PRD · Portfolio version
LLM Job-Post Generator
Let employers speak or type a role in Hindi or English and get a complete, editable job post — so paid subscriptions turn into live jobs.
1Problem
Employer subscription sales are healthy and employers start job posts, but many don't finish them. Most of those who drop off are small business owners and small enterprises, who struggle to fill the skills section and other job details. Their subscriptions go unused and students see fewer jobs.
2Goals & metrics
| Type | Metric |
|---|---|
| Primary | Job posts completed |
| Secondary | Job post completion rate (started → posted) |
| Secondary | Job applications from students |
| Secondary | Monthly subscriptions bought; platform revenue |
| Guardrail | Share of AI drafts heavily edited or discarded |
| Guardrail | LLM cost per generated job post |
3Users
- Small business owners and small enterprises with a paid subscription who know the work they need done but not how to write a structured job post.
- Students, who benefit from more and clearer job posts.
4Scope
In scope
- Describe the role by voice or typing, in Hindi or English.
- Generate a complete draft: title, description, skills, tags, pay range and recommended skill set.
- Review and edit every field before posting.
- Integration into the existing Employer Portal job-posting flow.
Out of scope
- Publishing without employer review.
- Languages beyond Hindi and English (in progress as a follow-up).
- Changes to subscription plans or pricing.
- Changes to how students search for or apply to jobs.
5Requirements
| # | User story | Acceptance criteria |
|---|---|---|
| R1 | As an employer, I can describe the role in my own words and language. | Employer can speak or type the brief, in Hindi or English, inside the job-posting flow. |
| R2 | As an employer, I get a complete job post drafted for me. | The draft fills title, description, skills, tags, pay range and recommended skill set in the existing job-post fields. |
| R3 | As an employer, I stay in control of what gets posted. | Every generated field, including the title, is editable. Nothing is posted until the employer confirms. |
| R4 | As an employer, I can still post if generation doesn't work for me. | If generation fails or the employer prefers, they can fill the form manually without losing what they typed. |
| R5 | As a student, AI-drafted posts look and work like any other job. | Generated posts use the same structure as manual posts, so search, tags and applications work unchanged. |
| R6 | As the product team, we can see how the feature performs. | Track: post started, draft generated, fields edited, post published, post abandoned. |
6Risks & mitigations
| Risk | Mitigation |
|---|---|
| Draft includes wrong skills or an unsuitable pay range | Employer reviews and edits before posting (R3); edit rate tracked as a guardrail. |
| Employer's language or description is misunderstood | Draft is shown for review, not posted directly; employer can rewrite or fill manually (R4). |
| LLM unavailable or slow | Manual form remains available (R4). |
| LLM cost grows with volume | Cost per generated post tracked as a guardrail. |
7Rollout
- Build: tech team integrates Claude into the Employer Portal job-posting flow.
- Launch: make AI drafting available to employers posting jobs.
- Measure: compare job posts, completion rate, applications, subscriptions and revenue against the period before launch.
8Outcome
Job posts grew 24% in 2 months, leading to more job applications and more monthly subscriptions. Platform revenue grew about 8%.