AI in HR and Recruitment in Oman: What to Automate and What to Keep Human
AI can take much of the paperwork out of HR:
- job descriptions, and job adverts in Arabic and English;
- CV summaries against agreed criteria;
- interview question banks;
- scheduling and candidate updates;
- onboarding material;
- answers to policy questions from the staff handbook.
What it should not do is decide who is hired, promoted or dismissed. Those decisions carry legal weight in Oman. Under the Labour Law, dismissal for reasons such as sex, origin, religion, disability or pregnancy is arbitrary. And since September 2026, people can object to decisions made by automated processing of their personal data, and the employer must then bring in human review.
The safe pattern is simple:
- Criteria are written down before anyone screens.
- Protected characteristics stay out of the process.
- AI summarises and drafts.
- People assess, decide and record their reasons.
- Outcomes are checked for patterns.
Payroll, contracts and the Wage Protection System are a separate subject, covered in HR and payroll systems in Oman.
The short answer
- Good uses: job descriptions and adverts; CV summaries against written criteria, with the evidence quoted; interview guides and scorecards; candidate communication; onboarding; policy questions.
- Keep human: shortlisting decisions, interviews, offers, performance ratings, discipline and dismissal.
- Oman law: Labour Law Articles 11, 12 and 23; Personal Data Protection Law Article 14 on automated decisions, and Article 5 on sensitive data.
- Controls: written criteria, protected characteristics removed, a person decides and records reasons, outcomes monitored, candidates informed, data deleted when the purpose ends.
Where AI helps in HR
| Task | How AI helps | What stays human |
|---|---|---|
| Job descriptions | Drafts from a role brief, in Arabic and English, in plain language | The manager confirms the duties and requirements |
| Job adverts | Adapts the description for different channels | HR removes requirements that exclude people without good reason |
| CV summaries | Summarises each CV against written criteria, quoting the evidence | People decide the shortlist |
| Interview guides | Drafts structured questions and scoring guides tied to the criteria | Interviewers ask, listen and score |
| Candidate communication | Drafts scheduling messages, updates and respectful rejections | HR sends them and answers questions personally |
| Onboarding | Turns policies into checklists and first-week guides | HR approves the content |
| Policy questions | Answers staff questions from the handbook, citing the section | HR owns the handbook and handles anything sensitive |
| Performance reviews | Turns a manager’s notes into a clearer draft | The manager’s judgement and rating |
Answering policy questions from the handbook is a good job for an internal AI assistant for company documents.
Why screening is the risky part
- Past decisions carry past bias. A system that learns from who was hired before can repeat whatever skewed those decisions.
- Proxies. A name can suggest nationality or gender. A career gap can reflect maternity or illness. An address can suggest background. Removing the obvious field is not enough if a proxy remains.
- Scores hide reasons. A ranked list looks objective, but nobody can say why one candidate is 82 and another 79.
- People can object. Since 7 September 2026, Article 14 of the amended Personal Data Protection Law gives data subjects the right to object to decisions resulting from automated processing. The controller must then bring in the human element when reviewing the decision.
- Europe treats it as high-risk. The EU AI Act lists AI systems intended for the recruitment or selection of people, and for decisions on promotion, dismissal and performance monitoring, among its high-risk uses. Those obligations now apply from 2 December 2027. They matter if you recruit in the EU or sell HR software there. The EU AI Act for GCC businesses covers the timetable.
What Omani law says
The Labour Law, issued by Royal Decree 53/2023 on 25 July 2023, has three provisions that bear directly on AI in HR:
- Article 12 makes a dismissal arbitrary if it is for sex, origin, colour, language, religion, belief, social status or disability, or, for women workers, pregnancy, childbirth or breastfeeding.
- Article 11 applies where a court finds a dismissal arbitrary or unlawful. The court orders reinstatement, or compensation of at least three and at most twelve months’ wage, on top of end-of-service benefits and other entitlements.
- Article 23 requires employers to employ Omanis, with Omanisation percentages set by ministerial decision. It also requires employers to treat all workers equally where the nature and conditions of work are the same.
If AI plays any part in performance, redundancy or disciplinary decisions, none of the Article 12 grounds may enter, directly or through a proxy. The reasons for each decision should be written by the people who made it.
The Personal Data Protection Law applies to every CV and employee file:
- Sensitive data — health, religious beliefs and racial origin among them — may not be processed without a permit from the Ministry (Article 5). Since September 2026, Article 5bis exempts employee data processed within the employer’s internal work, provided it is not disclosed to third parties without the employee’s written consent. Candidates are not yet employees, so don’t assume the exemption covers them. And check whether sending employee data to an outside AI provider fits the exemption.
- Automated decisions can be objected to, with human review (Article 14).
- Erasure: data must be erased when the purpose of processing ends, so unsuccessful candidates’ CVs need a defined retention period.
- Transfers abroad: a screening tool hosted outside Oman makes every CV a cross-border transfer.
What Royal Decree 68/2026 changed covers the amendments in detail.
A safe AI-assisted hiring process
- Write the criteria before advertising: must-haves, nice-to-haves, and how each one is evidenced.
- Remove protected characteristics and obvious proxies before AI sees the CV: photo, date of birth, marital status, religion, gender, family details. Nationality matters where a role is reserved for Omanis or Omanisation targets apply. Handle it as a rule-based check, never as an AI judgement.
- Ask for evidence, not scores. For each criterion, the summary quotes the CV or says “not shown”.
- A person shortlists and records the reasons.
- Interview with a structured guide, the same core questions for every candidate.
- Tell candidates how AI is used in the process, and how to ask for a human review.
- Monitor outcomes. Compare pass rates across groups where you lawfully hold the data, and investigate unexplained gaps.
- Delete on time. Erase unsuccessful candidates’ data when the purpose ends, unless they agree to be kept on file for future roles.
A prompt for CV summaries
You are helping a hiring manager review applications for a Senior Accountant role. Assess the CV below against each criterion in the list. For each criterion, quote the evidence from the CV, or write “not shown”. Do not give an overall score or a recommendation. Ignore age, gender, nationality, religion, marital status, photos and family details, and do not mention them.
Criteria: (paste the agreed criteria) CV: (paste the CV, with contact details removed)
The two refusals in that prompt do the work. Without a score, a person has to read the evidence and decide. Without protected details in the output, they are less likely to creep into the decision. Removing them from the input as well is better still. Checking AI output covers how to review the summaries.
Beyond hiring
- Performance reviews: AI can make a manager’s notes clearer and more specific. It should not generate ratings, or infer performance from activity data.
- Employee questions: an assistant that answers from the handbook saves HR time, as long as anything about health, pay disputes or discipline goes to a person.
- Staff use of AI tools: HR usually owns the policy. A policy for staff using AI tools gives a template.
Frequently asked questions
Can we use AI to screen CVs in Oman? As a summarising aid, against written criteria, with a person making the shortlisting decision. Keep protected characteristics out of the process, tell candidates how AI is used, and remember they can object to decisions made by automated processing.
Can we reject candidates automatically? It is risky. Since September 2026, people can object to decisions resulting from automated processing of their data, and the employer must bring in human review. A process where a person decides, and records why, is easier to defend.
What personal data in a CV is sensitive? Data about health, religious beliefs and racial origin, among other categories, is sensitive under Article 5 of the Personal Data Protection Law and needs a Ministry permit to process. Remove it before any AI processing and keep it out of decisions.
How long can we keep CVs of unsuccessful candidates? Only until the purpose of processing ends. Set a defined retention period, and ask candidates whether they want to be kept on file for future roles.
Can AI write job descriptions in Arabic? Yes, with review. Specify the register and terminology, and have HR check the requirements. Prompting in Arabic covers how.
Does the EU AI Act affect employers in Oman? Only if you recruit people in the EU or provide HR AI systems there. In that case, AI used for recruitment and selection is high-risk, with obligations applying from 2 December 2027.
This article states the position under the laws in force on 28 October 2026, the date of publication. It is not legal advice.
Sources: Royal Decree 53/2023 — Labour Law · Royal Decree 68/2026 amending the Personal Data Protection Law · Royal Decree 6/2022 — Personal Data Protection Law (Arabic text) · EU AI Act — Annex III, high-risk AI systems
Want AI that saves HR time without putting hiring decisions at risk? ZAWAT builds AI and automation and business systems with structured criteria, human decisions and audit trails designed in. Book a call.