to improve hiring efficiency, corporate recruiters often use technology in ways that narrow applicant pools so thoroughly that they end up “knocking out” qualified candidates. 

“Employers can and do put requirements inside of their ATSs which essentially exclude potentially qualified applicants,” says David Francis, vice president of research and product for Talent Tech Labs, a talent acquisition technology consulting firm that is based in New York City. “There probably is room for improvement across the board with that.”

Betsy Summers, principal analyst for research at advisory firm Forrester and a specialist in human capital management issues, agrees that the problem doesn’t rest with automated hiring technologies alone but in how those systems are configured by recruiters.

“The problem comes in setting up screening software in ways that search only for an exact match to job criteria and don’t factor in those candidates who might meet 80 percent of the job requirements, have transferable skills and could do the job with some additional training,” Summers says.

The authors of the Harvard study cited what they believe to be the paradox of the problem in their report. “Our conclusion rests on the striking irony that companies consistently express concern about the availability and quality of the talent available to them, while acknowledging that their hiring processes exclude qualified candidates from consideration,” the report authors wrote. “It’s unimaginable that management would tolerate an equivalent error rate in mission-critical processes associated with operations, supply chain management, distribution or customer service.” 

Screening on Skills, Not Just Pedigree

Some recruiting technology vendors have moved to address the problem of screening candidates primarily on pedigree rather than proven skills or knowledge by modifying their existing products or creating new ones.

LinkedIn’s new Apply Connect product is designed to identify the most qualified candidates based primarily on their skills, not on college degrees earned, former job titles or past employers. When job seekers apply for a position, the automated system identifies the skills and assessments data on the applicant’s LinkedIn profile that match the job requirements and then places those skills front and center in an ATS for recruiters to see next to traditional criteria such as job title or experience.

“Our vision is to help transition the hiring market from focusing solely on titles and companies, degrees and schools to also focusing on skills and abilities,” says Hari Srinivasan, vice president of product management at LinkedIn. 


‘Our vision is to help transition the hiring market from focusing solely on titles and companies, degrees and schools to also focusing on skills and abilities.’

Among the recommendations from the HBS report for addressing the hidden-worker issue is for recruiters to shift from using “negative” to “affirmative” filters in their ATS platforms, Fuller says, to help ensure candidates are hired on demonstrated competencies and skills, not just credentials.

Negative filters screen out candidates who have the necessary work or life experience to do a particular job but who lack certain proxies that employers seek, such as a college degree or an unbroken history of employment, Fuller says. Affirmative filters, on the other hand, are more skills-based. Rather than relying on criteria like “employed in a similar role within the last three months,” they use evaluation measures such as “cumulative five years of technical sales and service experience in B2B devices” or “multiple experiences working in team settings.”  

More recruiting technology providers also are deploying a capability called “skills inferencing” that can identify skills that aren’t explicitly listed on job candidates’ resumes or LinkedIn profiles, with the goal of providing a fuller picture of those candidates and making them more marketable.

The technology can interpret language on a resume that suggests an applicant has additional skills, scan publicly available information that might uncover other skills—for example, on a candidate’s Facebook profile—or infer that since other applicants with similar backgrounds have skills the candidate hasn’t listed, that particular candidate may have them, too.

Francis of Talent Tech Labs gives the example of technology that could infer a software developer applying for a job might have additional skills not listed on a resume, based on information the technology gleaned by scouring a database of other developers who worked in the same role at the same company as the candidate. 

Jerome Ternynck, founder and CEO of SmartRecruiters, a recruiting technology provider based in San Francisco, says his company’s systems have such skill-inferencing capabilities. 

“The artificial intelligence can extract much more information from a resume than a human recruiter often can,” he notes. “It can infer additional skills from past experiences and roles and turn what is a fairly basic resume into a richer document.”

SmartRecruiters’ platform also has specific filters that recruiters can use to help highlight candidate skills or knowledge that might be overlooked in favor of traditional credentials. 

“The filters could show you, for example, those candidates with the required skills but not the years of experience you’re looking for or those with the required skills but who may not have done the specific job before, and much more,” Ternynck explains. “We try to highlight people who are objectively a good match based on their skills but perhaps don’t exactly match the profile in a job description. They may not have worked at Facebook, received a degree from Harvard or have 20 years of experience, but they could have the skills or experience to do the job well.”

The Problem with Job Descriptions

Some recruiting experts believe issues with automated hiring systems screening out qualified candidates can be traced back to flawed job descriptions. The HBS report in particular found that the more organizations continue to add new requirements to job descriptions, the more they narrow their chances of finding the talent they need. Ballooning or outdated job descriptions is a growing problem, experts say.

“Companies need to revisit and simplify their job descriptions to focus on what’s essential to a role,” Fuller says. “If you look at job descriptions across industries, the number of different skill requirements for many entry-level jobs often approaches 40 today.”

He says companies too often simply bolt new job requirements onto old ones over time; as that list of preferences grows, the number of candidates likely to qualify shrinks. Reconfiguring job descriptions so more applicants can successfully meet the requirements can help address worker shortages for hard-to-fill roles. 

In a previous Harvard report titled Dismissed by Degrees, Fuller and his team found that recruiters were sometimes “inflicting a skills shortage on themselves” by writing job descriptions to state that a four-year college degree was required even though it wasn’t actually needed to perform the role successfully.

“That study found, for example, that 70 percent of job postings for executive assistants required college degrees but only 30 percent of those currently working successfully as executive assistants had degrees,” Fuller says. “Companies often raise the threshold of requirements they’re looking for when their own experience indicates those new requirements aren’t strictly necessary to perform roles well.”

Summers agrees that how job descriptions are written and presented contributes to the disqualification of qualified candidates in the initial stages of the hiring process. Lengthy or overly complex job descriptions also can discourage people from applying, as can cumbersome forms with too many questions or steps that cause promising candidates to abandon the online application process prior to completion, Summers says.

Caitlyn Metteer, director of recruiting for Lever, a provider of applicant tracking systems and other recruiting technologies that is based in San Francisco, says her recruiting team strives to overcome some of the historical problems with traditional job descriptions by creating “impact descriptions” for job roles instead.

Impact descriptions are designed to focus more on outcomes, impacts and the motivations of candidates rather than on a rigid set of job responsibilities or tasks, Metteer says. While all of the roles Lever hires for have a desired set of skills, experiences and qualifications, recruiters try to widen their lens when evaluating candidates.

“The focus is more on the impact a candidate will make on our organization rather than on a laundry list of degrees or credentials they bring to the job,” Metteer says. “We believe by shifting that focus, it helps our recruiters think more openly about the types of backgrounds or experiences that can make for a successful employee here.”

Dave Zielinski is a freelance business journalist in Minneapolis.

Illustration by Michael Korfhage.

Federal Scrutiny of AI-Based Hiring

The use of artificial intelligence for employment decisions is receiving increased scrutiny from government agencies concerned about the potential for bias in the use of algorithms to screen job candidates and a lack of transparency with candidates when that technology is used.

The U.S. Equal Employment Opportunity Commission (EEOC) launched an initiative in fall 2021 to ensure that AI and algorithmic decision-making tools don’t create new discriminatory barriers to jobs. The agency says the initiative, the latest in a series of related actions since 2016, will look closely at how technology is changing the way employment decisions are made and will help guide job applicants, employers and technology vendors in ensuring AI is used in ways that are consistent with federal equal employment opportunity laws.

“While the technology may be evolving, anti-discrimination laws still apply,” said EEOC Chair Charlotte A. Burrows in a public statement.

The EEOC initiative follows on the heels of a move by the White House’s Office of Science and Technology Policy to create a technology “bill of rights” to protect against harmful or faulty uses of AI and biometric tools for practices including hiring decisions or evaluations of individuals’ character or mental and emotional states.

In addition, the New York City Council passed a law in December that would bar city employers from using AI-based hiring tools unless those systems can survive an annual audit proving they don’t discriminate based on job candidates’ gender or race. The law takes effect in January 2023.

The law also requires companies that use such automated hiring tools to give job applicants the option of having their skills or personalities evaluated by a human rather than the technology. Additionally, the law requires companies to provide candidates with more information about how algorithmic hiring tools work and how their data will be used and stored.  —D.Z.


​An organization run by AI is not a futuristic concept. Such technology is already a part of many workplaces and will continue to shape the labor market and HR. Here's how employers and employees can successfully manage generative AI and other AI-powered systems.