Five AI Tool Categories
Dimension: Pursuit · Type: Stage
A taxonomy that separates the AI tools you can use during a job application by what they are actually good at, so you reach for the right one at the right step.
Introduced by Mian Nabeel Ahmed (IFAD) at the Mapping Professional Achievements session of the UN Inter-Agency Career Week 2026, on 5 May 2026, as part of the speaker’s “playbook for success” segment.
The framework
Each category solves a different problem. Use them in the order that matches the application workflow.
1. JD parsing
What it does. Breaks down a job description, identifies the key skills and keywords, and surfaces what your CV is missing. Helps you reverse-engineer the language of the JD into your own document.
When to reach for it. Right after you decide to apply. Run it before you start tailoring.
2. ATS readability checking
What it does. Simulates how an Applicant Tracking System will parse your CV. Flags formatting issues (multi-column layouts, embedded boxes, image-based text) and missing keywords.
Examples. Built into JobScan and Teal; some standalone tools (Resume Worded, Enhancv) include this.
When to reach for it. After your tailoring is done, before submission. If the JD is for an organisation that does not use ATS, the readability discipline still helps a human recruiter scan you faster, but skip the keyword-density push.
3. AI drafting
What it does. Helps you draft, restructure, and refine language. Catches tense mixing, weak verbs, missing transitions. Useful for unblocking when you have raw content and need a tighter version.
Examples. ChatGPT, Claude.
When to reach for it. When you have a complete first draft and want it tighter. Not as a starting point. The blank-page use case usually produces a generic application.
4. Polishing
What it does. Grammar, tone, clarity. Tells you whether your language sounds overconfident or underconfident, flags wordiness, suggests simpler phrasing.
Examples. Grammarly.
When to reach for it. As the last automated pass, after layer one and layer two of the Third Eye Principle. Treat its suggestions as proposals, not commands. It does not understand context.
5. Research and translation
What it does. This category is mostly LinkedIn, used as a research tool rather than a job board. It helps you understand how roles are described in other sectors and translate UN experience into the language a private-sector or IGO recruiter will recognise.
Examples. LinkedIn, plus organisational websites and strategy documents for the target organisation.
When to reach for it. Before tailoring, especially when you are applying outside your current sector and need to learn the vocabulary.
Worked example
A UN programme officer applies for a senior advisor role at an INGO.
- Category 5, research, comes first. The applicant spends an hour on LinkedIn looking at profiles of current INGO senior advisors and at the INGO’s published strategy. They notice that “advocacy” and “policy influence” are used in place of “policy support”.
- Category 1, JD parsing, runs the JD through JobScan. The tool flags “donor engagement” and “coalition building” as core terms not present in the CV.
- Category 3, AI drafting, is then used to restructure two CV bullets to surface honest examples of donor engagement and coalition building from the applicant’s actual work, using the language from the research step.
- Category 2, ATS readability, confirms the new CV reads cleanly and the JD-match score has moved from 58% to 79%.
- Category 4, polishing, runs Grammarly across the cover letter, fixing two tense errors and softening one sentence that sounded boastful.
The whole pass is about two hours. Without the categorisation, the same applicant would have spent the same two hours moving between tools without knowing which one to trust for which job.
Pitfalls
- Using AI drafting as a starting point. Cold-prompting an AI to write a cover letter from scratch produces text that sounds the same as everyone else’s. Use it on your own draft, not in place of it.
- Trusting category 4 to fix structural issues. Grammarly will not tell you that your cover letter is misaligned with the JD. Layer one of the Third Eye Principle does that.
- Pushing for high JD-match scores by adding keywords you cannot back up with experience. A high score on a dishonest CV will not survive an interview.
- Skipping the privacy check. Never paste sensitive personal data (full date of birth, ID numbers, financial information, passport numbers) into a consumer AI tool. Strip identifying details from any document before pasting.
- Verifying nothing. AI tools make confident, plausible mistakes. Read every suggestion before accepting it.
When not to use it
When you are applying to an organisation that explicitly discourages AI-drafted applications and reads every submission by a human. In that context, categories 1, 2, and 5 are still useful as research and discipline; categories 3 and 4 should be used minimally and only for cleanup, not generation.
How I use it
Personal note pending. Davide to fill.
Related frameworks
- Third Eye Principle, the review framework these tools support.
- BASIC Achievement Bank, the upstream content the AI drafting category works on.
- R-CAR, the structure the AI drafting category should preserve when refining bullets.
- Four Prompting Principles, the prompting discipline that makes categories 1, 3, and 5 work.
Notes compiled by Davide Piga. Last updated 2026-05-09.