Small Biz AI HQ

Category: HR & Recruiting AI

HR & Recruiting AI helps UK small businesses use artificial intelligence sensibly across the employee lifecycle, from writing job adverts and screening applications to onboarding, rotas, training and everyday people management. The articles are aimed at owners and managers who handle HR themselves, without a dedicated department, and who want to save time without creating legal or fairness problems. We explain what different tools do, where they help, where they can introduce bias and how to keep a human in charge of decisions about people. You will find practical templates, checklists and worked examples, along with notes on employment law, the Equality Act 2010 and handling candidate and staff data under UK GDPR. Employment rules can be complex and fact-specific, so for disputes, dismissals or anything unusual we recommend speaking to Acas or a qualified employment adviser.

  • AI Job Descriptions: A Proven Way to Attract the Right People

    AI Job Descriptions: A Proven Way to Attract the Right People

    Quick answer: Good AI job descriptions start with your own facts, not the tool: the real tasks, the must-have skills, the pay range, hours and location. Feed those into an AI assistant with clear instructions, then edit the draft for accuracy, plain English and fairness under the Equality Act 2010 before you publish it.

    AI Job Descriptions: A Proven Way to Attract the Right People

    Small businesses rarely have time to craft the perfect advert. The usual result is a hurried list of duties copied from an old post, or a generic template that could describe any job anywhere. Neither attracts the people you want. AI writing assistants can turn rough notes into a clear, well-structured description in minutes, but only if you give them the right material and check what comes back.

    Why most job descriptions miss the mark

    Candidates skim adverts quickly and decide within moments whether a role is for them. Descriptions fail when they are vague about the work, pile on requirements that are not really needed, bury practical details such as hours and pay, or use jargon that means something only inside the business.

    AI tools do not fix this on their own. Left to their defaults, they often produce enthusiastic but empty wording: phrases about exciting environments, rockstars and passion that say nothing specific. The value comes from using AI to organise and polish your own knowledge of the job. For a wider look at where automation helps and hinders recruitment, see our guide to using AI in hiring.

    Gather the facts before you open an AI tool

    Spend ten minutes on this and the draft will be far better. Write down:

    • The purpose of the role in one sentence: why it exists and what would be different if the right person joined.
    • A typical week: the five or six tasks that take most of the time.
    • Essential requirements: only the skills, licences or experience genuinely needed from day one.
    • Desirable extras: things that would help but can be learned.
    • Practical details: pay or pay range, hours, shift patterns, location, travel, contract type and start date.
    • What you offer: training, progression, flexibility, team size and anything that makes the job worth having.
    • How to apply and what happens next.

    Be honest about the less appealing parts too. If the role involves early starts or heavy lifting, saying so saves everyone time and reduces early departures.

    How to write AI job descriptions step by step

    1. Set the context. Tell the assistant you are a small UK business, what you do and who the advert is for.
    2. Paste in your facts. Include the notes above. The more specific they are, the more specific the draft will be.
    3. Give clear instructions. Ask for British English, plain language, short paragraphs, a clear essential/desirable split, and no clichés or buzzwords.
    4. Ask for structure. A summary, what you will do, what you will need, what we offer, and how to apply works well for most roles.
    5. Request a fairness check. Ask the assistant to point out wording that might put off particular groups or that sets requirements not linked to the job.
    6. Edit by hand. Correct anything inaccurate, remove anything you cannot deliver, and add your own voice.
    7. Test it on a colleague. Ask someone who does similar work whether it describes the job they would recognise.

    A simple instruction to try: “Using only the facts below, write a job advert for a small UK business in plain British English. Avoid buzzwords. Separate essential and desirable requirements. Flag any requirement that may not be necessary for the role.” The phrase “using only the facts below” matters, because it discourages the tool from inventing benefits or duties.

    Useful follow-up requests

    The first draft is a starting point. Short follow-up instructions can improve it quickly:

    • “Cut this to around half the length without losing any facts.”
    • “Rewrite the opening so it explains what the person will actually do in their first month.”
    • “List any statements here that are not supported by the facts I gave you.”
    • “Suggest a clearer job title that someone searching job boards would use.”
    • “Rewrite the requirements as skills rather than years of experience.”

    That third request is especially useful, because it asks the tool to audit its own additions. Treat its answer as a prompt for your own check rather than a guarantee.

    Tool types and effort

    You do not need specialist recruitment software to produce good AI job descriptions. A general AI assistant will do the drafting, and many job boards and applicant tracking systems now include built-in writing helpers. Specialist tools may add features such as bias checking or templates for particular sectors, usually on a subscription. For an occasional hire, a general assistant plus careful editing is normally enough; if you recruit often, a system that stores approved templates can save repeated effort.

    Keep your finished descriptions in one folder. Next time you hire for a similar role, give the AI your previous approved version and the changes, rather than starting from scratch. Over time, your AI job descriptions will build into a consistent, recognisable set that reflects how your business actually works.

    A worked example: a bakery in Milton Keynes

    Imagine a family bakery in Milton Keynes recruiting a part-time counter assistant. The owner’s first attempt says the business wants a “dynamic, customer-focused team player with at least two years’ retail experience and excellent communication skills”.

    She lists the real facts instead: early weekend shifts, serving customers, packing online orders, keeping the counter clean and following food hygiene rules. Experience is helpful, but she is happy to train someone reliable. She adds the hours, the pay, the location near the market and the free staff lunch.

    The AI draft that comes back is clear and specific, but it adds “must have a food hygiene certificate”, which she had not said. She changes it to “we will provide food hygiene training” and drops the two-year experience requirement, which would have excluded capable school leavers and people returning to work. The final advert is shorter, more honest and attracts applicants who know what the job involves. That is the right way to use AI job descriptions: the tool drafts, the owner decides.

    Fairness and the law

    Under the Equality Act 2010, adverts must not discriminate on protected characteristics such as age, sex, race, disability, religion or belief, sexual orientation and others. AI tools learn from existing text, and existing adverts contain plenty of biased habits, so check the wording carefully.

    • Avoid age-coded phrases such as “recent graduate”, “young team” or “digital native” unless there is a genuine, lawful reason.
    • Only list physical requirements that are truly part of the job, and mention that you will consider reasonable adjustments.
    • Question requirements for years of experience; describe the skills needed instead.
    • Ask for qualifications only where the role genuinely needs them.
    • Use gender-neutral job titles and wording.
    • Confirm the right to work in the UK is checked for all successful candidates in the same way.

    If you use AI later in the process, for example to screen applications, tell candidates, keep a person responsible for decisions and handle their data in line with UK GDPR. Acas offers free, practical guidance on recruitment and discrimination, and a qualified employment adviser can help with anything unusual.

    Common mistakes to avoid

    • Publishing the first draft. It will sound polished and may still be wrong.
    • Letting the tool invent details. Remove any benefit, duty or requirement you did not provide.
    • Endless requirement lists. Long lists put off good applicants who do not tick every box.
    • Hiding the pay. Clear pay information helps candidates decide quickly.
    • Copying a big company’s tone. Your advantage as a small business is being direct and human.
    • Pasting in personal data. You do not need names or details of current staff to write a job advert.

    Checklist before you publish

    • Every fact in the advert is accurate and approved by you.
    • Essential and desirable requirements are clearly separated.
    • Pay, hours, location and contract type are stated.
    • No buzzwords, jargon or internal abbreviations.
    • Wording checked for bias and Equality Act issues.
    • Reasonable adjustments mentioned.
    • Clear instructions on how to apply and what happens next.
    • Read aloud once to check it sounds like your business.

    Key takeaways

    • Strong AI job descriptions depend on specific facts you supply.
    • Tell the tool to use only your information and to avoid clichés.
    • Edit every draft for accuracy, tone and fairness.
    • Cut requirements that are not genuinely needed.
    • Keep people in charge of hiring decisions and seek advice when unsure.

    AI job descriptions: frequently asked questions

    Is it acceptable to use AI job descriptions?

    Yes. Many employers use AI to draft adverts. What matters is that the final text is accurate, fair and approved by a person, because the business is responsible for what it publishes.

    Will candidates notice the advert was written by AI?

    They may notice generic wording, which is why editing matters. A description full of specific detail about the real job reads as genuine, however it was drafted.

    Can AI help me decide the right pay?

    It can suggest factors to consider, but it may not have reliable, current local pay data. Check the legal minimum wage rules on GOV.UK, look at comparable local adverts and use your own judgement.

    How long does it take?

    Once your facts are gathered, drafting and editing AI job descriptions usually takes well under an hour, much of it spent on the careful review that makes the advert work.

  • AI in Hiring: Smart Screening and the Bias Problem That Won’t Go Away

    AI in Hiring: Smart Screening and the Bias Problem That Won’t Go Away

    Quick answer: AI in hiring can speed up job adverts, screening and scheduling, but it can also repeat or amplify bias, so UK employers remain responsible for fair decisions under the Equality Act 2010 and UK GDPR. Small employers should use it to assist people, never to make final decisions alone.

    Recruiting is time-consuming. A single job advert can bring in hundreds of applications, many unsuitable, and small businesses rarely have an HR team to handle them. AI tools promise to help: writing job adverts, screening CVs, scheduling interviews, and even assessing video interviews. They can save hours. But hiring is also where AI’s bias problem has caused real harm, and where the law sets clear limits. This briefing covers both.

    AI in Hiring: Smart Screening and the Bias Problem That Won't Go Away

    Where AI helps in hiring

    Writing better job adverts

    AI assistants can draft clear job descriptions from your notes, suggest more inclusive wording, and remove jargon. Well-written adverts attract more suitable applicants. Check every advert for accuracy, including pay, hours, location, and right-to-work requirements.

    Sourcing candidates

    Recruitment platforms use AI to suggest candidates whose profiles match a role and to recommend jobs to jobseekers. This widens reach but can also narrow it if the system learns to favour certain backgrounds.

    Screening applications

    CV screening tools rank or filter applications against the job requirements. For high-volume roles, this saves significant time. The risk is that a candidate is rejected unfairly because of how their CV is formatted, a career break, or factors linked to protected characteristics.

    Scheduling and communication

    Chatbots and automation tools answer candidate questions, book interviews, send reminders, and keep applicants informed. This is low-risk and improves candidate experience — nobody likes being left without a reply.

    Assessments and interviews

    Some tools use games, tests, or analysis of recorded interviews. Claims that AI can judge personality or suitability from facial expressions or voice have been heavily criticised by researchers, and some vendors have dropped facial analysis. Be sceptical of any tool that can’t explain what it measures and why it’s relevant to the job.

    The bias problem

    AI learns from historical data. If past hiring favoured certain groups, the model can learn to do the same. A well-known example is a large technology company that reportedly abandoned an experimental recruiting tool after finding it downgraded CVs that mentioned women’s activities. Bias can also creep in through proxies — postcodes, school names, gaps in employment, or language patterns — even if protected characteristics are removed.

    Bias isn’t only unfair; it’s a legal risk and a business one, because you miss good candidates.

    What the law expects (UK)

    • Equality Act 2010: employers are liable for discrimination in recruitment, including indirect discrimination caused by a tool they use.
    • UK GDPR: candidates must be told how their data is used. There are restrictions on decisions based solely on automated processing that have significant effects, such as automatic rejection, and candidates have rights to information and human review. The ICO has published guidance on AI in recruitment and audited recruitment AI providers.
    • Reasonable adjustments: disabled candidates may need adjustments to AI assessments.

    In the EU, the AI Act classifies AI used for recruitment as high-risk, with obligations for providers and employers. In the US, New York City requires bias audits for automated employment decision tools. Rules are tightening internationally.

    How small employers can use AI fairly

    1. Use AI for admin first: adverts, scheduling, and candidate communication.
    2. Keep humans in decisions. Don’t let a tool reject candidates automatically without review.
    3. Ask vendors hard questions: How was the tool tested for bias? What does it measure? Can you explain a decision to a candidate? Is there an audit?
    4. Define the criteria yourself based on what the job actually needs, and score all candidates consistently.
    5. Tell candidates where AI is used and how to request a human review or adjustment.
    6. Monitor outcomes: check whether particular groups are dropping out at particular stages.
    7. Protect data: don’t paste CVs into consumer AI tools without appropriate terms, and delete data you no longer need.

    Candidates are using AI too

    Applicants increasingly use AI to write CVs and cover letters, so polished applications are no longer a strong signal. Practical tasks, structured interviews, and good references are becoming more important.

    AI in hiring: frequently asked questions

    Can I use ChatGPT to shortlist CVs?

    You can use AI to summarise applications against your criteria, but review each candidate yourself, avoid feeding it unnecessary personal data, and don’t let it make rejection decisions alone.

    Is AI video interview analysis reliable?

    Evidence is weak for tools that judge suitability from facial expressions or tone of voice. Structured interviews assessed by trained people remain more defensible.

    Do I need to tell candidates I’m using AI?

    Yes. Transparency is required under data protection law and builds trust.

    *This article is general information, not legal advice.*

    A step-by-step plan for using AI in hiring responsibly

    For a small employer, recruitment is often squeezed in around everything else. AI tools can take some of the load, but only if you set them up with care. This plan assumes you are hiring a handful of people a year, not running a large recruitment team.

    1. Define the role before you touch a tool. Write down the genuine requirements: the skills, experience and qualifications the job needs, and which of these are essential rather than nice to have.
    2. Use AI to draft, then edit. Ask an assistant to draft the advert and interview questions from your requirements, then check for jargon, unnecessary requirements and wording that might put some groups off.
    3. Decide your scoring criteria in advance. Create a simple scoring sheet linked to the essential requirements, so every candidate is judged against the same yardstick.
    4. Keep a human reviewing every application. AI summaries can save reading time, but a person should look at each application before anyone is rejected.
    5. Record your reasons. Note why each candidate progressed or did not. If a decision is ever questioned, you will be glad you did.
    6. Review the outcomes. After each round, look at who applied, who was shortlisted and who was hired. If a pattern looks odd, investigate.

    A worked example: a small accountancy practice

    Imagine a six-person accountancy practice in Leicester hiring a trainee bookkeeper. The partners are busy with year-end work and expect a large number of applications.

    They use an AI assistant to turn their notes into a clear advert, then remove a line asking for “a degree” because the role does not genuinely need one. They set out four essential criteria and a scoring sheet. When applications arrive, the assistant produces a short summary of each one against those criteria, but one partner reads every application in full before deciding who to invite to interview.

    Their privacy notice for candidates explains that AI tools help with summarising and scheduling, and that a person makes all decisions. At interview, every candidate is asked the same core questions. After the hire, the partners check whether their shortlist looked noticeably narrower than the applicant pool. This is a hypothetical example, but it shows how AI in hiring can save time without handing over judgement.

    Questions to ask recruitment software suppliers

    • What exactly does the tool do with a candidate’s application, and can we see how it reaches a score or ranking?
    • Has the tool been tested for bias, and can you share what that testing involved?
    • Can we switch off automatic rejections so a person always reviews applications?
    • Where is candidate data stored, how long is it kept, and will you sign a data processing agreement?
    • Is candidate data used to train your models?
    • How do we handle a candidate’s request for reasonable adjustments, or to access their data?

    Common mistakes small employers make

    • Pasting CVs into a free chatbot without considering where that personal data goes.
    • Letting a tool filter out applicants on factors like employment gaps, which can disadvantage people for reasons linked to protected characteristics.
    • Copying an AI-written advert without checking for requirements the job does not really need.
    • Failing to offer an alternative for candidates who need adjustments to an online assessment.
    • Not telling candidates that AI tools are involved in the process.

    Effort and cost considerations

    Many general AI assistants can help with drafting adverts and questions at little or no extra cost if you already subscribe to one. Dedicated recruitment platforms vary widely in price and features, so compare current plans directly. Whichever route you take, factor in the time needed to set up clear criteria, write a candidate privacy notice and review outcomes. That effort is what makes AI in hiring fair rather than simply fast.

    A simple fairness checklist

    Before you publish an advert or reject anyone, ask yourself: are the requirements genuinely needed for the job? Has a person looked at every application? Are all candidates assessed against the same criteria? Can you explain each decision in plain terms? Have you offered reasonable adjustments? If any answer is no, pause and fix it first.

    Handling candidate feedback and data requests

    Candidates may ask why they were not shortlisted, or request a copy of the personal data you hold about them. Prepare for both. Keep your scoring notes in a form you could share, and set a clear retention period for unsuccessful applications, deleting them once it has passed. If a tool produced a summary or score, make sure you can retrieve it, because it may form part of the candidate’s data.

    Key takeaways

    • AI in hiring works best as an assistant for drafting, summarising and scheduling.
    • Set clear, job-related criteria before using any tool.
    • Keep a person reviewing applications and making final decisions.
    • Be open with candidates about how AI is used and protect their data under UK GDPR.
    • Check outcomes after each round for signs of unfair patterns.
    • Responsible AI in hiring saves time without handing over responsibility, which stays with you as the employer.

    Acas offers free, practical advice on fair recruitment at acas.org.uk, which is worth reading alongside any plans for AI in hiring.