AI Recruiting System
Where can AI take on repetitive recruiting work without removing human judgment from the decisions that need it?
An assistant runs first contact and screening. Rules hold the hard requirements, every recommendation keeps a decision trace, and a recruiter decides anything uncertain.
The short version.
Recruiters spend much of the day on first contact and the same screening questions, so the candidates who actually need judgment wait.
This system splits the work three ways. A model holds the conversation and pulls evidence out of it. Rules decide the hard requirements, the weights and the routing, so a requirement is never a model's opinion. A recruiter makes every call that's uncertain or consequential, and can override any rule.
Every recommendation keeps a decision trace: what the candidate said, which rule used it, and who decided. Pick a candidate below, watch the conversation run, then press Why? to read the trace.
Demo
Run the screening.
Runs in your browser on scripted, invented data. There are no live model calls, and every person, company and record here is fictional.
Pick a candidate
Role · Service Advisor, Northgate Auto GroupSystem
Conversation
Qualification
| Criterion | Source | Status |
|---|
System log
Decision trace
The model extracts and classifies evidence from the conversation. Rules apply the gates, the weights, the thresholds and the routing. People keep the consequential decisions. Every step keeps the evidence and the rule that produced it.
The weights and thresholds here are illustrative. In a real deployment they come from the employer's legitimate job requirements, are validated against past hiring outcomes where possible, and are tested for disparate impact before they influence any recruiting decision.
Build note.
The problem, the assumption, the system, and what comes next.
- The problem
- Recruiters spend much of their day on first contact and screening questions, so the candidates who need judgment wait.
- Riskiest assumption
- That candidates answer screening questions fully and honestly when an assistant asks them.
- What it does
- An assistant opens the conversation and asks the screening questions. Rules check each answer against the role's hard requirements, and the CRM and calendar are updated.
- Left to rules and people
- Hard requirements are rules, not model opinions. Work authorization is always asked, never inferred. Anything unclear goes to a recruiter, who can override any rule.
- What's next
- Run it with recruiters on a real opening, and measure where candidates drop out and which rules are stricter than they need to be.