'If HR is using a system that's even two years old and hasn't been updated … they're using systems that can't handle the mass mailings': AI hiring experts offer tips for better hiring
The rapid adoption of artificial intelligence (AI) use by job applicants has led to a surge in application volume, with many HR teams struggling to keep up.
A recent report by Remote found that businesses now spend an average of 9.24 days sorting through job applications for each opened role, many of which are irrelevant or misleading due to AI-assisted resume generation.
Plus, 73% of businesses say they are receiving AI-generated resumes containing false information.
It’s an “arms race,” says Terri Griffith, professor of innovation and entrepreneurship at SFU Beedie School of Business — and human resources departments are losing the battle.
“What I almost want to say is the HR groups jumped in early [with AI], and those tools have now been superseded by the applicants’ tools who came along a little bit later,” says Griffith.
“And so now, I think the HR leadership is going to have to leapfrog and perhaps upgrade. If they're using a system that's even two years old and hasn't been updated to take account of shifts in applications, then I'm going to say they're using systems that can't handle the mass mailings.”
AI job applicant screening – a catch-up game
A 2024 Robert Walters survey also found that while 66% of hiring managers can detect AI-written applications, 54% say that the sheer volume is slowing down hiring.
The concept isn’t new – AI has been integrated into recruitment processes for years, particularly for resume screening. However, its capabilities have changed drastically with the emergence of large language models (LLM), says Irina Mocanu, senior talent management and organizational development advisor at Workleap.
“Before LLM, AI rule-based filtering couldn't identify nuances,” she says.
“You might be rejected for an HRBP role if there was no mention of HRBP in your CV, even in a case where you had relevant experience that would qualify you for a role. This changes completely with genAI, because the technology is better at understanding context, nuances and even sentiment in content.”
However, Mocanu echoes Griffith’s point that HR screening technology isn’t keeping up with what applicants are able to do with genAI. Not only that, but HR also needs to up its game to use the tech it does have, optimally.
Source: Remote
“Generative AI requires guidance – it needs to be ‘taught’ what works for your organization to be effective. That means you need to define what’s important and configure your AI technology to understand what success looks like in your context,” says Mocanu.
“For example, this could involve outlining the core behaviours that represent high performance in your organization. When you do this, AI can help screen candidates more effectively, by focusing on what matters to your organizational culture and business objectives.”
Even with this more sophisticated screening, qualified candidates may still be overlooked, Griffith says, suggesting that clearer job criteria and improved filtering systems could help organizations cut through the noise.
“Having the higher flow is good, as long as you can still identify those people who have the capabilities that you need,” she says.
“So maybe it forces the organizational folks to be clearer about what it is they need, and make sure their systems are really scanning for the things they actually want.”
However, Griffith warns of going too far in the other direction. In attempts to pare down to specific qualifications to find ideal applicants, HR leaders may be raising the bar too high for all qualified candidates to make it over.
Beyond the resume: shifting to work samples
According to Griffith, employers have historically over-relied on resumes as a measure of candidate quality, and AI-assisted applications have highlighted long-standing flaws in resume-based hiring: “Looking at a resume has never been the best way to decide if someone's appropriate for the job.”
Instead, she suggests incorporating work samples or work tests into the hiring process, such as Automattic (the firm behind WordPress), which pays candidates for test projects, ensuring that screening processes remain fair.
“Looking at work product is a much better strategy, and finding ways to be able to do that at a higher scale would make sense to me,” says Griffith.
“If I'm going to go to all the trouble of updating my platform that's doing the screening, I want to update to one that allows me to have work samples as part of that … that would certainly be difficult, but not insurmountable.”
Reduce reliance on job postings
There are other, more creative ways for organizations to hire that don’t involve mass amounts of CVs, Griffith goes on, pointing to Google’s “hackathons” and Stack Overflow blog contributors being picked out for hire as examples of companies looking outside of the resume inbox to find talent.
“They would combine that with LinkedIn data, so they used that analysis, that deep analysis, to identify people who hadn't even applied.”
However, this strategy is vulnerable to bias, as performing work for job applications may be prohibitive for applicants who already are working or have caregiver duties, for example, she says.
“When people do use work tests or work products, they have to be very careful to not make that level of effort be so onerous that it really should be paid.”
Hiring the person and their tools
As AI-assisted job applications become the norm, Griffith suggests a shift in hiring mindset: rather than seeing AI-generated content as a negative, employers should assess how candidates use these tools strategically.
“I should be hiring the person and their tools,” she says.
“If I were going to hire a craftsman, a carpenter, to work on my house, they need to have high-quality tools to be able to do the best work. I want someone who knows how to use those tools to present the work in the best possible way. Why would I tie their hands behind their back, before letting them do that work sample?”
Instead of banning the use of AI in job applications, Griffith recommends allowing it, but then requiring applicants to disclose which tools they used, and how they used them.
“Give credit to your tools and then show me the work, and then that's going to be something that's a useful conversation in an interview,” she says.
“How did you work through this problem? What tools did you use? What tools would you expect to have available in our organization, and how much freedom do I have in deciding what tools to use in this particular job? Those are great interview questions, and they're things that the employee needs to know, right?”
Mocanu also notes that AI can improve candidate experience by reducing wait times in the hiring process and answering common applicant questions, helping both candidates and recruiters make better decisions.
“The recruitment process can be really stressful, and waiting time often adds to that for job seekers. AI can help accelerate the process, which can reduce the time it takes between applying to a role and receiving a response,” she says.
“That can have an important, positive impact for job seekers.”
The path forward for AI job applicant screening
Calling AI-assisted or AI-generated job applications the “genie-out of the bottle”, Griffith admits that even with all of these measures, the challenges involved with hiring through job posting platforms and postings are not going anywhere in the foreseeable future.
For this reason, she advises employers err on the side of acceptance by embracing the necessity of “random selection” from a large group of equally qualified candidates, rather than attempting to pare them down and inadvertently screening out the best.
“We really do have to take a step back and say, ‘How do we engage with the people who are going to be the best for this position? And how do I find that person?’” says Griffith.
“And maybe it is [about] being more proactive, rather than the reactive approach of having people apply … there is so much information about us, that that more proactive approach may become the thing, and then it'll be back on the applicant side, like, ‘Oh my god, I'm getting calls from too many headhunters. How do I filter all these calls?’”
