A hiring manager may decide whom to interview after reading every application on the shortlist. But the shortlist itself may have been shaped much earlier, when an AI tool screened the applications or assessed candidates’ skills.
That’s the part of CNA’s report that stayed with me. Two people can look at the same hiring process and both say a person made the decision. One may be thinking about who was invited to interview. The other may be asking who had a chance to get that far.
Where people come into the decision
A survey by the Singapore National Employers Federation found that in most of the uses of AI in HR it identified, a person confirmed the decision after AI provided input. When AI made a decision without prior approval, it was mostly for routine tasks or work handled in large volumes, such as screening résumés. Employers generally had someone designated to step in.
What I keep wondering is how that designated person would know when a decision needed another look. Would they have enough information to review it properly?
A routine task can still matter to the person affected
I can see why an employer receiving hundreds of applications would use AI to help with the first round of screening. From the employer’s side, the task is repetitive. From the applicant’s side, that first round can determine whether anyone sees their experience.
Imagine a candidate who has spent three years leading a shift but has never held the title of supervisor. If the candidate is screened out and later asks why, could HR check whether that experience was picked up? Could someone look at the original application and reconsider the outcome?
That example also shows why I think the timing of human review matters. A manager can make a careful, fair choice among the candidates on a shortlist and still have very little visibility into the applications that were filtered out before it reached them.
When a decision is challenged
The same issue can come up when a manager uses AI to summarise performance feedback or when an employer considers information produced by a tool during restructuring. If an employee raises a concern, the person handling it may need to go back to the original records and ask how much weight the recommendation carried.
For me, this is where the connection between AI and workplace fairness becomes clear. If a grievance comes in, can HR show who made the decision, what information they relied on and how they exercised their own judgement? Those questions are easier to answer when the reasons are recorded at the time, while the people involved can still explain what happened.
These are the questions we’ll explore with litigation practitioner Nadia Moynihan in our REL Alliance webinar, A Compliance Crossroad: Where AI Intersects Workplace Fairness, on 3 and 4 December 2026.
Comments