Candidates accept AI as an aid. They do not accept it as the judge. The expectation is that a person stays accountable for the outcome, and that automated steps inform that person rather than replace them. Saying so clearly matters as much as doing it. A candidate who knows a human reviewed their result feels respected. One who suspects an algorithm had the final word does not.

People want to be measured on things that relate to the job. Skills, aptitude and role-relevant traits feel fair because a candidate can understand them and a recruiter can explain them. Tools that infer personality from a face, or read emotion from a voice, feel like the opposite. They are opaque, they are hard to justify, and candidates increasingly see them as intrusive. Validated assessment that measures defined, job-relevant qualities is what meets the expectation.

An assessment asks candidates to hand over personal information, often before they have any relationship with the employer. They expect that information to be used only for the purpose at hand, kept secure, and not sent somewhere they would not expect. Where data is stored has become part of this expectation, particularly for roles that touch government or sensitive work. Being able to say plainly that data stays onshore answers a question candidates and clients are now asking.

Candidates know that bias exists in hiring, and many worry that AI could make it worse rather than better. The expectation is that your process is consistent and based on evidence, so every applicant is held to the same standard. A structured, validated assessment does this more reliably than unaided human judgement, which is one of the strongest reasons to use good tools well. The point candidates care about is that the same yardstick is applied to everyone.

Candidates do want a faster process. What they do not want is to feel like an input. AI-assisted hiring can remove delay and free your team to spend time where it counts, but the experience still has to feel human. Clear communication, realistic timeframes and a way to ask a question all signal that there are people behind the process. Efficiency and respect are not in tension. The best processes deliver both.

A few practices cover most of it:

• Tell candidates where AI and automation are used, in plain language.
• Keep a person accountable for every decision, and say so.
• Assess people on relevant, validated qualities, not opaque inference.
• Use no tools that read faces or judge tone of voice.
• Handle data carefully, and be clear about where it is stored.
• Communicate along the way, and give candidates a route to ask.

What candidates expect is fast becoming what the law requires. From 10 December 2026, Australian privacy rules will require organisations to disclose automated decisions that affect a person’s rights or interests, and the Government’s AI policy and the APSC principles set similar expectations for the public sector. The direction is the same as the candidate’s. Be transparent, keep people in control, and be able to explain how a decision was made. Meeting candidate expectations and meeting the new rules are, in practice, the same piece of work.

Does AI decide who gets hired?

It should not. In a well-designed process, AI and automation support the people making the decision rather than making it. A person stays accountable for the outcome.

Should candidates be told when AI is used?

Yes. Telling candidates in plain language where AI and automation are used is both what they expect and, from December 2026, what Australian privacy rules will require.

Is AI-assisted hiring fair?

It can be fairer than unaided judgement when it is built on validated, role-relevant methods and kept under human oversight. It is unfair when it relies on opaque inference that no one can explain.

How is my data used in an online assessment?

A responsible provider uses your information only for the assessment, keeps it secure, and is clear about where it is stored. It is reasonable to ask an employer or provider these questions before you start.