Services
Explore
Free AI assessmentAnswer 15 questions about one process and see where AI could help.
Start an assessmentHow do I use AI for hiring at a small business?
Where AI can help with drafting and preparation in hiring, and the legal, bias and data risks to understand before candidate information goes near it.
The short answer
Use AI to draft and prepare, and keep hiring decisions with people
Use AI to draft job descriptions, interview questions, correspondence and onboarding material. Do not let it screen, score or rank candidates, or make decisions about them, without a person reviewing each one. Before any candidate information goes into an AI tool, check that the tool is approved for that data and that your privacy notice covers the use.
Hiring decisions affect people's livelihoods and your business, and recruitment is covered by employment and data protection law. AI can reduce some of the repeated drafting and help you apply the same criteria to each candidate. Judging people stays with the people doing the hiring.
This guide covers five places where AI helps, five areas where it creates risk, and a six-step workflow with prompts. The employment and data protection points here are general information. Have your own process checked by a qualified adviser.
Where AI helps
Five places AI can save time in hiring
These tasks use AI to draft, structure and organise information. A person makes every decision about a candidate.
- Writing job descriptions
Brief the AI with the role title, key responsibilities and must-have requirements. You will still need to adjust the seniority level and check that the language is inclusive, and the draft gives you something to edit from the start.
- Generating structured interview questions
Give the AI the job description and ask for competency-based questions, follow-up probes and a scoring rubric. Using the same question set for each candidate at the same stage keeps the process consistent and records how each person was assessed if a decision is challenged.
- Summarising applications
Before you do this, confirm the AI tool is approved for candidate data, that your privacy notice covers the use, and that a person will make each longlisting decision. Then ask for a one-paragraph summary of each application against your stated criteria. Removing a name does not make a CV anonymous: addresses, dates, schools and employers can still identify someone or indicate a protected characteristic.
- Drafting offer letters and rejection emails
Both take time to do well. Give the AI your offer terms or your reasons for declining, and ask for a draft in your company tone. You add the specific context and check the details. A clear, courteous rejection email leaves a good impression on people who might apply again.
- Building onboarding documentation
First-day checklists, 30/60/90-day plans, role-specific training guides and FAQ documents for new starters can be drafted by AI from your existing process documents and a description of the role. This material uses little candidate data. Check it before use, because it describes your policies to a new employee.
Where it's risky
Five areas where AI creates legal, bias or data risk
The Equality Act 2010 and UK data protection law both apply to recruitment. These are the areas where AI use can cause problems. Have your process checked by a qualified adviser before using AI with candidate information.
RiskLegal
A tool that scores, ranks or filters candidates without human review can screen out people with protected characteristics, such as age, sex, race or disability. Using a tool does not move responsibility for the hiring decision away from your business. Data protection law also restricts decisions made solely by automated means. Keep a person reviewing each decision and take advice before using a screening tool.
RiskBias
Screening tools trained on past hiring data learn what past hires looked like, including patterns that reflected bias. Tools that assess 'culture fit', analyse communication style, or draw inferences from names and locations carry this risk. Before adopting any AI screening product, ask the vendor how they test for bias and what data the model was trained on.
RiskLegal + Bias
Some AI interview tools claim to assess personality, motivation or fit from video or audio. Ask the vendor for evidence that the assessment measures what it claims, and how it has been tested for bias against candidates with accents, disabilities or neurodivergent traits. Do not use such a tool to judge candidates without that evidence and specialist advice.
RiskCompliance
Right-to-work checks, DBS checks and reference checks follow set procedures with legal consequences. Follow the official guidance for each and keep the decision with a person. AI can draft the request letters.
RiskData
CVs contain personal data. Before pasting one into any AI tool, check the provider's terms for how data is stored and used, whether your business has an agreement covering that use, and whether your privacy notice tells candidates. Consumer plans, business plans and API access can have different terms. Removing a name does not necessarily anonymise a CV, because other details can still identify the person.
A sample hiring workflow
A six-step AI hiring workflow with prompts
These prompts work in general-purpose AI assistants such as ChatGPT, Claude and Gemini. Steps 3 and 4 use candidate information: before them, confirm the tool is approved for that data and that a person will make each decision. Fill in the bracketed sections with your own detail.
| Title | Num | Prompt | Note |
|---|---|---|---|
| Write the job description | 01 | You are helping a [company size] [company type] in [industry] hire a [job title]. Write a job description with: a 2-sentence company intro, a 5-point 'what you'll do' section, a 4-point 'what we need' section (must-haves only), and a 3-point 'what we offer' section. Tone: [direct / warm / formal]. Avoid corporate jargon. | Review the seniority level. Replace language that may put some candidates off, such as 'ninja' or 'rockstar'. Check that each must-have is a real requirement, and move the rest to a nice-to-have list. |
| Generate interview questions | 02 | Based on this job description [paste JD], write: (1) five competency-based interview questions relevant to the core responsibilities, (2) two follow-up probes for each question, (3) a brief scoring rubric (1 to 3 scale) for each question. Focus on evidence of past behaviour, and avoid hypothetical questions. | Use the same question set for each candidate at the same stage. It keeps the process consistent and gives you a record of how each candidate was assessed. |
| Summarise applications for longlisting | 03 | Summarise this job application against these criteria: [list your must-have criteria from the JD]. For each criterion, state: met / partially met / not evidenced. Then write a two-sentence overall summary. Do not make a hiring recommendation. | Use an approved tool only. Removing the name is a start, but addresses, dates, schools and employers can still identify someone or indicate age. Remove what you can, and make the longlisting decision yourself from the summaries. |
| Prepare for the interview | 04 | I am interviewing a candidate for [role]. Here is their CV with identifying details removed: [paste CV]. Here are my interview questions: [paste questions]. Flag any areas where their background is unclear and suggest follow-up questions to explore them. | Keep the core questions the same for each candidate, and use these follow-ups only to clarify points in the CV. |
| Draft the offer or rejection | 05 | Write an offer letter for the role of [title] at [company]. Salary: [amount]. Start date: [date]. The tone should be warm and clear. Include: role title, salary, start date, reporting line, and a brief note about what we value in their background. Leave a placeholder for the candidate's name, which I will add myself. | For rejections: 'Write a respectful rejection email for a candidate who reached final interview for [role] but was not selected. We noted strengths in [area]. We may have future roles that suit them.' Check each letter before sending, especially salary and dates in offers. |
| Build onboarding documentation | 06 | Create a 30/60/90-day plan for a new [job title] at a [company type]. By day 30 they should [goal]. By day 60 they should [goal]. By day 90 they should [goal]. Format as a table with: timeframe, key activities, success indicators, and who to meet. | Prepare the plan before the new starter's first day. Sharing it at offer stage gives the candidate a clear picture of the role. |
Common questions
FAQs
Is it legal to use AI in the hiring process in the UK?
It depends on how you use it. Using AI to help draft job descriptions, interview questions or correspondence carries less risk than using it to screen or score candidates. Screening and scoring raise questions under the Equality Act 2010 and UK data protection law. Keep a person responsible for each decision about a candidate, and have your process checked by a qualified adviser.
Can AI write a job description for me?
Yes, as a first draft. Give the AI the job title, the key responsibilities, the must-have and nice-to-have requirements, and your company tone. You'll still need to adjust it for your context, check the seniority level, and make sure the language is inclusive.
Will AI screening tools introduce bias into my hiring?
They can. Screening tools trained on historical hiring data can repeat patterns from that data, including any bias it contained. Tools that rate 'culture fit' or analyse language carry this risk. Using a tool does not move responsibility for the outcome away from your business, so ask vendors how they test for bias and check the outcomes yourself.
What's the best way to use AI for interview preparation?
Start with the job description, which contains no candidate data. Ask the AI for (1) structured competency-based questions for the role, (2) follow-up probes for each question, and (3) a scoring rubric so each interviewer uses the same criteria. If you want help with a specific CV, use a tool approved for candidate data and remove identifying details first. Give all candidates for the same role the same core questions.
Can AI help with onboarding after a hire?
Yes. AI can draft first-day checklists, role-specific 30/60/90-day plans, FAQ documents for new starters and training guides from your existing process documents. This uses little candidate data. Check the drafts before use, because they describe your policies to a new employee.
Do I need to tell candidates I'm using AI in the hiring process?
If any decision about a candidate is made by automated means, data protection law sets specific requirements, including telling people and offering human review. Take advice on what applies to your process. If AI is used only to draft and prepare, with each decision made by a person, your privacy notice should still explain how candidate data is used. Being open about your process is good practice whatever the legal minimum.
Talk to us
Want help applying this to your business?
Bring the task that takes your team the most time. We reply within one working day.