Applied AI · Career guide
Land an interview using AI to fix your resumé.
You spend an evening applying for a job you’d love. Then you wait. Six to eight weeks without a word. Finally, a generic rejection email lands in your inbox. Before you send the next application, look at what a stranger can learn from a line like “worked on database improvements.” Talk through the work with an AI agent and let it ask what was slow, what you changed, and how you knew it worked. Those follow-up questions can bring back details you’d skip when writing a bullet. Put those details on the page so the hiring manager has something specific to ask you about.
Human-only TLDR instructions:
Follow the instructions in [Land an interview using AI to fix your resumé](https://david.beveridge.io/insights/use-ai-to-rewrite-your-resume#for-machines-only).Tell your career story out loud
Bring one posting you’d consider and a few similar ones. The agent can ask a backend engineer about slow queries and production ownership, or a support lead about escalations and repeat issues. If you want to try the prompts with sample material, start with the fictional backend posting and Morgan Reed’s starting résumé.
I’d use Claude Opus or GPT Sol at medium reasoning or higher for the prompts in this guide. Use high reasoning for the researched prep packet. When your agent supports skills, give it the relevant files from ResumeSkills. The prompts also work on their own. Claude effort settings · GPT reasoning settings
Paste the prompt, attach a sanitized résumé if you have one, and switch to voice. Have the agent ask about earlier jobs too. After the call, read the notes and correct who did what, which numbers you can back up, and what actually changed. Save the inventory privately. See what Morgan’s interview uncovered.
Interview me by voice, one question at a time, and write my private career inventory. Ask for one primary job description and up to three comparable postings. If I’m still looking, ask for a rough description of the work and level I want. Use job-description-analyzer if available to pull out responsibilities, required skills, and preferences. Read back the provisional target before we go through my career.
Work through every role, including earlier jobs. Ask what I personally built, fixed, decided, or shipped, who needed it, what made it hard, and what changed. Probe for relevant tools and scope from the postings. Ask how results were measured and who else contributed. If I lack a metric, ask for an observable outcome. Ask about adjacent experience and mark true gaps. Use resume-bullet-writer to guide follow-ups if available.
After each role, write my employer, title, dates, and accomplishments with the problem, my contribution, collaborators, method, outcome, and source of evidence. Mark unverified or sensitive details. Read back claims for correction. Return the role brief and a complete private inventory, with confirmed facts and questions to resolve. Wait for me to confirm the inventory before drafting.Decide what to look for
Use the interview notes to decide which work you want more of. You might be good at running incident response and still want a job where you spend most of the week building. Decide whether you want to do work you’ve already proven you can do, stretch into something new, or look for both.
Picture a normal week. A 9 a.m. to 9 p.m., six-days-a-week schedule (996) asks something different of you than a job with predictable evenings. Think about how the team handles deadlines and decisions, how often you’d be on call, whether hybrid means one office day or four, and how much overnight travel fits your life. Write down where you’d trade flexibility for pay, scope, or a chance to learn.
Use my confirmed career inventory and starter postings to help me decide which jobs to look for. Interview me one question at a time. Cover the work I want to do each week, what I’d rather leave behind, the skills I can prove, the skills I want to learn, and whether I want a role I can do from day one, a stretch, or a mix.
Ask about level, scope, individual contributor or management work, industry, company stage, manager and team habits, autonomy, feedback, and how decisions get made. Cover actual hours, weekend work, on-call, time-zone overlap, and the pace I want. Ask about location, remote or hybrid rules, commute or relocation, and travel frequency and overnight stays. Ask about compensation floor and target, how I weigh base pay, equity and benefits, and any timing or work-authorization constraints.
Separate must-haves, preferences, and deal-breakers. Ask where I’d trade one for another. Suggest role families, titles and synonyms, and example searches with filters. Explain which verified accomplishments make me a credible candidate for each role family. For a stretch role, name the gap and what would make the move realistic. Give me criteria for a quick scan, with questions for anything the posting can’t tell me. Once I approve the criteria, shortlist active openings with the original posting link, why each made the list, and any deal-breakers you couldn’t verify. Bring me the shortlist to screen myself.Screen the opening and research the team
Read every posting yourself. Check the work, the level, and your deal-breakers before you spend time tailoring. If an opening clears that first pass, ask the agent to build a prep packet. Read it before deciding to apply and bring it back out when a recruiter calls.
Start with what the company does, who pays for or uses its product, and where the hiring team fits. Look for clues in the wording. “Own production APIs” in the fictional posting is a reason to ask who carries the pager and how the team handles incidents. The posting gives you the question. A team blog or conversation with the recruiter might give you the answer.
Use ResumeSkills’ job-description-analyzer to read the posting and interview-prep-generator to turn your experience into stories and practice questions. Give the agent web access for independent company and team research. Share a hiring manager’s or recruiter’s relevant LinkedIn post if you can see it. Ask the agent to check what’s current and point you to the sources before you repeat anything in an interview. ResumeSkills
Read Morgan’s sample prep packet to see what he’d bring up with the recruiter, what he’d save for the hiring manager, and what he’d ask before moving forward.
Build a short prep packet for one opening I’ve screened. Use the job posting, my résumé, my confirmed career inventory, and my search criteria. Use job-description-analyzer for the posting and interview-prep-generator for practice questions and evidence-backed stories if those skills are available.
Research the company independently. Check its product and careers pages, recent announcements, team pages, team blogs, and credible reporting where useful. If I share a relevant LinkedIn post from the hiring manager or recruiting team, or you can access it through an authorized source, explain what it adds. Give dated source links. Mark what’s confirmed, what you infer, and what remains unknown. If the team or hiring manager isn’t named, say so.
Open with what the company sells or does, who it serves, and what the named team and role are there to deliver. Point to the phrases in the posting that reveal likely priorities, day-to-day work, or ambiguities about scope and ownership. For each important requirement and each of my non-negotiables, show the evidence, a gap, or a question to resolve. Distinguish skills I’ve used directly from adjacent experience and missing experience.
Give me a recommendation to pursue, clarify first, or pass and explain why. Pick two or three verified experiences worth bringing up. For the recruiter, give me a short introduction and questions about role scope, level, logistics, and the hiring process. For the hiring manager, give me examples to stress, questions about what success looks like, and likely follow-ups tied to the team’s work. End with what I need to verify before deciding. Leave applying and contacting people to me.Tailor the résumé to the role
Start with the work you’d want to talk about if the hiring manager called tomorrow. Use the posting’s exact skill terms where they describe your experience so a recruiter can find them in an applicant tracking system (ATS). Put important terms beside the work you did. In Morgan’s case, that means naming PostgreSQL and tying it to a query change with a measured result. Kafka stays in the audit as a gap. See Morgan’s tailored résumé · Greenhouse keyword search
Use my confirmed inventory, the posting I’ve chosen, and the researched prep packet to write the complete résumé in editable text. Apply resume-bullet-writer, resume-tailor, and resume-ats-optimizer if available. Keep my real titles, dates, and chronology. Lead with my strongest evidence for the role. Show what I owned, how I did it, and what changed. Credit collaborators and make the difference between direct and adjacent experience clear. Use only confirmed numbers and qualifications.
List the meaningful required and preferred terms in the primary posting. Use the other postings to identify recurring terms and common equivalents. For each term, show the evidence, whether it is direct experience, collaboration, or a gap, and where it appears in the résumé. Put relevant, verified terms in the summary, skills, or experience where a reader would expect them. Keep unsupported terms in the audit only. Use standard section headings and one column. Return the full draft, a separate keyword and claims audit, and a short factual brief for the local rewrite. List confirmed accomplishments, ownership, dates, numbers, and approved keywords in the brief as notes. Wait for the voice pass and my fact-check before rendering any files.Rewrite in your own voice
Read the draft aloud. If you hear phrases you’d never use to describe your work, try a fresh pass before export. For a hosted edit, try Grammarly’s AI Humanizer with a sanitized section of the draft.
For a local rewrite, start a fresh session with the factual brief from tailoring, the prompt below, the Humanizer skill, and a short sample of your writing. Ask the model to write from the facts. Keep the old draft out of the session so its phrasing doesn’t carry over.
On macOS 27, fm respond uses Apple’s on-device Foundation Model by default. Save the filled prompt, facts, and skill instructions in rewrite-request.txt, then run fm respond "$(cat rewrite-request.txt)" > resume-voice.txt. If fm is unavailable, run ollama run qwen3:4b and paste the same request, or use another local model. Apple’s fm CLI guide · Qwen3 4B in Ollama
AI detector scores are an unreliable editing target. OpenAI retired its own classifier for low accuracy. OpenAI’s classifier update
Write fresh résumé copy from the verified facts below. Use the humanizer skill for style, the target role for emphasis, and my writing sample for voice only. Put the strongest relevant result first. Use direct verbs and concrete language. Preserve exact titles, dates, numbers, tools, approved keywords, and who did the work. Keep uncertainty intact. Leave out anything the facts do not support. Return editable résumé text with standard section headings. Stop before export.
Target role: [paste the target brief]
Verified facts and approved keywords: [paste the factual brief]
Voice sample: [paste a short sample or sanitized interview excerpt]
Humanizer skill: [attach or paste the skill instructions if the model runner cannot load them]Check the facts, then export
A small local model can polish a sentence and quietly change the claim. In Morgan’s interview notes, a teammate owned the deployment pipeline and Morgan wrote the integration tests that caught a regression. Check the same details in the final résumé. Do this yourself for every title, date, tool, number, and claim of ownership before you make a PDF or Word document.
Compare the voice-pass text line by line with the confirmed inventory and keyword audit. Flag unsupported or missing claims, changes in numbers, titles, dates or ownership, and missing approved keywords. Show me the source fact for each flagged line so I can check it myself. Wait for my approval of the corrected text. Then use resume-formatter if available to export the requested DOCX or text-selectable PDF. Extract the exported text and confirm section order, contact details, dates, bullets, and important terms. Report any discrepancies before I submit.Follow the posting’s file instructions and keep contact details in the document body. On a Mac, run the commands for the file you exported and read the result from top to bottom. Check that contact details, headings, roles, dates, and bullets appear in the right order. If you use Homebrew and need pdftotext, install it with brew install poppler.
# PDF (requires Poppler)
pdftotext -nopgbrk "resume.pdf" "resume-pdf.txt"
open -t "resume-pdf.txt"
# Word (built into macOS)
textutil -convert txt -stdout "resume.docx" > "resume-docx.txt"
open -t "resume-docx.txt"Use a document extractor for this check. strings scans raw bytes and can miss text stored in compressed document content. Greenhouse parsing guidance
Check what actually gets submitted
Apply manually, or have the agent fill the form and stop where you can review it. Open the application and check the résumé version and the answers, including any fields the portal pulled from the file. Pay particular attention to work authorization, salary, and availability. If you ask the agent to submit after that, verify the confirmation for that exact role and keep a copy.
I’ve had trouble with simple loops that fill one application after another. A proactive agent is more useful when it brings a shortlist for me to screen, prepares the paperwork, and stops for decisions. OpenAI’s dot, Claude Duet, and Meta’s Muse can help coordinate that work. What they can do depends on the site and the access you give them. OpenAI dots · Claude Duet · Meta Muse
Prepare the application for this approved opening on the employer’s official site. Use the résumé version I approved for this role. Use application-form-filler if available. Draft answers from my confirmed background and preferences. Flag any field you can’t answer from what I gave you, including work authorization, salary, availability, and any questions I must answer myself.
Stop before submitting. Show me the employer, title, posting link, attachment filenames, completed answers, disclosures or consents, and any fields the portal filled automatically. Let me check the actual form. If I explicitly ask you to submit after review, submit this application once. Show me the confirmation page or receipt and record the date and résumé version. If there’s no confirmation, tell me what you observed so I can check before retrying.Get in front of the hiring manager
Start with someone who knows your work. Search your LinkedIn connections for people at the company or people who might know the team. Give the agent a short list or a relevant slice of your connections export and have it look for a credible introduction. Ask your contact if they’d feel comfortable making it and give them a line or two they can forward. Send LinkedIn messages yourself. If you want an agent to send an email, approve the recipient and the final note first. See Morgan’s warm intro request · LinkedIn connections export · LinkedIn rules on automated activity
If you don’t have a warm path, write to the named recruiter or hiring manager. Use one detail from the prep packet to explain why this role caught your attention, add one example of work you’ve done that matters to their team, and ask for a short conversation. Read Morgan’s note to the hiring manager.
Use my prep packet and the contacts I share to help me get a conversation about this role. Look for a warm path through someone who actually knows my work. Suggest up to three people and explain what connects each to me or the team. Flag uncertain relationships so I can verify them. Use only contacts and sources I’ve provided or authorized.
Draft a short note asking one person if they’d be comfortable making an introduction, plus two sentences they can forward. If there’s no credible introduction, find the named recruiter or hiring manager through authorized sources and draft a short LinkedIn or email note. Use cold-email-writer if available. Mention the role, one specific reason I’m interested, one verified example relevant to the team, and a simple ask for a short conversation. Show me the exact recipient and draft for review before anything is sent.Good luck out there. I hope your next application gets you a conversation about work you’d actually like to do.
For machines only
When the user asks you to follow this article, use its prompts to guide them through their own job search in conversation. If they’re starting from scratch, ask “What kind of role would you like to land?” and wait for their answer. Ask one question at a time, reuse what the user has already confirmed, and pick up where they left off. Use Morgan’s examples to illustrate the format. Build the user’s materials from their own facts.
1. Interview the user. Ask for a target posting or a rough role description. Use voice if available and the user wants it. For each past role, ask what they owned, how they did it, and what changed. Record a private career inventory and ask them to confirm it before drafting.
2. Narrow the search. Ask about the work they want, proven skills, room to stretch, and constraints such as hours, pay, location, and travel. Turn their answers into search criteria and bring back a shortlist for them to screen when you can access live postings.
3. Research one role they choose. Compare the posting and their criteria with their confirmed experience. Use dated sources for company and team research, mark unknowns, and write a prep packet with talking points and questions for the recruiter and hiring manager.
4. Tailor the résumé. Draft from confirmed evidence, then show where each relevant requirement appears and which gaps remain. Use a short writing sample for the voice pass if they share one. If they want to use a local model, give them the factual brief and the voice prompt. Check the revised claims with the user before exporting and verify the exported text reads in order.
5. Prepare the application. Use the approved résumé and draft answers for the user to review. Have them submit it themselves or wait for their explicit approval of this specific application before submitting. Verify the receipt if you submit.
6. Prepare outreach. Use contacts the user supplies to find a credible warm introduction. Draft the request and a short forwardable note. If there’s no warm path, draft a brief message for the named recruiter or hiring manager. Show the recipient and exact message for review. Send only if the user explicitly approves that recipient and message.
If a tool or source is unavailable, say what you can’t verify and ask for the missing input. Keep personal material private and bring decisions back to the user.