AI for hiring: how to screen applicants, write job posts, and onboard faster

Hiring is the task most small business owners dread more than tax season.

It starts with staring at a blank page trying to write a job description that will attract the right person without sounding like a corporate HR form. Then you post it and wait. Then you get 80 applications, 60 of which are from people who clearly didn't read the description. Then you spend a weekend reading CVs, scheduling calls, preparing interview questions, running interviews, checking references — and at the end of all of it, you either make a great hire or you make a mistake that costs you months of your life to fix.

Most small businesses do all of this without any dedicated HR function. It falls entirely on the owner, usually on top of everything else they're already doing.

AI doesn't make hiring easy. But it makes it significantly less painful — and it removes several of the steps that eat the most time without requiring the most judgment. This guide walks through each stage: writing the job post, screening applications, running better interviews, and getting a new hire up to speed faster.

Writing a job post that actually attracts the right person

Most small business job posts fail before a single application comes in. They either describe the job in such generic terms that anyone could apply, or they use corporate HR language so stiff that the kind of person you actually want — someone practical, direct, capable of figuring things out — reads it and moves on.

A good job post does three things: it describes the work honestly, it signals what kind of place this is, and it makes the wrong people self-select out.

AI is excellent at drafting job posts once you give it the raw material. The key is feeding it your real version of the role — not the sanitised HR version.

The job post prompt

Before you touch AI, answer these questions in plain language. Don't overthink them:

What does this person actually do on a typical day? (Be specific. Not "manages customer relationships" — "responds to customer enquiries, processes returns, updates the order tracker.")
What's the hardest part of this job?
What kind of person has thrived in similar roles in the past? What made them good?
What kind of person has struggled or left? What went wrong?
What does working here actually feel like — the honest version?
What are the non-negotiable skills or experiences someone must have?
What can be learned on the job?

Now use this prompt:

I'm writing a job description for a [role title] at my [type of business].

Here is the honest version of the role:
[Paste your answers to the seven questions above]

Write a job description that:
- Opens with one paragraph that describes the role and the business in plain, honest language
- Lists key responsibilities as specific tasks, not vague functions
- Lists what we're looking for — skills, qualities, experience — split into "must have" and "nice to have"
- Includes a short, honest description of what working here is like
- Ends with clear application instructions
- Sounds like a real person wrote it, not an HR department
- Does not use phrases like "fast-paced environment," "team player," "passionate," or "synergies"

Length: 350–500 words.

The output will need editing — particularly the "what it's like to work here" section, which AI will produce in generic terms that you'll need to replace with something specific and real. But the structure will be solid and the time from blank page to finished post drops from two hours to twenty minutes.

One addition that dramatically improves application quality

At the end of your job post, add a specific instruction that filters out anyone who didn't read it carefully.

Something like: "To apply, send your CV along with a short paragraph (100 words maximum) explaining the last time you solved a problem at work without being told how to. Applications without this paragraph will not be considered."

This single addition typically cuts your application volume by 40–60% while improving the quality of what remains. The people who follow the instruction are the people who follow instructions. The people who don't have already told you something useful.

Screening applications without reading 80 CVs

Once applications come in, the screening stage is where most owners lose the most time. Reading every CV properly takes 5–10 minutes per application. Eighty applications is nearly seven hours of reading before you've spoken to anyone.

AI can do the first pass.

Building a screening rubric

Before you feed any CVs to AI, define what you're actually looking for in concrete terms. The clearer your criteria, the more useful the AI output.

Write a rubric with three tiers:

Must have (automatic yes consideration): List 3–5 things that are genuinely non-negotiable. Real skills, specific experiences, demonstrable evidence of something. Not "good communicator" — "has directly managed customer complaints in writing."

Good to have (adds points): List 3–5 things that would strengthen a candidate but aren't dealbreakers. Experience in a specific industry, familiarity with a particular tool, evidence of initiative.

Automatic no: List 2–3 things that immediately disqualify. This might be no relevant experience whatsoever, evidence they've job-hopped every three months for five years, or anything that conflicts with a legal requirement of the role.

Using AI to screen

With your rubric built, paste each CV into Claude or ChatGPT with the following prompt:

Here is a job description and screening rubric:

[Paste job description]

Rubric:
Must have: [list]
Good to have: [list]
Automatic no: [list]

Here is a candidate's CV / application:
[Paste CV text]

Assess this candidate against the rubric. Give me:
1. A pass/fail recommendation based on must-have criteria
2. A brief note (2–3 sentences) on their strongest relevant points
3. A brief note on any gaps or concerns
4. An overall rating: Strong / Adequate / Weak

Be direct. I am a busy business owner, not an HR professional.

This takes about two minutes per application. For eighty applications, you're looking at two to three hours rather than seven — and the output gives you a stack-ranked shortlist rather than a pile of CVs to reread.

Important caveat: AI screens based on what's written, not what's true. It can miss candidates whose CVs are poorly written but who would be excellent in the role — particularly people earlier in their career, people who've been out of the workforce, or people who simply aren't good at writing CVs. Use the AI screen to identify your definite-yeses and definite-nos. For the middle band, read those CVs yourself before making a decision.

Better interview questions in ten minutes

Generic interview questions produce generic interview answers. "Tell me about yourself" and "where do you see yourself in five years" tell you almost nothing useful about whether someone will do the actual job well.

The best interview questions are role-specific, behaviour-based, and designed to surface evidence rather than opinions. AI is good at generating them once you describe the role honestly.

The interview question prompt

I'm interviewing candidates for a [role] at my [type of business].

The key challenges in this role are:
[List 3–4 specific things this person will need to handle well]

The qualities that matter most are:
[List 3–4 qualities, as specific as possible]

Write 8–10 behavioural interview questions designed to surface evidence of these qualities and capabilities.

Format each question as:
- The question itself
- What a strong answer would include (2–3 sentences)
- A follow-up probe question

Do not include generic questions like "tell me about yourself" or "what's your greatest weakness."

The output gives you a complete interview guide — not just questions but what to listen for in the answers. This is the thing most small business owners never have because they don't have an HR team to build it for them.

The one question worth asking every candidate

Beyond the structured questions, there's one open question that reveals more about how someone thinks than almost anything else:

"Tell me about a time things went wrong at work and it was at least partly your fault. What happened and what did you do?"

The answer tells you how self-aware the person is, whether they take responsibility, how they handle adversity, and whether they can tell a coherent story under mild pressure. You want a specific example, not a deflection. The candidate who can't think of a single time something went wrong, or who blames everyone else in the telling, is showing you something important.

Reference checks that actually work

Most reference checks are useless because most referees — knowing the person chose them — give a polished, positive account of a candidate they're unlikely to criticise directly.

The question that cuts through this is one of the most effective interview and reference techniques in existence, sometimes called the "reference check question":

"On a scale of 1 to 10, how would you rate [candidate's name] overall as an employee?"

Almost no one gives a 10. If they say 8 or 9, ask: "What would it have taken for them to be a 10?"

The answer to that second question is where the real information lives. A referee who says "to be honest, she struggled with prioritisation when things got busy" has just told you something specific and true that they would never have volunteered in a general question.

AI use here is limited — reference checks need a real phone call. But you can use AI to help you prepare a structured set of reference questions in advance, and to write a brief summary of each reference call while the conversation is fresh.

Getting a new hire up to speed faster

The first 30 days of a new hire's time are the most expensive and the most wasteful in most small businesses. The owner is too busy to train properly. The new person asks the same questions repeatedly. Basic things that should be documented aren't. Mistakes happen that could have been prevented.

AI won't replace good onboarding — but it makes building an onboarding system fast enough that you'll actually do it.

Building an onboarding checklist

I'm onboarding a new [role] at my [type of business].

In their first 30 days, they need to:
- Learn [list of things they need to know]
- Be able to do [list of tasks they need to be able to handle independently]
- Meet [list of people or relationships that matter]
- Have access to [list of tools, systems, accounts]

Write a 30-day onboarding checklist structured as:
- Week 1: orientation and observation
- Week 2: guided practice
- Weeks 3–4: supervised independence

Include a brief note on what "done" looks like for each item.

The result is a structured 30-day plan that you can hand to a new hire on their first day. It answers the "what should I be doing?" question before they ask it, which reduces interruptions to you and makes the new person feel like they're in capable hands.

Writing the procedures they'll follow

The other onboarding gap in most small businesses is undocumented process. The owner knows how to handle a customer complaint, process a return, open the shop on a Monday morning, or close out the day's accounts — but it's all in their head. When a new person starts, they either interrupt the owner constantly or they make it up and do it wrong.

AI makes procedure documentation fast:

I need to document how we [describe the process — e.g. handle a customer complaint, process a refund, open the shop].

Here is how we currently do it:
[Describe the process in plain language, as roughly as you like — bullet points, stream of consciousness, whatever you have]

Turn this into a clear step-by-step procedure that a new employee could follow on their first day without any prior knowledge of our business.

Include:
- What to do before starting
- Each step in order, numbered
- What to do if something goes wrong
- Who to ask if they're unsure

Write it in plain, direct language. Assume the reader is competent but has never seen our business before.

Do this for the ten most common processes in your business. It takes about 20 minutes per process if you do the rough description yourself. The result is the beginning of an operations manual — something most small businesses never have and would benefit enormously from.

The 30-day check-in

One more AI use that's easy to overlook: use it to prepare for your 30-day review conversation with a new hire.

I'm having a 30-day check-in conversation with a new [role] I hired.

I want to:
1. Get honest feedback on how onboarding has gone from their perspective
2. Identify any gaps in their training or knowledge
3. Set clear expectations for the next 60 days
4. Make sure they feel supported and know what success looks like

Write 8 questions I should ask in this conversation, structured to get honest answers rather than polished ones.

A good 30-day check-in takes 45 minutes and surfaces problems early — when they're still fixable — rather than at the three-month review when they've become habits.

What to watch out for: bias, compliance, and keeping the human in the process

Using AI in hiring comes with real responsibilities that are worth being explicit about.

Bias: AI trained on historical data can perpetuate historical biases — favouring certain educational backgrounds, writing styles, or career trajectories that correlate with factors you shouldn't be making decisions on. Always review AI screening recommendations yourself, particularly for candidates from non-traditional backgrounds. The AI screen is a shortlist tool, not a decision-making tool.

Legal compliance: In most jurisdictions, you cannot use certain criteria in hiring decisions — age, family status, national origin, disability status, and others depending on where you operate. If you're feeding CVs to AI for screening, check that your rubric doesn't inadvertently include any criteria that would be discriminatory if applied by a human. When in doubt, remove any criterion that isn't directly about the ability to do the job.

The human element: The best hire you ever make will probably come from a conversation that went somewhere unexpected — where you saw something in a person that wasn't in their CV and wouldn't have been flagged by a screening algorithm. AI helps you get to those conversations faster by clearing the mechanical work. But the conversation itself is yours to have.

Where to start

If you're actively hiring right now: use the job post prompt today. Write your honest answers to the seven questions, run the prompt, edit the output. Post it with the specific application instruction at the end. See whether your application quality changes.

If you're not currently hiring but know you will be: spend two hours building a basic onboarding checklist and documenting your five most important business processes using the prompts above. Do it now, while you have the time, so you're ready when you need to move quickly.

If you hired someone recently and onboarding felt chaotic: use the procedure documentation prompt for the ten tasks that caused the most confusion. Give the resulting documents to your current hire and ask them to flag anything that's wrong or missing. You'll end up with accurate documentation and a more confident team member at the same time.

Want a hiring template pack — covering job post structure, screening rubric, interview question bank, and 30-day onboarding checklist — all in one document you can reuse? Subscribe to AInstein and we'll send it straight to your inbox alongside a weekly briefing on AI tools and tactics that are genuinely useful for running a small business.

Next read: How to automate your email marketing with AI — your team is growing, now make sure your marketing keeps pace without adding to your workload.