The job description generator is one of the most-used AI features on Jadiroon. It is also one where the difference between a good output and a poor one comes down almost entirely to the prompt — that is, what you put in the Description field on the role wizard.
What the generator sees
Four inputs go to the model:
- The title you typed.
- The function.
- Your free-form description — the most important input.
- The required competencies with their levels.
The model is also given guardrails on tone (formal Saudi business register), structure (four named sections), and length (around 250 words).
What it does well
- Producing a readable, conventionally structured JD from a brief seed.
- Writing Arabic and English together in a single generation. One click produces both versions as faithful counterparts — the same responsibilities and requirements, localized into each language — shown side by side for review.
- Tying responsibilities back to the required competencies you selected, so the JD reads as coherent rather than generic.
- Removing repetition and tightening phrasing.
What it does poorly
- Inventing requirements you did not ask for — the model occasionally adds a Bachelor's degree in Computer Science required line when no degree was implied. Read every JD before publishing.
- Specific compensation and benefits — the model will hedge or fabricate. Add these manually after generation, in your own words.
- Local Saudi context — references to Vision 2030, Saudi Labor Law, and Saudization phrasing are woven in when the sector or the Saudization preference warrants them, in both languages. Review them; the model still occasionally over- or under-applies them.
- Personality — the JD reads competent and bland. If your organisation has a strong tone, edit the output to match.
How to get a better output
Two practical moves:
- Be specific in the seed. We need someone to lead our hiring funnel from sourcing to offer letter, with three direct reports produces a much better JD than Recruitment manager.
- Pick the right competencies before generating. The competencies feed directly into the responsibilities. A role with five well-chosen competencies produces a more focused JD than one with twelve loosely-related ones.
Regenerating
Click Regenerate on the wizard and the model produces a fresh draft. Each generation is cached against the inputs, so regenerating with the same inputs returns the same output (and does not re-spend AI quota). If you change one of the inputs — even a single word in the description — a new generation is produced and the new output is cached.
AI quota
Every JD generation counts against your workspace's AI quota. The default in v0.5 is generous; most workspaces never come close to the cap. If you do, AI usage in your workspace explains how the quota works and how to monitor consumption.
Generating a single section in the job-description editor
Beyond the whole-JD draft above, the job-description editor can generate each MOMH-card section on its own, so you can build the card up section by section rather than accepting or rejecting one big draft. Four sections have a generator:
- Overview → Purpose — a Generate purpose with AI button fills the Arabic and English purpose.
- Responsibilities — generates 7–9 core responsibilities plus a set of general ones, each following the ministry's action verb + subject + purpose/impact formula.
- KPIs — generates a mix of quantitative and qualitative indicators, each tied to a strategic objective and a target.
- Qualifications — fills the five requirement fields (minimum education, specialization, field of experience, years, and preferred certifications).
Every section generator is grounded in two things: the ministry's job-description authoring rules (the MOMH methodology — the responsibility formula, the 7–9 rule, "describe the job not the person", the verb catalog, and the formal register), and the fields you have already filled in on the same card. Generation chains downstream: the purpose is written from the title, job type, sub-sector and job family; the responsibilities are written to serve the purpose; the KPIs are written from the purpose and the responsibilities you already have; the qualifications from the role and its purpose. The more of the card you fill before generating a later section, the better the fit — so a good order is title → purpose → responsibilities → KPIs → qualifications.
Preview before it lands
Purpose and qualifications fill their fields directly (you can then edit them). Responsibilities and KPIs open a preview first: the generated items appear with checkboxes, the main text is editable in place, and only the items you keep are added. Insertion is non-destructive — accepted items are appended to whatever you have already written, never replacing it. Nothing is saved to the server until you press the editor's Save, exactly like the rest of the card.
Each section generator counts against the same AI quota and is cached against its inputs (including the section-specific prompt), so re-running with the same card returns the same suggestions without re-spending quota.
Next
Mark the role succession-critical if it qualifies — covered next.