A typical recruitment consultant's week is split between work clients value (conversations, judgement, closing) and work nobody values: rewriting job adverts, typing up call notes, reformatting CVs, chasing timesheets. AI is useful for the second list. It is also where agencies get into trouble with data protection and discrimination law.
This guide covers where AI earns its keep in a UK agency, what the regulators have said, and how to roll it out without creating a problem you then have to explain to a candidate or the ICO.
How many UK recruitment agencies are using AI already?
The Recruitment & Employment Confederation published those figures in its UK recruitment industry status report 2023/24 in November 2024. The same report puts the sector's contribution to the UK economy at £44.4 billion in 2023, so this is a large industry, and it is taking AI on unevenly.
For a five or ten person agency, the headline figure matters less than the split underneath it. The firms getting value have picked narrow, repetitive tasks. Nobody is handing over the whole desk.
Which recruitment tasks is AI actually good at?
A realistic list for a small agency:
- Job adverts and LinkedIn posts. Paste the client brief, salary band and location. Get three advert variants back and edit the best one.
- Candidate summaries. Turn a CV and your screening call notes into a one-page profile in the client's preferred format.
- Call and interview notes. Transcribe and summarise a registration call, then pull out notice period, salary expectation and right-to-work status for the CRM.
- Boolean strings. Describe the role in plain English and ask for a search string for LinkedIn Recruiter or a job board, then test and tighten it.
- Admin and credit control. Timesheet reminders and invoice follow-ups, which matter in a sector where the REC's 2024/25 status report found 35% of member firms experienced bad debt in the past year. The approach in this guide to using AI to chase late payments applies directly to agency invoicing.
Where should a small agency start?
Pick one consultant and one week. Every advert they write goes through an AI draft first, then gets edited. Track two things: minutes per advert before and after, and whether response quality changes. If the drafts need heavy rewriting, the problem is usually the prompt. Give the tool a house style, three of your best past adverts and a list of banned phrases ("rockstar", "fast-paced environment").
Only once that works should you move on to tasks involving candidate data.
What did the ICO find when it audited AI recruitment tools?
The ICO's announcement of the audit outcomes is worth reading in full before you buy anything. Three findings stand out for agencies:
- Some tools inferred characteristics such as gender and ethnicity from a candidate's name instead of asking.
- Some let recruiters filter out candidates with certain protected characteristics.
- Some collected far more personal information than they needed and kept it indefinitely to build candidate databases, without candidates knowing.
The audited providers accepted or partially accepted the recommendations. The ICO also published key questions for organisations buying AI recruitment tools, which make a ready-made vendor checklist.
What does UK government guidance say about AI in hiring?
The DSIT guide is written for non-technical readers and warns that AI hiring tools can perpetuate existing biases, cause digital exclusion and enable discriminatory job advertising. Three points translate directly into agency practice:
- Pilot first. Run a new tool alongside your normal process on one vacancy type and compare the shortlists.
- Plan for adjustments. A candidate with a disability may be disadvantaged by an AI video interview or a timed online assessment. Offer an alternative route, and say so in the advert.
- Keep monitoring. A tool that looked fair in month one can drift. Check the makeup of AI-assisted shortlists against your applicant pool every quarter.
How do you stay compliant with UK GDPR when using AI?
A practical order of work:
- List every AI tool in use, including the free ones consultants signed up to on their own. That shadow use is usually the biggest gap.
- For each, record what candidate data goes in, where it is processed, and the retention period.
- Switch off model training on your data wherever the tool allows it. Business tiers usually offer this; free consumer tiers often do not.
- Complete a DPIA for anything that screens, scores or ranks candidates.
- Update the privacy notice in plain English.
For a fuller walkthrough, the AI data protection checklist for UK businesses covers DPIAs, processor agreements and retention in more depth.
Should you use general AI assistants or specialist recruitment software?
Before buying anything new, check what your existing ATS, CRM, Microsoft or Google tools already offer. Many agencies already pay for AI features they have never switched on. The REC also publishes guidance on using AI in the recruitment sector for members. If you are weighing custom workflows against off-the-shelf software, the trade-offs in this piece on whether to build or buy AI apply here too.
What should stay human in a recruitment agency?
The reason is partly legal and partly commercial. Legally, a person reviewing each rejection is your best protection against the discrimination problems the ICO found. Commercially, clients pay agency fees for judgement and relationships. If a candidate feels they were screened out by software, or a client senses a templated shortlist, the fee is harder to defend.
A useful rule: AI can write anything a consultant will read before it leaves the building. It should not send anything to a candidate or client on its own, and it should never make the final call on a person.
How do you measure whether AI is working?
Pick three tasks, time them for a fortnight, introduce AI, then time them again. If a consultant saves 20 minutes on each of ten summaries a week, that is over three hours returned to billable work. Check the other side too: interview-to-offer ratios, client complaints about shortlist quality, and candidate feedback. If time falls but quality slips, fix the prompts or pull the tool back from that task.
For help setting up Claude properly across a small team, with shared prompts, data settings and a sensible rollout, the Claude hub collects guides and setup options.
AI for UK Recruitment Agencies — FAQ
How many UK recruitment agencies already use AI?
The Recruitment & Employment Confederation's UK recruitment industry status report 2023/24, published in November 2024, found that 41% of recruitment businesses already use AI for recruitment. Another 30% said they intended to adopt AI within the next year, and 26% said they did not plan to use it. So most agencies are either using it or planning to. It is not universal yet, and a meaningful minority are deliberately staying out. For a small agency, the practical question is less about keeping up with the sector and more about which tasks in your own week are repetitive enough to hand over, and which ones clients pay you to do personally.
Is it legal to use AI to screen CVs in the UK?
Yes, but UK GDPR and the Equality Act 2010 still apply in full. The ICO's November 2024 audit of AI recruitment tools found some systems inferring gender and ethnicity from candidate names and letting recruiters filter out people with protected characteristics. Both are serious problems. If you use AI to screen, you need a lawful basis, a data protection impact assessment, clear privacy information for candidates, and a human who reviews decisions rather than rubber-stamping a score. Where a decision about a candidate is made solely by automated means and has a significant effect on them, extra rules apply. The safest pattern for a small agency is AI that ranks or summarises, with a consultant making every rejection.
What is the best first AI task for a small recruitment agency?
Job adverts and client-facing writing. It is the lowest-risk starting point, because no candidate personal data needs to go into the tool, and the time saving is easy to measure. You feed in the client brief, salary band, location and must-have skills, and get a first draft advert, a LinkedIn post and a shortlist summary template. A consultant then edits for accuracy and tone. Once that saves real time each week, move on to meeting notes and candidate summaries, where you will need to think carefully about data protection, lawful basis and what your privacy notice tells candidates.
Do candidates have to be told that AI is being used?
Under UK GDPR, candidates have a right to be informed about how their personal data is processed, which includes processing by AI tools. The ICO's recommendations after its recruitment audit called for transparent privacy information, including a clear retention period. In practice, update your candidate privacy notice to name the categories of tool you use, what they do with CVs and call notes, how long data is kept, and how a candidate can ask for a human review. Keep the wording plain. A candidate who finds out later that an unnamed system ranked their CV is far more likely to complain than one told upfront.
Should a recruitment agency build its own AI tools or buy them?
Most small agencies should start by switching on the AI features already inside their existing tools, such as the ATS, CRM, Microsoft 365 or Google Workspace. Many firms already pay for features they have never used. Buying a specialist screening product makes sense only when you have volume problems and can get proper assurances from the vendor on bias testing and data retention. Building custom workflows, for example connecting a general model like Claude to your CRM, suits agencies with a repeatable process that off-the-shelf tools handle badly. Whichever route you choose, pilot it on one desk before rolling it out.
