Where recruitment AI is going

From features to agents

The first wave of recruitment AI added features to existing tools: a better search box, a drafted email. The current wave replaces whole workflows with agents: brief in, shortlist and replies out. The next step is agents that own a requisition end to end: sourcing, screening against structured criteria, scheduling and candidate communication, with a recruiter supervising several at once. The sourcing half of that already exists.

Candidate-side AI

Candidates now use AI to write applications, prepare for interviews and, increasingly, to apply at scale. Application volume is rising and signal per application is falling. Expect employers to respond by relying less on inbound applications and more on outbound sourcing and verified skills, and by moving assessment to live, structured conversations that are hard to automate from the candidate side.

The end of Boolean and of the seat

Semantic search has already made Boolean strings optional. Pricing is following the work: from per-recruiter seats towards per-team or per-outcome pricing, because an agent's output does not scale with the number of humans logged in.

Regulation will shape product design

The EU AI Act's requirements for logging, explanation and human oversight are becoming default features globally, because vendors build once. Expect every serious screening and ranking product to ship audit reports and explanation views as standard.

What will not change

• People take jobs from people. The conversation that closes a candidate stays human.
• A clear brief beats a clever model.
• Candidates remember how they were treated, whoever (or whatever) sent the message.