The seven types of recruitment AI
Products are easier to evaluate when you know which type you are looking at. Each type replaces a specific chunk of work.
| Type | What it replaces | Who buys it | Examples |
|---|---|---|---|
| AI sourcing agents | The sourcer's day: searching, ranking, finding contacts, writing and sending outreach | Startups, lean talent teams, agencies | Noon |
| AI talent search | Boolean search across profile databases; contact finding | Enterprise TA teams, agencies | SeekOut, hireEZ, LinkedIn Recruiter AI features |
| Screening and matching | Reading and ranking applications; knockout questions | High-volume employers, large ATS users | Eightfold, HireVue, ATS-native matching |
| Conversational AI | Answering applicants, collecting details, booking interviews | Retail, hospitality, healthcare, logistics | Paradox, Sense |
| Interview intelligence | Note-taking, debrief summaries, interviewer coaching | Any team with structured interviews | Metaview, BrightHire |
| Scheduling automation | Coordinating panels across calendars and time zones | Mid-market and enterprise | GoodTime, ATS schedulers |
| Writing and workflow assistants | Drafting job descriptions, emails, scorecards; summarising profiles | Everyone, usually inside the ATS | Ashby AI, Greenhouse AI, Workable AI |
AI sourcing agents
The newest and most autonomous type. You give the agent a role brief; it searches, ranks, verifies contact details, writes personalised messages and sends them by email and SMS, then returns the people who replied. The value is that the work happens whether or not anyone had time for it, which is why it suits teams without a dedicated sourcer. Judge an agent on reply rate, the relevance of the shortlist, and how transparent it is about why each person was chosen. Noon is the clearest example of the category.
AI talent search
Databases of hundreds of millions of profiles with AI-assisted search: paste a job description and get a ranked list, with filters LinkedIn lacks (code repositories, patents, clearances, diversity indicators). The human still runs the search and the outreach. Strong for specialist roles; expensive; requires a skilled user.
Screening and matching
Models score applications against the job and surface the best ones first, or ask structured knockout questions. This is the type most exposed to bias and regulation, because it affects who is rejected. Insist on audit results, explanations and a human decision point before rejection.
Conversational AI
Text and chat assistants that respond to applicants in seconds, ask qualifying questions and schedule interviews directly onto managers' calendars. In hourly hiring, where applicants take the first offer of an interview, this measurably raises hire rates.
Interview intelligence
Joins video calls, transcribes, writes notes structured to your scorecard and syncs them to the ATS. Also produces interviewer metrics (talk time, question consistency). Improves decision quality more than it saves time; needs candidate consent.
Scheduling automation
Less glamorous, large payoff: interviewer pools, load balancing, self-scheduling and automatic rescheduling for complex loops.
Writing and workflow assistants
Job-description drafts, outreach templates, profile summaries, candidate-facing updates. Useful, cheap, and increasingly free inside the tools you already have; rarely a reason to buy a separate product.
Questions on this chapter
Which type of recruitment AI should a small company start with?
An AI sourcing agent if the problem is finding candidates (it replaces hours no one has), or conversational/scheduling automation if the problem is responding to too many applicants. Screening models matter most at high volume.
Is LinkedIn Recruiter recruitment AI?
It includes AI features (search assistance, message drafting) on top of a profile database, so it belongs in the AI talent search type. The recruiter still does the searching and messaging.