What recruitment AI is (and is not)

Recruitment AI is software that performs a recruiting task that used to need a person's judgement: finding candidates, writing to them, reading applications, answering applicants, scheduling interviews, taking notes, or predicting who will succeed. The word covers everything from a chatbot that answers "is this role remote?" to an autonomous agent that sources and contacts a shortlist overnight.

It is useful to separate three things that get lumped together.

• Automation follows rules a human wrote: "if the applicant lacks a driving licence, reject". It has been in applicant tracking systems for twenty years and involves no learning.
• Machine learning finds patterns in past data: which résumés led to hires, which outreach got replies. It predicts; it does not understand.
• Generative and agentic AI (large language models) read and write natural language and can plan multi-step work. This is what changed between 2022 and today, and it is why the term "recruitment AI" suddenly means something different.

The four jobs recruitment AI actually does

1. Finding and contacting people (sourcing). AI searches profile databases and the open web, ranks people against a brief, finds contact details and writes outreach. The most advanced version is an AI sourcing agent that does the whole loop and returns replies.
2. Reading and sorting applications (screening and matching). Models score how well an application matches a job, or ask knockout questions by chat.
3. Talking to candidates (conversational AI). Chatbots and SMS assistants answer questions, collect information and book interviews, mostly in high-volume hourly hiring.
4. Supporting interviews and decisions (interview intelligence). AI transcribes interviews, writes notes against a scorecard, and flags inconsistent questioning.

What it is not

It is not a robot that hires. Almost no organisation lets software make a final hiring decision, and in several jurisdictions doing so without human review is unlawful. It is also not, usually, a single product. Most companies run several narrow AI features inside tools they already own, plus one or two dedicated AI products for sourcing or screening.