A system, not a shortcut

Stop Winging Your Job Search

By Dan Lee

A framework for using AI as an operating partner during an active search, with persistent context, structured workflows, and compounding prep.

See the steps ↓

Job searching is brutal. The market is competitive, and companies are evolving fast. Running a serious search means tailoring resumes, tracking each company's context, and prepping for interviews at different stages, and that complexity compounds when you're crossing industries or applying to different types of roles.

For experienced candidates, there's the added challenge of having too much history — a decade-plus of work is hard to synthesize and even harder to tailor for each role.

This is a system I built and ran during my own job search. It uses AI as an operating partner, not a chat assistant you ping for one-off advice. It knows your background, tracks your pipeline, and does all your prep work so you can focus on the interviews themselves.

How this impacted my job search:
100%
screen rate
7 for 7 applications
38
interviews across 7 companies
~3x
productivity multiplier
~24 hrs prepping, saving ~53 hrs

Your AI partner knows who you are

The core problem with using AI for job searching is that it doesn't know you. You explain yourself every session, it gives generic advice, nothing compounds. This system fixes that by giving the AI a persistent knowledge base about you.

Handles

  • Resume tailoring — paste a job description, get a tailored version sourced from your master CV
  • Interview prep — research the company and interviewer, map your stories to what they'll probe
  • Debrief processing — turn a transcript into structured analysis: what you learned, what landed, what to fix
  • Pattern tracking — spot recurring issues across companies and address them systematically
  • Pipeline management — one place tracking every company, stage, and next step

Doesn't do

  • Get you in the door. — networking, referrals, cold outreach are still critical for the job search.
  • Replace knowing your material. — it helps you prep. It can't prep for you.
  • Affect how others behave. — recruiters can still ghost you, interviewers can still not like you, and headcount can still change.

Every company runs the same cycle

Each stage feeds the next. The improvement layer is what makes this compound — you get better across companies, not just within one.

feeds the next company 📋 APPLY 🔍 RESEARCH 📝 PREP 🎤 INTERVIEW 📊 DEBRIEF ⬆️ IMPROVE
Research 🔍 Builds the company hub — overview, role, people, and your "why this company" answer. Written once, reused every round.
Prep 📝 Pulls from the hub and your story bank. Maps your best stories to what this specific interviewer will likely ask.
Debrief 📊 Updates the hub with what you learned — intel on the team, process, red flags, anything to address before the next round.
Improve ⬆️ Patterns recurring across companies go into a shared learnings file. You stop relearning the same lesson at every screen.

Six files. The AI reads them so you don't repeat yourself.

Each session, the AI loads these files for full context. You never re-explain your background. You never paste your resume again.

context.md
Foundation
Who you are, what you're looking for, your background snapshot, communication preferences. The AI's north star. Everything else is filtered through this.
cv-master.md
Foundation
Your comprehensive career record — more detail than any resume you'd send. Full scope, metrics, context, bullets you'd cut for a specific role. Never edited for a company. Everything tailors from this.
stories.md
Foundation
Your core interview stories: the problem, what you did, the outcome, what it signals. 5–8 stories mapped to different competencies. The AI pulls the right ones for each interviewer.
[company]-hub.md
Per Company
One per company. Overview, the role, people you've met, your "why this company" answer, confirmed logistics, debrief summaries. Updated after every round.
[company]-round-N.md
Per Interview
One per interview. Interviewer research, story map for this conversation, questions to ask, tone notes. References the hub rather than repeating it.
learnings.md
Ongoing
Delivery and communication patterns observed across multiple interviews — recurring issues, status (open / in progress / resolved), and what "fixed" looks like. Checked before every round.
Folder structure
📁 job-search/
📄 context.md foundation
📄 cv-master.md foundation
📄 stories.md foundation
📄 learnings.md ongoing
📁 [company]/ one per company
📄 [company]-hub.md
📄 [company]-round-1.md
📄 [company]-round-2.md
📄 cv-[company].md
You don't need all of these on day one. Start with context.md, cv-master.md, and stories.md. Everything else you build as companies move forward.

Do the foundation once. Everything else builds on it.

Start with steps 1–2. Come back as needed.

🏗️
Foundation
~2 hours · Do this before anything else
1
Write your career snapshot. A 1-paragraph summary of who you are and what you're looking for. The AI uses this as its lens for everything — get it right.
2
Build your master CV and story bank. The AI interviews you to build a comprehensive career record — richer than anything you'd send to a company. It also walks you through 3–5 core stories one at a time. Start with 3; add as you go. These two files stay consistent with each other: no contradictory metrics, no duplicate claims.
🏢
Per company
~1 hour per company · When you apply, or they reach out
3
Tailor your resume for the role. Paste the job description. The AI tailors from your master CV — not your generic resume. You review and approve.
4
Set up the company hub and research doc. Once they respond and an interview is scheduled, build the research doc and hub file. These feed every round from here.
🗣️
Per interview
~1 hour · Before and after every round
5
Prep each round. Run the round prep prompt with the interviewer's name and role. It pulls from the hub and your story bank.
6
Debrief within 24 hours. Paste your notes (or transcript). Update the hub. If a pattern is recurring across companies, add it to learnings.md.
🔄
Ongoing — after each interview
10–30 min per touchpoint · The compounding layer
7
Weekly pipeline check. 10 minutes — review every active company, surface next steps, flag anything slipping.
8
Periodic story bank refresh. When your bank feels thin or a new panel type surfaces, do a story enrichment session. Not per-interview — maybe every few weeks.

Start simple. Upgrade when it makes sense.

Claude Code or Cowork adds the most power, and they're easy to use.

Claude.ai / ChatGPT
Ok to start
Same workflows, more friction. You paste your context at the start of each session instead of the AI reading files automatically. Works fine, just slower as the number of active companies grows.
Note-taking tool
Add early
Take the best notes you can. Ideally you can transcribe the calls, but hand-taken notes are better than nothing.

What a focused, AI-assisted search actually looks like

I ran this system across my full job search. Before using this, I applied to 4 companies and didn't hear back from any of them.

The pipeline

Most attrition was voluntary.

APPLIED SCREENED DEEP ROUNDS OFFERS ACCEPTED 7 Connection (5) Cold apply (1) Recruiter outreach (1) 7 5 3 1 Withdrew (had offers) Withdrew (comp gap) No offer Ghosted Declined offers (2)
The timeline

Eight weeks of overlapping processes. Each dot is a conversation, color-coded by type. Hover or tap a week to see hours.

Swipe to scroll →
W1 W2 W3 W4 W5 W6 W7 W8 W9 Jan 26 Feb 2 Feb 9 Feb 16 Feb 23 Mar 2 Mar 9 Mar 16 Mar 23 PEAK WEEK Co. A Co. B Co. C Co. D Co. E Co. F Co. G
Interviewing
AI prep
Time saved
Screen
1:1 Interview
Live Exercise
Presentation
Offer
Recruiter/Prep