Most employers we speak with already use AI somewhere in their hiring process, even if they don't describe it that way. An applicant tracking system ranks applications, one tool drafts rejection emails, another schedules interviews. It rarely arrives as a single decision; it builds up one tool at a time.
Used well, AI takes administration off recruiters' desks and gives them more time for the conversations that matter. Used carelessly, it can screen out people you would have wanted to meet. This is where we see the line.
What AI does well
The tasks worth automating look similar in most companies: repetitive, rules-based and easy to check afterwards. Sorting and routing applications, scheduling interviews and confirmations, sending status updates so candidates are not left waiting, and producing recruitment reports without rebuilding a spreadsheet every month.
AI-assisted tools can also summarise large batches of CVs or highlight profiles that match set criteria. That is useful, provided everyone treats the output as a shortlist for review rather than a decision already made. A score should start a conversation, not end it.
Where it gets candidates wrong
A CV is an incomplete picture of a person, and keyword filters make it more so. Career changers, people returning after a break, candidates whose previous employer used different job titles, and those who did similar work in another industry can all look weaker on paper than they are in person.
The most common case we see is a candidate whose experience is genuine but described in different terms. Someone who managed vendor governance at a bank may never use the exact phrase in your job description, and the filter removes them before anyone reads the CV. The candidate is lost on wording, not ability.
Motivation, judgement, how someone handles mistakes and whether they will cope in a demanding team do not appear on a CV. They come out in conversation, which is why final shortlisting and hiring decisions need someone who has actually spoken to the candidate.
Candidate data deserves more attention
Everything a candidate sends you is personal data, from the moment it arrives until it is deleted. Hong Kong's Office of the Privacy Commissioner for Personal Data has published a model framework for AI covering governance, risk assessment, implementation controls and communication with the people affected. It is worth reading before you expand a tool, not after a problem arises.
Before introducing or expanding an AI recruitment tool, document:
- What the tool is for, and which decisions it supports
- What candidate data it collects, where it is stored and who can access it
- How results are checked for accuracy, relevance and bias
- Which decisions require human review, and who is accountable
- How candidates can ask questions or correct inaccurate information
- How long data is kept and when it is deleted
Vendors deserve the same scrutiny. Contract terms, security controls, data location and incident handling matter more than anything shown in a demo.
For candidates
Assume your application may be processed by a system before a person reads it. That is not a reason to fill your CV with keywords; it reads poorly and does not help with the software either.
Keep the layout clean and the headings standard. Use job titles and terms that accurately describe your work, and where the vacancy uses different wording for the same thing, use theirs if it is accurate. Quantify results where you can. Check the details before submitting, because an incorrect date or a missing role can raise questions you will not be there to answer.
If you think a system has misread your background, a recruiter can explain the context to the employer directly. This is one of the more practical ways we help.
For employers
Test your screening rules from time to time. Take a sample of applications the system rejected and have someone review them properly. If strong candidates keep appearing in that group, the criteria need adjusting.
Be careful about what you measure. Speed is easy to report and easy to mistake for improvement. Candidate drop-off, time between stages, offer acceptance, hiring-manager satisfaction and six-month retention will tell you more about whether the process is working.
Our view
The recruitment teams getting real value from AI use it to free up time for the work that needs judgement. Those that struggle are trying to automate the judgement itself. The two approaches can look similar on a process chart but lead to very different hires.
We help employers set selection criteria that reflect the actual role, review profiles a filter may have missed, and keep a person involved where it matters. For candidates, we explain the role clearly, help present experience in a way that works with screening tools, and give employers the context a CV cannot.
If you are reviewing your hiring process or would like a confidential conversation about your next move, contact the iTalent team. You can also browse our current openings.
