Both employers and candidates are turning to AI in the recruitment process
A leading workplace academic has warned of an "arms race" between between employers and job seekers as both groups accelerate the use of AI in recruitment.
While the technology promises more efficient screening of applications, it also allows candidates to draft specific CVs that can make it harder for employers to find top talent.
Worries of AI usage in recruitment
“It’s almost a game – seeing so many CVs coming in and trying to work your way through them. AI was being used to make sure keywords were in an application and it’s almost about trying to work the system using technology,” David Cheng, Associate Professor at the Australian National University Research School of Management, told HRD.
“The danger here is trying to find loopholes. Like any new technology we don’t yet know where is too much or where is too little.”
It’s this worry, Cheng noted, that is creating an “arms race” between employers and job seekers.
While some organisations, particularly in tech, may see value in AI-assisted hiring Cheng cautions that this doesn’t translate to every industry or role.
“At the CEO level, you’re hiring from a very small pool and won’t need to rely on AI filtering,” he explained. “But when you have 1,000 applicants for 10 positions, the temptation is to let AI take the first pass — even though you risk cutting out suitable candidates.”
Therein lies the issue – overreliance solely on AI to screen candidates. Lack of input, whilst improving efficiency may decrease efficacy.
“We’re losing those moments where someone learns from trial and error, because AI smooths everything out. That’s part of developing real expertise.”
“In the past, I could see people skills just by meeting someone or looking at their work. AI removes those natural touchpoints,” he added.
Rethinking assessment
As AI becomes more embedded in hiring, traditional methods of evaluating candidates — such as resumes, cover letters, and standardised tests — are losing their reliability.
With AI able to generate polished documents and near-identical test responses, Cheng believes HR teams will need to rethink how they assess skills.
“In the past, I could judge communication skills from a cover letter. Now, that might be written by ChatGPT,” he explained.
This could drive a resurgence in in-person evaluations, assessment centres, and live problem-solving sessions. “If you need someone with strong people skills, you may need to bring them in and see it firsthand.”
Cheng predicts a dual focus: technical competence in using AI effectively, and human strengths like critical thinking, creativity, and relationship building.
“Prompting AI may become like managing junior staff — and the best AI operators will also develop the higher-level skills we need in leadership roles.”
Eroding skills and shifting workforce pathways
Beyond recruitment, one of the longer-term implications of AI in the workplace is the impact on capability and entry-level roles that have traditionally provided the first step on the career ladder where junior employees could learn - often through repetitive, lower-skilled work.
Cheng pointed to parallels in law, where AI now outperforms graduates in legal research.
“If those graduate roles disappear, what happens in five or 10 years when you need experienced lawyers?” he asked.
This is a similar challenge across other white-collar jobs, where junior jobs, generally, develop talent.
“Back in the day, you started at the bottom of the food chain – that experience of learning day-to-day operations is invaluable. If junior positions are lost because of AI to improve efficiency, what does the workforce look like in 20 years? How are we getting skills into the workforce?” Cheng questioned.
For HR leaders, this creates a strategic tension: how to balance efficiency gains from automation with the need to cultivate future talent.
“Large organisations may be able to take the risk of training juniors even if AI can do their job better,” Cheng noted. “Smaller businesses, under financial pressure, may not.”
