Claude for HR and Recruitment: Screening, Drafting and Staying Compliant

HR and recruitment teams handle a constant flow of writing and judgement calls: job adverts, screening notes, offer letters, policy questions and the awkward grey areas that sit between them. Claude, the AI assistant built by Anthropic, can take a large share of that drafting and triage off your team's plate. The important caveat, and the reason this guide exists, is that hiring and people decisions in the UK are governed by employment law, equality law and data protection rules. Claude can support the work, but a human must own every decision that affects a person.

This post explains where Claude genuinely helps an HR or talent function, where it must stay firmly in an assistant role, and how to keep your use of it compliant. If you want the wider picture first, our overview of Claude AI agents for business sets out how these tools fit into day to day operations.

Where Claude helps an HR and recruitment team

The strongest use cases are the ones where Claude drafts, summarises or answers, and a person reviews and signs off. In practice that covers a lot of the working week.

• Drafting job adverts and descriptions. Claude can turn a rough brief into a clear, inclusive advert, suggest more neutral wording, and adapt the same role for different channels. You keep editorial control and check the claims are accurate.
• Screening support. Claude can summarise a CV against a defined set of role requirements, flag where evidence is missing, and produce consistent interview notes. It supports the shortlisting conversation; it does not make the cut.
• Interview preparation. It can generate structured, role specific question sets and scoring guidance so that every candidate is assessed against the same criteria, which is good practice for fairness and for defending a decision later.
• Policy and process questions. Pointed at your own staff handbook and policies, Claude can answer routine "how do I book parental leave" or "what is our notice period" questions, freeing HR from repetitive queries.
• Correspondence. Offer letters, rejection notes, reference requests and onboarding emails can all be drafted to a consistent tone and then reviewed before sending.

The common thread is that Claude reduces the time spent on first drafts and lookups, while the experienced person keeps responsibility for the outcome.

The compliance line you must not cross

This is the part HR leaders should read twice. UK employment and equality law treats hiring, promotion, discipline and dismissal as decisions that need a human, accountable owner. There are three risks to manage deliberately.

Automated decisions and human oversight

Do not let Claude reject candidates, rank them into a hire or no hire list, or make any decision that materially affects someone, on its own. UK GDPR gives people rights around solely automated decisions that have a significant effect on them, and recruitment is exactly that kind of context. The safe model is simple: Claude advises, a named person decides, and that decision is recorded. Keep the human in the loop and make sure they have genuine authority to disagree with the model.

Bias and the Equality Act

An AI model can reflect patterns in the text it is given, so it can carry bias if you let it. Never ask Claude to infer or weigh protected characteristics such as age, sex, race, disability, religion or pregnancy. Constrain screening to the documented requirements of the role, review outputs for skewed language, and use structured, criteria based scoring so decisions are explainable. Treating Claude as a drafting and consistency aid, rather than a judge of people, is what keeps you on the right side of the Equality Act 2010.

Data protection and candidate data

CVs and HR records are personal data, and often special category data. Be clear about what you put into the tool, where it is processed, and how long anything is retained. Use a business grade configuration with appropriate data handling rather than a personal account, minimise the data you share to what the task needs, and update your privacy notices and records of processing to reflect AI assisted screening. We cover the underlying controls in Is Claude safe for business, which is worth reading before any rollout that touches employee data.

A sensible way to start

You do not need to automate the whole function to get value. A measured sequence works best for HR teams, where the cost of a mistake is high.

1. Pick low risk drafting first. Job adverts, policy FAQ answers and internal templates carry little legal risk and build confidence quickly.
2. Write the guardrails down. Agree what Claude may and may not do, who reviews what, and which tasks are off limits. Put it in a short internal policy.
3. Ground it in your own documents. Connecting Claude to your handbook and approved templates makes answers accurate and on brand, rather than generic.
4. Add screening support carefully. Introduce CV summarisation and interview prep only once the review process and audit trail are clear.

If the same approach appeals for your numbers and controls work, our guide to Claude for finance teams follows the same assist, review, sign off pattern.

How SpotDev approaches this

SpotDev is a UK consultancy that specialises in Anthropic's Claude. We design and build the workflow around the tool, the part that turns a capable assistant into something your HR team can rely on: the prompts and templates, the connection to your own policies, the review steps, and the guardrails that keep a human accountable for every people decision. Our engineering is in house, we have delivered more than 300 technology projects, and a first rollout is typically live in two to three weeks.

When you are ready to scope something concrete, our Claude implementation packages are fixed price, so you know the cost before you commit and the scope does not creep.

Frequently asked questions

Can Claude reject candidates automatically?

No, and you should not configure it to. Under UK GDPR people have rights around solely automated decisions that significantly affect them, and recruitment decisions fall into that category. Claude can summarise CVs and draft notes, but a named person must make and record the shortlisting and hiring decisions.

Is it legal to use Claude for screening CVs in the UK?

Yes, provided you use it as a support tool with a human decision maker, you constrain screening to the documented role requirements, you avoid weighing protected characteristics, and you handle candidate data in line with UK GDPR. The risk comes from removing the human and accountable owner, not from the technology itself.

How do we stop AI introducing bias into hiring?

Keep Claude away from protected characteristics, base screening only on the explicit requirements of the role, use structured and consistent scoring so each candidate is assessed the same way, and have people review outputs for skewed language. Treating the model as a drafting and consistency aid rather than a judge of people is the practical safeguard.

What HR tasks is Claude best suited to?

Drafting job adverts and descriptions, summarising CVs against fixed criteria, preparing structured interview questions, answering routine policy questions from your own handbook, and drafting correspondence such as offer and rejection letters. These are tasks where Claude saves time and a person retains the final say.

Work with a Claude specialist

SpotDev designs, builds and deploys custom Claude agents and enterprise Claude rollouts for UK businesses, with fixed packages from £8,000 to £45,000 and a first rollout live in two to three weeks. Explore our Claude implementation packages or talk to one of our engineers.