AI in recruitment. What people still decide.

What AI is genuinely good at

Recruitment involves a lot of reading. A role that attracts a few hundred applications means a few hundred sets of answers to check against the same requirements. That is exactly the kind of repetitive, criteria-based work where AI can help:

• Checking every application against the same criteria, in the same way, without tiring at application two hundred.
• Summarising long answers so a reviewer can see the key points quickly.
• Pointing to the evidence in an answer that relates to each criterion.
• Flagging missing information, such as an unanswered question or a broken link.
• Sorting applications so a person can start with the strongest matches.

Notice that every item on that list supports a person’s decision. None of them replaces it.

What it shouldn’t decide

AI should not be the thing that decides whether someone makes a shortlist or gets rejected. Hiring decisions affect people’s livelihoods, and the reasons behind them need to be understood and owned by a person who can explain them.

There are also things AI is poorly placed to judge at all:

• Whether someone will succeed in a role. Nobody can reliably predict that, and a tool that claims to is overselling.
• Personality from a face, a voice or a background. Appearance has nothing to do with the job.
• Potential that doesn’t follow the usual path, such as a strong applicant whose experience comes from a different industry.
• Anything outside the criteria the role actually needs.

The real risk is the criteria

Most problems with AI in hiring aren’t really about the technology. They come from what it is asked to look for. A tool set up to find people who resemble past hires will repeat whatever shaped those hires, good and bad. A tool that filters on keywords rewards people who write keyword-heavy resumes, not people who can do the job.

That is why the criteria have to come first, and they have to come from the job. What does this person need to be able to do in their first few months? What would you see in an application that shows it? Write that down before any tool is switched on, and make sure the tool assesses those things and nothing else.

How to use it fairly

If you use AI-assisted screening, a few principles keep it honest:

1. Set the criteria from the role: availability, work rights, relevant skills and judgement. Never proxies such as age, gender or where someone went to school.
2. Ask for evidence, not just scores. A reviewer should be able to see why an application was assessed the way it was.
3. Keep a person in every decision. A person reviews the shortlist, and no applicant is turned away by AI alone.
4. Tell applicants how it works. People deserve to know their application may be screened with AI assistance, and that a person makes the decisions.
5. Look after the data. Keep what you need for the role, protect it and delete it on a set schedule.

For your privacy obligations, the Office of the Australian Information Commissioner publishes guidance for organisations. For discrimination in employment, the Australian Human Rights Commission is the place to start.

Keep checking the results

Switching a tool on isn’t the end of the job. Every so often, read a sample of the applications it rated lower and see whether a person agrees. If good applicants are being missed, the criteria or the setup need work. A tool that nobody checks can drift away from what the role needs without anyone noticing.

Done well, this makes screening fairer, not just faster. Every application is read against the same criteria, including the ones that arrive last, and reviewers spend their time where judgement matters: listening to how someone explains things, or weighing an unusual background properly.

How we use it

At Supercharged Talent, hard requirements the employer sets, such as the right to work in Australia or the hours a role needs, are checked by simple rules. AI-assisted screening then checks the rest of each application against the role’s real criteria, which we agree with the employer in the brief, and it gives reasons drawn from what the applicant actually wrote.

Then a person takes over. A person reads the answers, watches the 90-second video intros and makes every shortlist decision. The AI never turns anyone away on its own. The employer interviews the shortlist and makes the hire.

Questions to ask any tool or provider

If you’re weighing up a screening tool, or a recruiter who uses one, ask:

• What exactly is it assessing, and where did those criteria come from?
• Can it show the evidence behind each assessment?
• Who reviews the results, and can a person override them?
• What happens to applicants’ data, and for how long is it kept?
• How are applicants told that AI is involved?

If a provider can’t answer those clearly, the tool is making decisions nobody can explain. That is a risk to your applicants and to your business, and no amount of speed makes up for it.