You type a sentence. The agent filters candidates, builds the campaign and prepares the message — on the actual screens, not in a chat log. You keep every Send button. Three steps, about 90 seconds.
| Name | Role | Location | Stage | Last reply | Phone |
|---|
| Name | Source | Fit | Verdict |
|---|---|---|---|
| The pool is empty. | |||
| Recipient | Preview |
|---|
Most "AI in your ATS" features are a chatbot bolted onto a search box: you ask, it reads thousands of rows into a prompt, it summarizes, you still do the clicking. The ATSBrain agent works the other way around. It drives the interface: sets the filters on the Candidates page, opens the campaign it created, fills the pool while you watch, and prepares the bulk message with recipients and previews on the real screen. Only counts and ids ever enter the model's context — which keeps it fast, cheap and auditable.
Is this a real AI? This page is a scripted demo with sample data so you can see the interaction without an account. In the product, the agent runs on the model of your choice (OpenAI, Anthropic, or your own key) and calls the ATS's own tools.
Can it send without me? No. Sending is a human action by design; the agent prepares and hands over.
Does it see my whole database? It sees what you see: your permissions, your page, your selection. It reads one record at a time only when you ask it to review someone.