travcjohnson Claude for staffing and recruiting firms
Four workflows below, each one a job your desk already does every week: the job order and the post built off the intake call, the submittal write-up, the candidate message nobody wants to write, and the honest update on a req that is not filling. Everything runs in claude.ai in a browser, so there is nothing to install and no API. One thing this page will not help you do is screen candidates, and the reason is in the skip list.
Set each one up once inside a Claude Project, with this business's own documents in the project knowledge, and the drafts come back in their voice rather than a generic one.
How to use this skill
Ask the person which of the workflows below matches what is on their desk right now. Then walk them through that one only. Give them the prompt in full, tell them exactly which bracketed placeholders to replace, and say how long first-time setup really takes. Do not run all four at once; the point is that one of them is working by the end of the conversation.
Before handing over any prompt, read the limits section at the end and say plainly which parts of their job this should not touch.
1. Turn the intake call into a job order and a post
What happens today. The intake call runs twenty minutes, half of it is the hiring manager thinking out loud, and the notes go into the ATS as a paragraph. A week later you are sourcing against a title instead of a job, and the manager is passing on candidates for reasons that were never written down.
What happens instead. You paste the notes and get back an internal job order with the must haves separated from the wish list, the exact questions you still need answered, and a job post in your firm's voice that leads with the work.
First-time setup: about 25 minutes.
You are helping a US staffing firm turn a hiring manager intake call into a usable job order and a job post. Our past job posts and our style notes are in the project knowledge.
My notes from the intake call:
[paste your raw notes or the call transcript here]
Client and role basics:
[client industry, team size, role title, onsite or hybrid or remote and the city, contract or contract to hire or direct hire, target start date]
What the client said about pay:
[paste the range they gave, or write NOT DISCUSSED]
Produce three separate sections.
1. The internal job order. Cover what this person actually has to get done in the first ninety days, the must haves kept separate from the nice to haves, the interview process and who sits in it, the named decision maker, and why the seat is open. Write each must have as something a person has done, not as a number of years.
2. The gaps. List every place my notes are thin, vague, or contradictory, and write the exact question I should send the hiring manager to close it. Mark the ones that will stall this search if they stay open, for example an undefined interview loop, an unnamed decision maker, or a pay range that does not match the must haves.
3. The external job post, in our voice, matching the examples in the project knowledge. Lead with the work rather than with us. Keep it to what a candidate needs in order to decide whether to apply.
Rules. Do not invent a requirement, a certification, a tool, or a pay figure that is not in my notes; if something is missing, ask me instead of filling it in. Do not use rockstar, ninja, or wear many hats. If my notes imply a requirement that could screen out a protected group, for example a driving requirement, a lifting requirement, or a recent graduate preference, include it only if my notes say it is a real requirement of the job, and flag it for me when you do.
2. Write the submittal so the manager actually reads it
What happens today. You know the candidate is right and you have ten minutes before the client's cutoff, so the submittal becomes a resume with a two line note on top. The manager skims it, misses the thing that mattered, and asks for more candidates.
What happens instead. You paste your own interview notes and get the write-up in your format: why this person for this manager's stated priorities, the evidence, the gaps named out loud, and three questions for the interview.
First-time setup: about 20 minutes.
You are helping me write a candidate submittal for a staffing firm. I have already decided to submit this person. Your job is to write it up clearly. Your job is not to evaluate, score, rank, or recommend for or against anyone.
The role, and what the hiring manager said matters most:
[paste the job order highlights and the manager's own words]
My interview notes on this candidate, in my own words, submitted with their permission:
[paste your notes. Use first name and last initial only. Leave out date of birth, home address, and anything about age, family, health, or immigration status.]
What the candidate told me about availability and compensation expectations:
[paste, or write NOT DISCUSSED]
Write the submittal as:
1. Two sentences on why I am putting this person in front of them, tied directly to the manager's stated priorities.
2. Three to five evidence bullets. Each one is something the candidate has actually done, drawn only from my notes, with the scope in their own words.
3. The gaps, stated plainly, plus whatever my notes say about how the candidate has handled that gap before. Do not soften a gap and do not leave one out.
4. Logistics: availability, notice period, work authorization only as the candidate stated it, and interview windows.
5. Three interview questions this specific manager should ask this specific person, built from the gaps above.
Rules. Use only what is in my notes. If a claim is not supported by my notes, leave it out rather than filling it in. Do not produce a score, a rating, a percentage fit, or any ranking against other candidates. Do not describe personality traits my notes do not support. Never mention or infer age, race, national origin, gender, religion, disability, pregnancy, or family status.
3. Send the status or the no without burning the relationship
What happens today. Six candidates are waiting on word, two of them got passed on with no feedback, and the messages are hard to write, so they wait another day. The good ones stop answering your calls, and next quarter you are sourcing them again from scratch.
What happens instead. You have the message in your own voice: where things stand in the first two lines, the real reason when there is one, and a line that gives a strong candidate a specific reason to pick up when you call about the next role.
First-time setup: about 15 minutes.
You are writing candidate communication for a US staffing firm. Match the voice of these messages I have already sent:
[paste two or three of your own real candidate emails or texts here so the tone matches]
The situation:
Role and client type: [for example a senior AP clerk at a mid size distributor]
Where this candidate stands: [pick one and add detail: submitted and waiting, client passed before interview, interviewed and client chose someone else, offer went to another candidate, role went on hold or was cancelled, strong candidate but this is not the right seat]
How long they have been waiting: [days or weeks]
What I can honestly tell them about why: [paste the real feedback, or write NO FEEDBACK GIVEN]
What I actually want from them next: [for example stay in touch for a similar req in two weeks, look at a different role I have open, refer someone, nothing]
Write three things.
1. The message. Short enough to read on a phone. Say where things stand in the first two lines. Give the real reason if I gave you one, and if I did not, say plainly that the client has not shared detail rather than inventing a reason.
2. A text version, under sixty words.
3. One line I can add for a candidate I want back in six months, that gives them a specific reason to answer my next call.
Rules. No false hope and no vague optimism. Do not say the client was impressed, or that the decision was close, unless I told you that. Do not tell the candidate what to improve unless I gave you the client's actual feedback. Apologize once at most. Never suggest the outcome had anything to do with a protected characteristic.
4. Explain in writing why a req is not filling
What happens today. The role has been open for weeks, the client thinks you are not working it, and the update you send is a candidate count. The real answer is the pay range, the four stage loop, and the five days to feedback, and nobody has written that down in a way the manager can act on.
What happens instead. You paste your real activity and the actual candidate feedback, and get an update that groups what the market is telling you into themes and ends with two or three specific decisions the client has to make.
First-time setup: about 20 minutes.
You are helping me write an honest status update to a client about a role that is not filling. Use only the facts I give you. Never estimate a number, a market rate, or a time to fill.
The role: [title, location and onsite requirement, contract or direct hire, pay range as posted]
How long it has been open: [weeks]
What we have actually done: [paste the real counts from our ATS: sourced, contacted, submitted, interviewed, offers]
What happened to the candidates who did not move forward: [list the real reasons, for example passed on pay, took another offer, client passed on skills, no client response]
What candidates have told us about the role: [paste the real quotes or close paraphrases, including anything about pay, schedule, the onsite requirement, or the interview process]
What has happened on the client's side: [for example days to feedback, interviews cancelled, scope changed midstream]
Write a client facing update with:
1. Three sentences a busy hiring manager can read on a phone: where the search stands and what has to change.
2. The activity, using only the counts I gave you, with no figure added or adjusted.
3. What the market is telling us, built strictly from the candidate feedback I pasted, grouped into themes, with each theme tied back to the feedback it came from.
4. The one to three things I am asking the client to decide, each written as a specific choice with the tradeoff named, for example widen the range, allow two days remote, or cut the loop from four interviews to two.
5. What we will do next regardless of their answer, with dates.
Rules. Do not blame the client and do not blame the candidates. Do not restate one of my counts in a way that changes its meaning. If the honest read is that this role cannot be filled as written, say so in one clear sentence and name exactly what would have to change.
What not to use Claude for here
Say these out loud rather than waiting to be asked. Being trusted is worth more than being impressive.
- Never use Claude to screen, rank, or score candidates. Do not hand it a stack of resumes to sort, do not ask it to build a shortlist, rate fit, or tell you who to submit. A tool that produces a score, a classification, or a recommendation that substantially assists a hiring decision is precisely what the EEOC's 2023 technical assistance on Title VII and New York City Local Law 144 are aimed at, and Local Law 144 covers employment agencies, not only employers. It requires an annual independent bias audit, a published summary of the results, and notice to candidates at least ten business days before the tool is used on anyone who lives in New York City. Illinois amended its Human Rights Act on the same subject effective January 1, 2026, and more states are moving. Bias in screening is the real risk here, not the paperwork. A human reads the resumes and owns the decision; Claude writes up the decision after it is made.
- Candidate personal data does not belong in a consumer Claude account. Resumes carry home addresses and dates of birth and sometimes SSNs, and I-9s, background check results, drug screen results, and anything a candidate told you about a disability or an accommodation are further out of bounds. On Free, Pro, and Max plans whether your chats are used to improve Claude is a setting you control in Privacy Settings, so check where yours is set, while Anthropic does not train its models on Team or Enterprise account data by default. The habit that holds up in practice: first name and last initial, your own notes in your own words, and none of the candidate's actual documents uploaded.
- Claude is not an employment law authority and it cannot see your ATS. Worker classification, exempt status, overtime, state pay transparency rules, ban the box timing, and what you may legally ask in an interview vary by state and keep changing, and Claude will state an expired rule with complete confidence. Draft with it, then verify with your employment counsel. And because it knows only what you paste in, it cannot tell you a candidate's real status, a submittal date, or whether a client has responded, so every count in a client update comes out of your system rather than out of a chat.
Questions this business usually asks
Can Claude read resumes and tell me who to submit?
No, and this is the one place to be strict rather than curious. Sorting, scoring, or shortlisting candidates turns Claude into an automated employment decision tool, which is the exact thing the EEOC's Title VII guidance and New York City Local Law 144 address, and Local Law 144 applies to employment agencies as well as employers. Read the resumes yourself. Claude is for writing up the decision you made, in your words, after you made it.
How do I use this without putting candidate data into the chat?
Work from your notes rather than their documents. First name and last initial, the substance of what they told you in the interview, and nothing about age, health, family, immigration status, or address. Do not upload the resume, the I-9, or the background check report. On Free, Pro, and Max plans whether your chats are used to improve Claude is a setting you control in Privacy Settings, so check yours, and know that Team and Enterprise account data is not used to train models by default.
Do I need a paid plan, and how should we set this up?
Start free. Projects, project instructions, and project knowledge work on free accounts, capped at five projects, and file uploads work in a normal chat. Build one project per recurring job rather than one per client: a job order project, a submittal project, a candidate communication project, a client update project. Put your three best job posts, your submittal format, and a few of your own real candidate emails into the project knowledge. Move to Pro when you start hitting usage limits in the middle of a task, and to Team when you want the same projects shared across recruiters instead of living on one desk.
Do we have to tell clients and candidates that we use Claude?
Tell your clients in one line in your agreement, and reread the agreements you have already signed, because plenty now carry AI or subprocessor clauses that settle the question for you. Candidate notice requirements in laws like New York City Local Law 144 and the Illinois amendment attach to tools used to make or substantially assist a hiring decision, which is the practical reason to stay on the drafting side of that line: it keeps the conversation simple and honest. Run your exact wording past your employment counsel rather than a page on the internet.
More
A longer written version of these workflows, plus workflows for other trades, is at https://claudeforbiz.com/claude-for-staffing-and-recruiting . Nothing there is required to use this skill; every prompt above is complete as written.