Automated screening promises speed, fairness and protection from fraud. Whether it delivers on any of them depends on one design choice: is verification done to the candidate, or with them?
Consider a job seeker in Manila who applies for a role she is well qualified for. She uploads a resume, sits through a one-way video interview recorded for an algorithm to score, and waits. A week later a templated email arrives: the company is moving forward with other candidates. She never spoke to a person. She does not know whether a human or a model decided, and she has no way to ask.
She is not an outlier. In a 2026 survey, half of job seekers said they had been rejected at least once in the past year without a single word from a human, and most of that group believed a machine made the call. Only about one in ten had ever been clearly told that AI was involved at all.
From paper trails to predictions
For most of corporate history, trust in hiring rested on two artefacts: the resume and the reference call. Both looked backward at things a candidate had already done. That model is being replaced quickly. By the end of 2025, an estimated 83 per cent of companies were using AI to review resumes, nearly double the year before, and a majority now use it to filter applicants before any human sees them.
Background verification is shifting the same way. The old process was an HR officer playing phone tag with former employers for days. The new one is software that cross-references government records, credential authorities and employment data in hours. At Smile Checks, where I work, automated cross-referencing compresses a check that traditionally took two weeks into roughly two days. The pattern is industry-wide: hiring is moving from static records toward systems that score, rank and predict.
Why the employer's worry is real
It would be easy to read this as employers chasing efficiency at candidates' expense. The fuller picture is that the threats driving it are also real. Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake: not embellished, but entirely fabricated. The FBI has documented more than 300 US companies that unknowingly hired North Korean operatives using stolen identities and AI-generated personas.
For remote-first employers, and for the Philippines' large outsourcing sector in particular, a fabricated hire is not a hypothetical. Verification, done well, protects honest candidates as much as it protects the company. The problem is not that screening exists. It is how it is built.
The cost of a frictionless rejection
Automated screening tends to optimise for the employer's experience and quietly tax the candidate's. Three costs stand out.
The first is the black box. When an algorithm filters someone out, it rarely tells them why, and almost never offers a route to appeal. A rejection with no reason denies the candidate the one thing that would let them improve, and it conveniently hides the error on the days the model is simply wrong. The second is bias laundering. A decade ago, Amazon scrapped an AI recruiting tool after finding it penalised resumes that referred to "women's" activities: trained on a male-dominated hiring history, it had simply learned to repeat it. "Objective data" can encode the very bias it claims to remove. The third is consent. Deep digital-footprint screening can analyse far more of a person's life than they knowingly agreed to share.
Candidates feel all of it. Nearly half believe AI hiring tools are built with biases working against their age, race, gender or background, and almost a third have abandoned an application rather than face a one-way AI interview.
Verification with the candidate, not to them
The encouraging news is that these failure points are design choices, and design choices can be changed.
Regulation is setting a floor. New York City's Local Law 144 now requires annual independent bias audits of automated hiring tools, advance notice to candidates, and public disclosure of the results. The European Union's AI Act classifies recruitment systems as "high-risk," with compliance obligations taking effect in August 2026.
Beyond compliance, the strongest signal is consent-based design: verification a candidate can see and authorise, rather than surveillance they never opted into. At Smile Checks, that means the candidate grants explicit permission, connects their own employment and government records, and has the data deleted once the report is delivered, in line with the Philippine Data Privacy Act of 2012. The principle generalises well beyond any single tool: tell candidates when AI is used, give a real human the authority to review a negative decision, audit models for bias before they ship, and collect only what the person knowingly agreed to share.
The choice in front of HR leaders
AI in hiring is not going to slow down, and for fraud-exposed employers it should not. But speed and dignity are not opposites. The organisations that earn trust will be the ones that treat candidates as participants in verification, not as data points to be processed. That is not a soft preference. In a market where rejected applicants compare notes publicly, a process experienced as a black box becomes a hiring disadvantage in its own right.
For HR leaders, founders and compliance officers, the test is concrete. Can a rejected applicant find out why? Can a qualified person reach a human? Does the system collect only what was knowingly given? A hiring process that cannot answer those three questions is not efficient. It is only fast. The work now is to build screening that uses AI to widen the door, rather than to quietly close it.