AI in recruitment: where automation helps and where it is high-risk.
Recruiters can safely automate interview scheduling, candidate updates, CV formatting and compliance-document chasing. Using AI to screen, rank or reject candidates is different: the EU AI Act treats many recruitment uses as high-risk, and discrimination and GDPR rules apply. Take advice before automating any decision about a candidate.
What takes a recruiter’s day?
Recruiters are paid to match people and roles. A large share of the day goes elsewhere: arranging interviews across three diaries, telling candidates where they stand, reformatting CVs into the agency template, chasing references, right-to-work documents and permits, and updating the applicant tracking system after every call. Candidates who hear nothing for a week drift to another agency.
For agencies placing overseas candidates, add employment permit applications and the documents that go with them. Each step is a message to write, a document to request or a field to update, and none of it requires a recruiter’s judgement. That is the work to automate. The judgement itself, about who is right for a role, is the part that needs the most care, and is covered further down this page.
Which jobs are safe to automate?
The safer jobs are the ones that move information around without judging candidates:
- Interview scheduling. Find times across candidate, client and recruiter calendars, send invitations and reminders, handle reschedules. See appointment booking.
- Candidate updates. Tell every candidate what stage they are at, using the stages recorded in your ATS, so nobody is left waiting. See follow-up automation.
- CV formatting. Convert CVs to your template and remove details clients should not see, with a recruiter checking the result. See document processing.
- Compliance chasing. Request references, ID and right-to-work documents, and track what is outstanding before a start date.
- Notes and ATS updates. Turn call notes into structured ATS records for the recruiter to confirm. See data entry automation.
What could automation save recruitment?
Put your own numbers in.
The starting values are an illustrative example, not a benchmark or a neu figure. Use your own volume, the time each one takes today and the share you expect the system to handle without a person.
Time saved is only part of the case: response times, missed work and errors matter too. The ROI guide covers how to set a baseline, and the cost guide what the build and running costs include.
Why is AI screening different?
The EU AI Act lists AI systems used in recruitment and selection, for example to filter applications or evaluate candidates, among its high-risk uses. High-risk systems carry obligations for both the provider and the business using them, around risk management, data quality, human oversight, record-keeping and transparency. The obligations are being phased in, and guidance is still developing.
Separately from the AI Act, equality law prohibits discrimination on protected grounds, and a model trained on past hiring decisions can reproduce past bias without anyone noticing. GDPR also gives people rights around decisions based solely on automated processing. In practice, keep a recruiter making and recording every decision about a candidate, and design any assistance so it can be explained and checked. Our human-in-the-loop guide covers the design side.
Which systems are involved?
The applicant tracking system is the centre. Whether an agent can read stages and write notes depends on its API and your plan. Calendars are usually Microsoft 365 or Google Workspace. Client relationships sometimes live in a separate CRM such as HubSpot. Many candidates prefer WhatsApp or text for interview reminders.
What should an agency decide before building?
- Draw the line in writing. List the tasks the automation may do (schedule, remind, format, chase) and the ones it may not (score, rank, reject, shortlist). Review the list whenever the scope changes.
- Check your suppliers. If your ATS or a job board already uses AI to rank or filter candidates, you may already be using a high-risk system. Ask the supplier what it does and what documentation they provide.
- Update your privacy notice. Candidates should know which tools process their data and why.
- Keep the records. Who made each decision, when and on what basis. Good records protect candidates and the agency.
- Plan for questions. Decide who answers a candidate who asks how a decision was made.
None of this stops you saving time on the safer jobs. It means you know where the boundary is before a tool quietly moves it.
What should a recruitment agency automate first?
Start with interview scheduling and candidate status updates. Neither involves judging a candidate, both save recruiter time every day and both improve the candidate experience in a way you can measure: time to schedule, candidates chased, complaints about silence. Add CV formatting next, with a recruiter checking every output. Leave anything that assesses candidates until you have taken advice and can show how a person stays in control.
Common questions
Is using AI to write job adverts high-risk?
Drafting an advert is not the same as evaluating candidates, so it is generally lower risk. Check that adverts do not use language that discourages people on protected grounds, and have a recruiter edit every one. Where an advert is targeted at particular groups automatically, the position may differ, so take advice if you plan to do that.
Can we use AI to summarise CVs for recruiters?
Summaries that help a recruiter read faster, with the full CV still available, keep the decision with a person. The risk rises if summaries leave out or emphasise information in a way that steers decisions, or if recruiters start relying on them alone. Treat this as an assistance tool, test it for consistency and take advice before rolling it out widely.
Do candidates need to be told AI is involved?
Transparency is expected under GDPR, and the AI Act adds transparency obligations for certain uses. At minimum, your privacy notice should explain how candidate data is processed and by which tools. Telling candidates when they are messaging an automated assistant is good practice in any case. Confirm the exact wording with your advisor.
Bring us the task.
Tell us what happens today, which tools are involved and where it goes wrong. On a free 30-minute review, we’ll work out whether an agent, an automation or neither is the right fix, and what a sensible first release would be.
Start a project.

Book a call or send us a message. We read everything.
tash@neu.ie
Working with teams in Ireland and worldwide.
Free 30-minute AI review.