Recruiting is famously repetitive. You post a job, applications flood in, someone has to read every CV, schedule calls with a fraction of the candidates, coordinate interview slots with three different managers, send rejections to everyone else, and then — once you’ve hired someone — run through the same onboarding checklist you’ve been using for three years.

A significant chunk of that work is mechanical. And AI agents are genuinely good at mechanical work that involves reading documents, sending messages, and coordinating schedules. Here’s what’s actually working in 2026, without the hype.

CV Screening: Where AI Agents Add Real Value

The most mature use case is initial CV triage. An AI agent connected to your applicant tracking system (Workable, Lever, Greenhouse, or even a shared inbox folder) can read incoming applications, extract structured information — skills, years of experience, location, relevant qualifications — and score candidates against a job brief.

The important caveat: you’re not letting the agent make decisions. You’re using it to sort 200 applications into “worth a human look” and “clearly doesn’t meet the criteria” piles. That’s a different job, and it’s a defensible one. The human still reviews everything in the top pile; the agent just means a recruiter isn’t manually reading applications from candidates who applied from the wrong country or with no relevant experience.

A basic version of this is surprisingly buildable without custom development. Tools like Make or n8n can connect to your ATS via API or email parsing, pass CV content through Claude or GPT-4o with a prompt that includes your job requirements, and tag or score applications automatically. The prompt engineering matters a lot here — you need to be explicit about what you’re looking for, what disqualifies a candidate, and what you want as output (a score, a summary, a pass/fail flag).

For anything more sophisticated — multi-stage scoring, integration with assessment results, panel feedback aggregation — purpose-built platforms like Workable AI, Ashby, or Paradox (for high-volume recruiting) handle this without requiring you to build an agent yourself.

One thing to think about carefully

Automated CV screening has genuine legal risk in the UK. The ICO has been clear that GDPR applies to automated decision-making about job applicants, and the Equality Act 2010 requires that hiring decisions don’t discriminate on protected characteristics. If your screening criteria are written badly — filtering out candidates from certain universities, penalising employment gaps that could reflect caring responsibilities — an AI agent will apply those biased criteria at scale.

The fix isn’t to avoid AI in screening; it’s to design your criteria carefully, test your prompts against diverse candidate sets before deploying, and treat the agent output as a first filter that humans then review, not a final decision. Keep records of your AI-assisted screening process — you may need to explain how you reached shortlisting decisions if challenged.

Interview Scheduling

This is probably the lowest-controversy use case, and it’s already nearly fully automated at many companies. An AI scheduling agent can coordinate interview slots without a human in the loop for the back-and-forth.

The pattern works like this: candidate confirms interest in proceeding → agent checks interviewer calendars via Google Calendar or Microsoft 365 API → agent proposes slots → candidate selects a time → calendar invites go to all parties. Confirmations, reminders, and reschedule handling all automated.

Tools like Calendly AI, Motion, and Reclaim handle this well for straightforward cases. For enterprise-scale recruiting with panel interviews, video conferencing links, assessment centre coordination, and candidates in different time zones, dedicated scheduling tools from Greenhouse or Lever handle the complexity better.

The genuine efficiency gain here is real. Scheduling a three-stage interview process for a senior hire across six interviewers used to take days of back-and-forth email. Automated scheduling handles it in minutes.

Candidate Communication

Recruiting has a reputation problem: candidates often hear nothing after applying, or receive generic rejection emails written in the tone of a form letter from 1995. An AI agent can do better without much effort.

Automated but personalised rejection messages — referencing the role, thanking the candidate specifically, and where appropriate noting a specific reason — are meaningfully better than generic rejections, and they protect employer brand. An agent can draft these based on the application and the rejection reason, with a human approving the batch before sending.

Status updates (“your application is progressing to the next stage,” “we’re still reviewing applications for this role”) can be sent automatically based on ATS stage changes. Candidates get timely communication; recruiters don’t have to manually send dozens of updates.

Onboarding Automation

Once a candidate accepts, there’s a different category of repetitive work: the onboarding checklist. IT equipment requests, payroll setup, right to work verification, system access provisioning, training module assignments, first-week schedule coordination.

Most of this is triggered by the same event — someone accepting an offer — and follows a predictable sequence. An AI agent connected to your HRIS (BambooHR, Personio, HiBob) and IT ticketing system (Jira Service Management, Freshservice) can trigger this sequence automatically when an offer is marked as accepted.

The agent creates an IT equipment request, sends the new hire a right-to-work document checklist, schedules their first-week meetings, assigns mandatory training modules in your LMS, and notifies the relevant managers. A human HR coordinator reviews the package before it goes out — but they’re reviewing rather than creating, which is much faster.

For companies hiring regularly, this kind of onboarding automation saves hours per new hire. At 50 hires a year, that’s a meaningful number.

What Doesn’t Work Well Yet

Unstructured candidate research — where an agent trawls LinkedIn and other sources to supplement application information — is legally murky and operationally inconsistent. You’re better off asking candidates for what you need rather than having an agent find it elsewhere.

Full autonomous hiring — where an agent makes shortlisting, interview, or offer decisions without human review — isn’t appropriate today, both legally and practically. The value of AI in recruiting is compressing the mechanical work; the judgement calls about cultural fit, growth trajectory, and team dynamics are still human decisions.

And AI agents are only as good as the processes they’re automating. If your onboarding checklist is incomplete, inconsistent, or varies by department, an agent will automate the inconsistency. Fixing the underlying process before automating it is usually the right call.