Claude for recruiters: a practical guide
How recruiters in Ethiopia can use Claude and other AI assistants to write better briefs, screen faster and stay fair.
AI assistants like Claude are now part of many recruiters' daily work. Used well, they save hours every week and raise the quality of everything a recruiter produces: job descriptions, outreach, interview plans and candidate summaries. Used badly, they produce generic copy, repeat bias and put candidate data at risk.
This guide is for recruiters and talent teams in Ethiopia who want to use AI well. It focuses on practical workflows, the prompts that work, and the rules that keep you fair and trustworthy.
What AI is good at in recruiting
- Turning a messy hiring manager conversation into a clear role brief.
- Drafting job descriptions that are specific, inclusive and readable.
- Designing structured interview plans and scorecards.
- Summarising interview notes into a consistent format.
- Drafting outreach messages you then personalise.
- Preparing market briefs from information you provide.
What AI should not do
- Make hiring decisions or rank candidates on its own.
- Screen people out based on names, photos, age, gender, ethnicity, religion or other protected characteristics.
- Receive personal data you do not have permission to share.
- Invent facts about a company, a salary range or a candidate.
The simple rule: AI drafts, a person decides. Every decision about a candidate should be made by a person, using criteria that were agreed before anyone applied.
Before you start: set up your workspace
A few minutes of set up makes every later request better. Most AI assistants, including Claude, let you save standing instructions or reference files so you do not repeat yourself in every chat.
- Write a short house style: your company's tone, preferred spelling, words to avoid and the standard sections of your job descriptions.
- Save an approved company description and a summary of your benefits, so the assistant never has to guess.
- Keep templates for your role brief, scorecard and interview summary.
- Note your data rules at the top: no personal details in prompts unless the tool is approved for them.
How to write prompts that work
- Give context: the role, the company type, the level and the audience.
- Give the material: your notes, the brief or the scorecard. The assistant can only work with what you provide.
- Say exactly what you want back: the sections, the format and a word limit.
- Ask it to flag gaps and assumptions instead of filling them in.
- Review, then ask for specific changes: "shorter", "less formal", "remove the second requirement". Two or three rounds usually get you there.
Workflow 1: from intake call to role brief
After your intake call with the hiring manager, paste your notes into Claude and ask for a structured brief.
The last part is the most useful. AI is good at noticing gaps: a missing budget, an unclear reporting line, conflicting requirements. Send those questions back to the hiring manager before you write the job description.
Workflow 2: job descriptions people want to read
Give Claude the approved brief and ask for a job description in a specific structure, with a word limit. Ask it to remove jargon and requirements that are not truly needed. Long lists of requirements discourage strong candidates, especially women and people changing sector, from applying.
Workflow 3: structured interviews and scorecards
Structured interviews are fairer and more predictive than unstructured conversations. Ask Claude to design one for you.
A good scorecard answer describes behaviour you could observe, not a personality trait. "Explains the trade offs of two options and picks one with a reason" can be scored. "Is a strategic thinker" cannot. If the draft uses vague descriptions, ask Claude to rewrite them as observable behaviour.
Review the output with the hiring manager and agree the final version before interviews start. Use the same questions for every candidate.
Workflow 4: consistent interview summaries
After each interview, paste your own notes and ask for a summary against the scorecard. Remove names and personal details before you paste if your policy requires it. Always check the summary against your notes, and never let the summary replace your own judgment.
Workflow 5: outreach that sounds like you
Ask for three short versions of an outreach message, then edit the best one in your own voice. Mention something specific about the person's work. Generic AI outreach is easy to spot and damages your reputation.
Workflow 6: candidate communication
Candidates remember how they were treated long after they forget the questions. AI can help you write clear, respectful messages quickly: interview invitations with full logistics, updates when a process is delayed, and rejections that are kind and specific.
Always check that every fact is right, add anything personal you know, and send it from your own account. Never send a message you have not read in full.
Workflow 7: briefing the hiring manager
At the end of a search, hiring managers want a clear picture of the shortlist, not a pile of notes. Paste your anonymised scorecard summaries and ask for a side by side comparison against the agreed competencies, with open questions for the final round. The assistant organises the evidence. The hiring manager and panel make the decision.
Protecting candidate data
Candidates trust you with their information. Before you use any AI tool, check your company's policy and the tool's data settings. As a default:
- Remove names, phone numbers, emails and ID numbers from anything you paste.
- Do not upload full CVs unless your company has approved the tool for that use.
- Keep final records in your own systems, not in chat histories.
Also check whether the tool you use keeps chat history, whether it can be turned off, and who in your company can see shared projects. Settings differ between products and plans, so read them rather than assume.
Staying fair
AI models can reflect biases in the data they learned from. Keep a person responsible for every decision, use structured criteria agreed in advance, and regularly check whether your shortlists are diverse. If an AI suggestion ever relies on a protected characteristic, discard it.
Working in Amharic and other languages
Many roles in Ethiopia are advertised in English, but outreach, candidate messages and some job adverts work better in Amharic or another local language. AI assistants can draft and translate, but quality varies by language and by topic. Treat any translation as a first draft and have a fluent speaker review it before it goes out, especially for job adverts and offers, where a small error can change the meaning.
Worked example: one role, end to end
Take a search for an operations manager at a manufacturing company. After the intake call, the recruiter pastes her notes and gets a role brief with four open questions, including one about whether the role manages the warehouse team. She sends those questions to the hiring manager and gets answers the same day.
She then asks for a job description under 450 words and removes two requirements the assistant flagged: a specific master's degree and familiarity with one software product the team could teach. Next, she drafts a structured interview with five competencies and agrees it with the hiring manager before the first call.
After each interview she writes her own notes and uses the assistant to format them against the scorecard, checking every line. At the end, she shares a comparison of the finalists with the panel, who score independently and then decide together. Every judgment is made by a person. The assistant saved time on writing and formatting, which is where most of the hours used to go.
Common mistakes to avoid
- Publishing AI drafts without editing. Readers notice, and the role sounds like every other.
- Pasting full CVs or interview recordings into tools that are not approved for personal data.
- Asking the assistant which candidate is best. It does not know your candidates, and it should not decide.
- Trusting salary figures or market claims from the assistant. Use your own data and sources.
- Changing interview questions between candidates after the plan is agreed.
A starter prompt library
- Turn these intake notes into a role brief and list the questions I should clarify.
- Write a job description under 450 words in this structure. Flag unnecessary requirements.
- Design a structured interview with five competencies and scoring guidance.
- Summarise my notes against this scorecard. Do not add information that is not in my notes.
- Write three short outreach messages for this role. Keep each under 90 words.
AI will not replace good recruiters. It will make good recruiters faster and more consistent. If you hire through HireHabesha, our team uses the same principles: people decide, criteria come first, and candidate data is protected.
Frequently asked questions
Can I paste CVs into Claude to compare candidates?
Only if your company has approved the tool for personal data and you have a lawful basis to process it that way. Even then, use it to organise information against agreed criteria, not to rank people. When in doubt, remove personal details first.
Should I tell candidates we use AI in our hiring process?
Being open builds trust. A short line in your candidate information, saying that you use AI tools to help draft materials and organise notes, and that all decisions are made by people, is a good default.
Can AI screen hundreds of applications for me?
It can help you organise applications against criteria you have defined, but screening decisions should stay with a person. If you use any automated step, test it for unfair patterns, document how it works, and make sure a person reviews borderline cases.
Which AI assistant should we use?
Choose the tool your company has approved for your data, then learn it well. Claude, ChatGPT and similar assistants can all support the workflows in this guide. Clear prompts and good review habits matter more than the brand.