Claude Code for Recruiters: Sourcing to Submittals

Claude Code for recruiters takes on the writing and research that fill a recruiter's day: job descriptions, candidate research, outreach messages, interview briefs, notes and client updates. It does not decide who gets hired. That stays with people, and the rules below keep it that way.

Claude Code for recruiters means keeping each search in a folder an AI agent reads before every job: the role, the must-haves, your outreach voice and your notes. It drafts job descriptions, researches public profiles, writes outreach and turns interview notes into structured summaries. Recruiters make every decision.

Here is how it fits across a search.

One folder per search

• role.md: the job, the team, the must-haves and nice-to-haves agreed with the hiring manager, salary range and location rules.
• scorecard.md: the outcomes the hire must deliver and how each will be judged.
• voice.md: outreach messages of yours that got replies.
• candidates/: one file per candidate with what they shared and your notes.

The search, step by step

Intake and the scorecard

After the intake call, give Claude Code your notes and ask for a role summary and a scorecard: three to five outcomes the hire must deliver in their first year, each with how it will be judged. Read it back to the hiring manager. A clear scorecard makes every later step faster and fairer.

A good scorecard keeps everyone honest later. When a hiring manager falls for a candidate who interviews well but misses two of the outcomes, the scorecard is what lets you have that conversation calmly. It also protects candidates, because each one is measured against the same outcomes rather than against a gut feel.

Job descriptions

Ask for a job description built from the scorecard rather than a list of buzzwords: what the person will do, what success looks like, the real requirements and the salary range where you have to or choose to share it. Ask it to flag requirements that may be unnecessary, such as degree requirements for work that does not need one, and language that might put qualified people off.

Ask for two versions while you are at it: the full description for your careers page or job board, and a short version of four or five lines for outreach messages and social posts. The short version should lead with the problem the hire will solve, because that is what makes a passive candidate curious, and leave the requirements list for the full post.

Sourcing and outreach

For each candidate you have found, paste what is on their public profile and ask for one specific observation and a short message: why this role fits what they have done, one question, no wall of text. The same rule as sales outreach: if the first line could go to a thousand people, rewrite it. Follow-ups add something new, such as a detail about the team or the problem they would solve.

Interview briefs and notes

Before an interview, ask for a one-page brief: what the candidate has shared, the scorecard outcomes, and questions mapped to each outcome. After it, give Claude Code your notes and ask for a structured summary against the scorecard, with direct quotes. It summarises; you assess.

Structured notes also make your submittals better. When you present a candidate to a client, a short summary that maps their own words to each outcome on the scorecard is far more persuasive than "great culture fit". It shows the hiring manager exactly why you think this person can do the job, and it gives them specific things to probe in their own interview.

Client updates

Once a week, ask for a short update per search: pipeline by stage, what changed, what you need from the client, and the next step. Same shape every week.

A worked example: intake notes to a scorecard

Here is the kind of input and output you can expect. The company and role below are made up.

Intake notes, Brightline Logistics (fictional), Operations Manager
- Two warehouses, 40 staff, night shift is the problem
- Late shipments up this year, manager left in spring
- Must have run a shift team of 15+, WMS experience
- "We need someone who stays calm when the trucks are late"
- Salary range agreed with HR, hybrid not possible

Read role.md and these intake notes. Write a role summary and a scorecard
with 3 to 5 outcomes for the first 12 months. For each outcome: what
success looks like and how we will judge it. Flag anything in the notes
that is vague or might be an unnecessary requirement.

A good result looks something like this:

Outcome 1: On-time shipments back above target within 6 months.
  Judged by: weekly on-time rate, reviewed monthly with the GM.
Outcome 2: A stable night shift with a named lead on every shift.
  Judged by: night shift turnover and absence rate by month 9.
Outcome 3: The WMS used for every pick and count, no paper workarounds.
  Judged by: system audit at month 4.
Flags: "stays calm when trucks are late" needs a concrete interview
question; confirm whether the WMS brand matters or any WMS will do.

Notice what it does with the vague line from the intake. It does not invent a requirement; it turns it into a question for you to settle with the hiring manager. That habit, flagging instead of guessing, is what you want from every draft.

Keeping hiring managers aligned

Most searches go wrong in the gap between what the hiring manager said at intake and what they react to in interviews. Claude Code helps close that gap. After each round of interviews, ask for a one-page calibration note: which scorecard outcomes the candidates met, where the manager's feedback drifted from the scorecard, and the one question to settle before the next round. Send it before the debrief call so the conversation starts from evidence rather than impressions.

The same note is useful when a search stalls. If the last five candidates were rejected for a reason that is not on the scorecard, the scorecard needs updating, or the manager needs a gentle reminder of what was agreed. Either way, you can see the pattern in writing instead of arguing from memory.

A sample outreach sequence

For a passive candidate, ask for three short messages rather than one long one. The first opens with something specific from their public profile, such as a project they led or a move they made, connects it to the problem the role solves, and asks one easy question. The second, a few days later, adds one concrete detail about the team or the challenge. The third, a week after that, closes the loop politely and leaves the door open.

Keep each message short enough to read on a phone. Candidates who are not looking will give you ten seconds, so the first line has to show you read their background, and the rest has to make the role sound like a real problem worth solving rather than a list of perks.

Rules for recruiting work

• People make hiring decisions. Claude Code drafts and summarises. It does not rank, reject or recommend candidates on its own.
• Fair and lawful. Job descriptions and screening must follow anti-discrimination law. Some places regulate automated hiring tools; New York City requires bias audits for automated employment decision tools. Check the rules where you hire.
• Candidate privacy. Collect only what you need, keep candidate files on your own machine, turn off model training in your Claude settings, and follow the privacy rules that apply to you.
• Public information only for sourcing, and the platform rules of wherever you find candidates. No scraping.

Mistakes to avoid

• Letting the AI write requirements. Requirements come from the hiring manager and the scorecard. A model will happily add "5+ years" or a degree requirement nobody asked for, which narrows your pool for no reason.
• Generic outreach at scale. Fifty near-identical messages damage your reputation and your reply rate. Fewer, specific messages win.
• Summaries without quotes. A summary that says "strong communicator" is an opinion. A summary that quotes what the candidate said is evidence.
• Skipping the fairness check. Read every job description and screening question for anything that could exclude people unfairly, and check the rules where you hire.
• Putting candidate data everywhere. Keep candidate files in one place on your own machine and delete what you no longer need.

How to set this up in an afternoon

1. Create a folder for one live search and write role.md and scorecard.md from your intake notes.
2. Paste three outreach messages that got replies into voice.md.
3. Ask for a job description from the scorecard and edit it with the hiring manager.
4. Research five candidates and draft one first message each.
5. After your next interview, turn your notes into a structured summary.

Once that works for one search, copy the folder structure for every new search.

Claude Code for recruiters: FAQ

Is Claude or ChatGPT better for recruiters?

Both can draft and summarise well. Claude Code suits recruiters who manage many searches, because it keeps each search in a folder and reads the role, scorecard and notes before every job, so drafts stay specific to the role.

Which AI tool is best for recruiters?

The best setup pairs your applicant tracking system for candidate records with an AI that drafts and summarises from those records. Claude Code fits the second role, especially for outreach and interview summaries.

Can Claude Code screen candidates for me?

Claude Code can summarise what a candidate shared against your scorecard, but the screening decision should stay with a person. Automated screening raises fairness and legal risks, and some places regulate it directly.

Can I use Claude Code to apply for jobs?

Job seekers can use Claude to tailor a CV and cover letter to a role. Keep it truthful, and write the final version in your own words.

Can AI write a job description?

AI can write a strong first draft of a job description when you give it the scorecard and the real requirements. Edit it with the hiring manager, remove anything unnecessary, and check it for exclusionary language before posting.

How do recruiters use AI for sourcing?

Recruiters use AI to research candidates from public profiles and company news, summarise backgrounds and write personal first messages. The search strategy and the decision about whom to contact stay with the recruiter.

Get the hiring and outreach prompts

The AI Operator Kit includes a hiring pack (scorecard, job post and work sample), delegation briefs, the /outreach command and a 107-prompt library. For the sales side of recruiting, read Claude Code for sales.