A hiring pipeline that reads every application. Gmail Trigger watches the inbox and fires on new application emails — the filter is yours to set, for example subject contains "application." Download Resume grabs the attachment, Parse Resume with AI has GPT read it and return clean structured JSON (name, email, skills, years of experience, a short summary), and the Job Description Set node holds the job posting you paste in once. Score Candidate compares the parsed resume against that JD and returns a 0–100 score plus strengths and risks; Normalize Result tidies everything into clean fields; Append to Sheet adds one ranked row per candidate to Google Sheets, highest scores on top when sorted. High Score? continues only at 80 or above, and Create Interview Draft creates a Gmail draft interview invite — a human reviews and sends every invite, because the workflow never emails anyone automatically. It exists for the hiring manager who reads the first twenty applications and ignores the other one hundred eighty.
The startup founder hiring the first engineers
A founder posts a backend role and gets 150 applications with no HR team and no ATS. The ranked sheet is the whole shortlist process: sort by score, read the top fifteen summaries, check the risks column for red flags like job-hopping or missing core skills. The strengths field tells her why each candidate scored well, so she is not trusting a bare number. Because the Job Description node holds the actual posting, the scoring is anchored to what she asked for, not a generic "good engineer" judgment. Practical tip: edit the system prompt in Score Candidate to weight skills, experience, or education differently — for a first engineering hire, she might weight demonstrated shipping experience above credentials.
The 5-person agency hiring an intern
An agency needs a design intern, gets 90 applications, and the partners cannot justify a week of resume reading. The default 80 threshold is too strict for an intern role, so they open High Score? and lower the cutoff — the workflow's value here is the ranked sheet and the parsed summaries, not the interview gate. The draft invite text in Create Interview Draft gets rewritten in the agency's casual voice, and it still stays a draft until a partner hits send. Practical tip: keep the threshold modest and let the risks column do the filtering — for junior roles, a clear, honest resume often matters more than a high score.
The recruiter running five roles at once
A solo recruiter is hiring a frontend dev, a PM, a designer, and two sales reps, and candidates keep landing in the wrong pile. The fix is one copy of the workflow per role: each copy gets its own Job Description node, its own Gmail filter (subject contains "frontend application" versus "PM application"), and its own sheet tab. Applications sort themselves by role, score themselves against the right JD, and rank themselves. Practical tip: for PDF resumes, add an Extract from File node between Download Resume and Parse Resume with AI, and point the parsing prompt at its extracted text — scanned or image-based PDFs otherwise parse poorly.
The retail chain hiring seasonal staff
A retail ops manager needs forty seasonal hires in two weeks and receives 500 applications across three stores. Speed is the entire game: the workflow parses and scores overnight, and the morning sheet is a ranked list per store. The summary field lets her scan a candidate in ten seconds, and the risks field flags the obvious mismatches. She lowers the High Score? threshold for volume roles — the interview itself is the real filter at this scale, and the workflow's job is just to order the queue sensibly. Practical tip: sort the sheet by score descending and work top-down; the normalize step keeps every row in identical shape, so the sheet stays scannable at 500 rows.
The non-technical founder screening developers
A non-technical founder cannot tell a strong React resume from a padded one, which used to mean either reading everything anxiously or delegating blindly. The parsed skills and years of experience fields give her structured facts, the summary gives her a plain-language read, and the score measures fit against the JD her technical advisor helped write. She reviews the ranked sheet, reads the top summaries, and personally sends the interview invites from the drafts — the human-in-the-loop design means the workflow advises but never acts on her behalf. Practical tip: have the technical advisor review the Job Description node text once — the entire scoring quality flows from how precise that posting is.