Back to the Campfire Blog

Steal My AI Prompts For Customizing Your Application to Any Job Ad

Written by: Chai Fisher Tags: jobs-career

Published: Jul 16, 2025

One of the strategies that helped me secure my 1 in 4 application to interview rate during my job search was to essentially treat my application like a CRO project.

This meant EVERYTHING in my resume and cover letters had to help work toward a conversion, from passing through ATS to passing human reviewer.

Here's how I optimized my application to convert

1. EVERY SINGLE BULLET POINT ON THE RESUME HAS A STATISTIC. No one cares that you work well with others, but they do care if you "trained x team members". It doesn't matter if what you have to say isn't much. "Trained two people" will beat out "Trained people" any day. Everything needs a statistic. I cannot stress this enough. It is the best way to prove to recruiters that you're not full of it early on by explaining WHY you're good, not just claiming you ARE.
   
   
2. I used keywords on my resume that I saw job postings use. Cross-collaboration, keyword research, Surfer, Screaming Frog. I don't necessarily use words like cross-collaboration on a day to day basis, but the software does.
   
   
3. Formatting was plain as hell. I'd previously used a custom template for InDesign that was gorgeous. No interviews. Switched to the most boring and computer friendly template, and it worked great.
   
   
4. Send a cover letter for each job. Your application will look complete, get picked up more by ATS, so more likely to get an interview over someone that hasn't. 

Cover letter writer AI prompt 

Prompt 1 (Attach your resume and job description to this)

You are my ANALYZE & MATCH agent. I will give you:
{JOB_POSTING_TEXT} ← entire job ad
{COMPANY_NAME}
{USER_RESUME_TEXT} ← your full résumé / work history
{OPTIONAL_COMPANY_DOCS} ← paste About-page text, press releases, etc. (or leave blank) Your job (do ALL of this):

1. Extract from the job posting: – "role_title" – 3-5 "top_objectives" (plain-language success goals) – lists of "must_have_skills" and "nice_to_have_skills" – up to 8 "cultural_values_keywords" (tone, verbs, adjectives)
1. For every "must_have_skill" and each "top_objective," pull 1-2 concrete résumé achievements that prove I can deliver. Include metrics if possible.
2. Note any obvious gaps where my résumé lacks proof.
3. Derive a high-level writing style for the company (“tone_guidelines”) plus two recent company themes (from supplied docs or the cultural keywords).
4. Craft two punchy hook sentences that tie my background to their mission.

Return only this JSON: { "role_title": "", "top_objectives": [ "" ], "must_have_skills": [ "" ], "nice_to_have_skills": [ "" ], "cultural_values_keywords": [ "" ], "skill_alignment": [ { "requirement": "", "matching_experience": "", "metric_or_result": "" } ], "gaps": [ "" ], "tone_guidelines": "", "two_recent_company_themes": [ "", "" ], "hook_sentence_options": [ "", "" ] }

Prompt 2

You are my OUTLINE agent. Input:
{ANALYSIS_JSON} ← paste the full JSON output from Prompt 1.
Task: – Build a 4-part cover-letter framework:
hook – choose ONE of the provided hook sentences.
proof_points – weave 2-3 of the strongest aligned achievements around the top objective.
culture_future – sentence(s) linking my values to one recent company theme.
close – confident CTA (“I'd love to discuss…”, etc.).
– Keep each section as a single string (no newlines inside values).
Return only this JSON: { "outline": { "hook": "", "proof_points": [ "", "", "" ], "culture_future": "", "close": "" }, "tone_guidelines": "" // pass through from Prompt 1 so the next agent sees it }

Prompt 3

You are my DRAFT & POLISH agent.
Inputs:
{OUTLINE_JSON} ← from Prompt 2
{COMPANY_NAME} {SPECIFIC_HIRING_MANAGER_NAME} (optional; default “Hiring Team”)

Tasks:
Write a complete cover letter (~300–350 words) in first-person singular, active voice, following the outline and tone_guidelines.
Address the letter to the manager name provided (or “Hiring Team”).
Self-proofread: fix typos, tighten wording, and ensure it stays ≤350 words.
Return only the finished letter in Markdown—no notes, no JSON.