Brief the AI with the role title, key responsibilities, and must-have requirements. It produces a first draft in under a minute - typically 70-80% complete. You still need to adjust the seniority calibration and check the language is genuinely inclusive, but you're editing rather than staring at a blank page.

Give the AI the job description and ask for competency-based questions, follow-up probes, and a scoring rubric. The same question set for every candidate at the same stage makes decisions more consistent and significantly easier to defend if challenged. Most hiring managers don't do this because it takes too long - AI removes that excuse.

For roles that attract high volume, paste each application into the AI and ask for a one-paragraph summary against your stated criteria. Use the summaries to help you decide who to longlist - but make the longlisting decision yourself. This is assisting human judgement, not replacing it.

Both are time-consuming to do well and easy to get wrong. Give the AI your offer terms or your reasoning for declining, and ask for a draft in your company tone. This handles the templated parts - you add the specific context. Consistent, well-written rejection emails protect your employer brand with people who might apply for future roles.

Any AI tool that scores, ranks, or filters candidates without human review creates exposure under the UK Equality Act 2010. If a protected group - based on age, gender, race, disability, or other characteristics - is systematically screened out, your business carries the legal liability regardless of whether a tool made the decision. The ICO is explicit: automated decisions about people require human oversight.

AI screening tools trained on past hiring data learn what your past hires looked like - including any patterns that reflected unconscious bias. Tools that assess 'culture fit', analyse communication style, or make inferences from names and locations are particularly prone to this. Before adopting any AI screening product, ask the vendor for their bias testing methodology and what datasets the model was trained on.

Several AI interview tools claim to assess personality, motivation, or cultural fit from video or audio analysis. These tools have poor validity evidence and significant documented bias against candidates with accents, disabilities, or neurodivergence. Using them exposes you to Equality Act claims and ICO scrutiny. The ICO has specifically flagged this category of tool as high risk.

AI should not be used to determine whether a candidate has the right to work in the UK. This is a legal obligation that requires a human to verify specific documents. The same applies to DBS checks, reference verification, and any other compliance step that carries a legal duty. AI can draft the request letters - not make the compliance determination.