AI Automated Job Application System — Research Every Role, Tailor Every Application and Track Every Submission Without Spending 3 Hours Per
Built by RadianMatrix ·
About this build
A Claude and n8n powered workflow that takes a job posting URL, researches the company in depth, analyses the role requirements against your CV, rewrites your CV and cover letter specifically for that role, scores your application fit before you submit and logs everything to a tracking dashboard — collapsing a 3 hour m
What was built
Built this during a period of active job searching when I realised I was spending more time preparing applications than actually applying. The research phase alone was taking 45 to 60 minutes per role. The CV tailoring was another 45 minutes. The cover letter was another hour if I was doing it properly. By the time I had prepared one strong application the day was gone. The workflow runs from a single input — paste a job posting URL and your base CV into a form and everything else happens automatically. Step 1 — Job posting extraction n8n fetches the job posting URL and extracts the full text. Claude then parses the posting into a structured format — role title, company, key responsibilities, required skills, preferred skills, seniority level, and any language or values signals in the posting that suggest what this company actually cares about beyond the listed requirements. The language signal analysis is the part most applicants skip — companies consistently use specific words in job postings that reveal their culture and priorities, and mirroring that language in an application significantly increases match rate with applicant tracking systems. Step 2 — Company research via Perplexity The company name gets passed to Perplexity which runs 4 targeted searches. Recent news and announcements from the last 90 days. Their current product or service offering and any recent changes. Their stated values and culture from their website and reviews on Glassdoor or similar. Any funding, growth or strategic developments that suggest why they are hiring for this role right now. The why they are hiring now question is the most valuable output of the research step. A company that just raised a Series B and is hiring a Head of Marketing is in a completely different situation to a company that has been flat for 2 years and is replacing a departing employee. The application should acknowledge which situation applies — not explicitly but in the framing of what you offer. Step 3 — CV tailoring via Claude The base CV and the structured job posting analysis get passed to Claude with a prompt that does three things. Scores each experience on the CV for relevance to this specific role on a scale of 1 to 10 and flags any experience that should be expanded, reduced or removed for this application. Rewrites the professional summary at the top of the CV to be specific to this role — referencing the company's current situation, the role's primary challenge and how the applicant's background positions them to address it. Reorders the skills and achievements within each role to lead with the most relevant items for this specific posting rather than the most impressive items in absolute terms. The distinction matters — what is most impressive in your career may not be most relevant to this hiring manager today. Step 4 — Cover letter generation via Claude The cover letter is written using a strict 4 paragraph structure that consistently outperforms generic cover letters in hiring manager feedback. Paragraph 1 — opens with something specific about the company from the research step that signals genuine interest rather than generic enthusiasm. Not I have always admired your company — a specific recent development that connects to why this role is interesting right now. Paragraph 2 — the most relevant single achievement from the CV, told as a brief story with a specific result. Not responsible for X — did X which resulted in Y measurable outcome. Paragraph 3 — how the applicant's specific background addresses the primary challenge implied by the job posting. This is where the language signal analysis from Step 1 pays off — using the company's own language to describe how you would approach their problem. Paragraph 4 — brief close with a specific call to action. Not I look forward to hearing from you — a specific proposed next step. Total length — 250 to 300 words. Hiring managers report that cover letters under 300 words get read significantly more often than longer ones. Step 5 — Fit score and application decision Before submitting anything Claude generates a fit score for this application out of 100 across 5 dimensions — skills match, experience level match, industry match, culture signal match and application strength. Any dimension scoring below 6 gets flagged with a specific explanation of the gap and a suggestion for whether to apply anyway, address the gap directly in the cover letter or skip this role. The fit score has prevented several applications to roles that looked right on the surface but had a fundamental mismatch that would have been obvious to the hiring manager — saving the time of preparing and submitting an application that was unlikely to progress. Step 6 — Airtable tracking Every application gets logged to Airtable automatically with the company name, role, application date, fit score, which version of the CV was used, whether a cover letter was submitted and the current status. Follow-up reminders fire automatically 7 days after submission if no response has been received — a single Slack message prompting a brief polite follow-up email. Results after 8 weeks of active job searching: 47 applications submitted. Average time per application — 7 minutes and 43 seconds from URL input to submission ready. Previous average without the system — 2 hours 51 minutes. Interview rate — 34 percent of applications, compared to an estimated 8 to 12 percent industry average for similar roles. Two hiring managers specifically mentioned the cover letter as standing out during the interview. The fit score proved its value most clearly when it flagged a role I was excited about as a 4 out of 10 on experience level match — the role wanted 8 plus years of experience and I had 4. Applied anyway. Did not get an interview. Would do the same again because the data was there to make the decision consciously rather than hoping for the best. What broke during development: The CV reordering was too aggressive in early versions — Claude was removing experience it deemed irrelevant rather than just reordering. Fixed by changing the instruction from optimise to reorder and expand or contract — the base experiences must stay, only their prominence and detail level changes. The cover letter first paragraph kept defaulting to generic admiration despite specific instructions. Fixed by providing 3 examples of bad first paragraphs and 3 examples of good first paragraphs directly in the prompt — showing the contrast was more effective than describing it. Job postings with very little text — some companies post extremely brief descriptions — gave insufficient data for the language signal analysis. Fixed by adding a fallback that uses Perplexity to search for more information about the role and the team before running the analysis. What I would build next: An interview preparation module that takes the job posting and the submitted application and generates a set of likely interview questions with suggested answer frameworks based on the specific experiences highlighted in the tailored CV. The application is the first step — the interview is where the job is won or lost and the same research that powers the application can power the preparation.