Recruitment is a writing-heavy, admin-heavy business, which makes it a natural fit for AI. It's also a business where AI decisions about people carry legal and ethical risk. Here's the honest split.
The line that matters most
AI drafts and organises. It does not decide about people. Using AI to screen, rank or shortlist candidates introduces real risk — of discrimination, of bias baked into the system, and of decisions you cannot explain if challenged. Every judgement about a candidate stays with a human who can justify it. Everything below sits inside that rule.
This isn't excessive caution. Automated decision-making about individuals carries specific obligations under UK data protection law, and bias in AI screening is a well-documented problem. The commercial risk to a small agency of getting this wrong is significant.
Where AI genuinely earns its place
Job adverts that actually attract people
Most job adverts are duty lists. AI is good at turning them into something a candidate wants to respond to.
Prompt — job advert
Write a job advert for a [role] at a [type of company] in [location]. Lead with why someone would want this job rather than a list of duties. Here's the information: [paste brief]. Around 250 words, warm and human, British English. Include a clear application instruction. Avoid gendered or age-coded language, and flag anything in my brief that might unintentionally deter good candidates.
That final clause is genuinely useful — AI is reasonably good at spotting phrases like "young dynamic team" or "recent graduate" that narrow your pool unnecessarily and create legal exposure.
Rewriting client briefs into candidate-facing copy
Clients send briefs written for internal use. Translating them is daily work.
Prompt — brief to advert
Turn this internal client job brief into candidate-facing advert copy. Keep every factual detail accurate, remove internal jargon, and make it appealing without overselling. If anything in the brief is unclear or missing — salary, location, hybrid arrangements — list it as a question for the client rather than guessing: [paste brief, client name removed].
Candidate communication at volume
The part agencies are most criticised for — leaving candidates without a reply. AI removes the excuse.
Prompt — candidate messages
Write template messages for: acknowledging an application, inviting to a first conversation, a positive update when a client is interested, a rejection after CV review, a rejection after interview, and keeping a good candidate warm when nothing suitable is live. Respectful, human, never corporate boilerplate. Each under 100 words with clear placeholders.
The rejection templates matter more than agencies think. Candidates rejected well come back and refer others; candidates ghosted tell people.
Prompt — post-interview rejection
Write a rejection message for a candidate who interviewed well but wasn't selected. Be genuinely kind, give one piece of useful specific feedback where I supply it, don't be vague or evasive, and make clear we'd like to stay in touch. Under 120 words.
Business development
Often the weakest area in a small agency, because it competes with delivery for time.
Prompt — client outreach
Write a short outreach message to a hiring manager at a company in [sector]. Lead with something relevant to them rather than about us. No generic 'I hope this finds you well'. Under 100 words, ending with a low-commitment question. Confident, not salesy.
Prompt — client update
Write a weekly update email to a client on a live role: candidates submitted, stage each is at, market feedback on the brief, and anything I need from them. Professional, scannable, under 150 words: [paste your notes].
Admin and internal
Summarising notes, drafting terms of business explanations in plain English, writing internal procedures, and turning a placement into a case study for marketing. All low-risk, all time-consuming. See onboarding and training staff with AI.
What to keep well away from AI
- Screening, ranking or scoring candidates — the core risk
- CVs and candidate personal data in consumer tools — that's personal data, often including protected characteristics
- Writing candidate assessments — your professional judgement, in your words
- Anything that becomes an automated decision about an individual
- Client confidential information in unapproved tools
See is it a GDPR breach to put customer data into free AI — the same logic applies squarely to candidate data.
The bias question, honestly
AI systems learn patterns from historical data, and historical recruitment data reflects historical discrimination. A tool trained on who got hired before will tend to favour people like those who got hired before. This is not hypothetical and has caused well-publicised problems.
For a small agency the practical response is straightforward: don't use AI to evaluate people, do use it to communicate better and work faster, and be able to explain every decision you make about a candidate in human terms.
Where the actual advantage lies
Not in processing more candidates — in treating the ones you have better. The agencies that win repeat business are the ones that reply quickly, communicate honestly and reject people decently. All three are admin problems, and all three are exactly what AI fixes. That's a more durable advantage than speed of screening, and it carries none of the risk.
Running an agency and want this set up safely?
We help professional services firms use AI where it helps and avoid where it doesn't. AskColin gives small businesses low-cost monthly support with one-to-one help whenever you need it — so when you're stuck on a real job, there's someone to ask. Practical, jargon-free, built around how you actually work.
Get one-to-one AI support for your businessFrequently asked questions
Can recruitment agencies use AI to screen candidates?
No — screening, ranking or scoring candidates with AI introduces real risk of discrimination, embedded bias, and decisions you can't explain if challenged. Automated decision-making about individuals also carries specific obligations under UK data protection law. Keep every judgement with a human.
What can AI safely do in recruitment?
Write job adverts that attract rather than list duties, translate client briefs into candidate-facing copy, produce candidate communication templates including good rejections, support business development outreach and client updates, and handle internal admin and procedures.
Is it safe to put CVs into ChatGPT?
No. CVs are personal data and often include protected characteristics. Don't put candidate personal data into consumer AI tools. Work generically with role requirements rather than individual candidate information.
Why is AI biased in recruitment?
AI learns patterns from historical data, and historical recruitment data reflects historical discrimination — so a tool trained on who got hired before tends to favour people like them. This has caused well-publicised problems and is why AI shouldn't evaluate people.
Where's the real AI advantage for a recruitment agency?
Not processing more candidates — treating the ones you have better. Agencies win repeat business by replying quickly, communicating honestly and rejecting people decently. All three are admin problems AI fixes, with none of the risk of automated screening.