# The late-night application is a full candidate record before your first coffee.

You automate candidate screening by putting a conversational agent in front of the job application. It asks what your intake form already asks, one question at a time, types the answers into a structured candidate record, and posts that record to your ATS the moment it is complete, with no score or decision attached.

## The process this page is about

Not "AI for recruiting". One process, the one that decides whether a good candidate finishes applying or closes the tab: the candidate intake, the conversation that starts the moment someone finds your posting.

- **Who starts it**: Someone who just found your posting, usually at night or on a break, filling in a form between other things.
- **What it costs today**: A static application form with no follow-up logic, or an ATS knockout question that forces a blunt yes or no where the real answer is "it depends."
- **What it has to produce**: A typed candidate record: the role, what they do now, their relevant experience in their own words, and whether the must-have basics are there.
- **of recruiters say back-and-forth scheduling coordination is the most time consuming part of their day**: 35% (source: The Interview Guys)

## What it asks, and what it never asks

You name the fields your application already has. It does the asking, in whatever order the story comes out, and it keeps asking until the required ones are filled.

| Field | Type | Required | Why |
|---|---|---|---|
| full_name | text | yes | Identity on the candidate record |
| phone | phone | yes | How the recruiter calls back |
| email | email | yes | Where the follow-up lands |
| role_applied_for | text | yes | Which opening this is about |
| current_employer_or_status | text | yes | Context for the recruiter |
| relevant_experience_summary | text | yes | The pitch, in their own words |
| work_authorized_in_country | boolean | yes | The one legal question, asked plainly and once |
| referral_source | text | no | Where the posting reached them |
| earliest_start_date | text | no | Timing a recruiter will confirm, never promised here |

### And what it will not do

- It never asks about race, religion, national origin, disability, age, or marital and family status, and it never records one if a candidate volunteers it.
- It never uses a scheduling or availability question to find out about childcare, religious observance, or a medical reason for a gap. "What hours work for you" is about hours, nothing else.
- It never tells anyone they have been screened out, rejected, or passed over, and it never scores, ranks, or renders a hire decision. A recruiter reviews every record.
- It never asks for salary history, only what someone is looking for going forward.
- It never invents a field. If they did not say it, the record says "Not captured" instead of a guess.

## Try the run yourself

Six situations a People Team actually handles, each running its own agent with its own fields. Pick one and answer it the way a real candidate or employee would, vaguely and out of order, and watch it ask the follow-up instead of accepting the blank.

- [Applying for a role](https://joina.chat/l/lm-intake): Chat widget. The core candidate intake, in the corner of the careers page.
- [Just started, first paperwork](https://joina.chat/l/lm-i9): Conversation page. I-9 document details, with a hard stop to HR compliance on anything off.
- [Would I even qualify](https://joina.chat/l/lm-prescreen): Conversation page. Pre-screening against a specific opening, no score revealed.
- [Interviewed here before](https://joina.chat/l/lm-silver): Conversation page. Checking if a past candidate is still interested, warmly.
- [Handing in notice](https://joina.chat/l/lm-exit): Conversation page. The exit interview, routed to HR and only HR.
- [Open enrollment](https://joina.chat/l/lm-benefits): Conversation page. Benefits elections for an employee who already decided.

## Frequently asked questions

**How do I automate candidate screening and intake?**

Name the fields your application already has, point the agent at your careers page and postings, and put the link or widget on your site. The agent runs the conversation, produces a typed candidate record, and posts it to your ATS through a webhook, Zapier or Make. Setup is minutes, not an implementation project.

**Will it ever tell a candidate they were rejected?**

No. It never scores, ranks, or renders a hire decision, and it never tells anyone they have been screened out or passed over. It records what was said and says plainly that a recruiter reviews every record and follows up.

**Is it safe under equal employment law?**

It is instructed never to ask about or record race, religion, national origin, disability, age, or marital and family status, and never to use a scheduling question as a way of finding out about any of those. Work authorization is asked once, as a plain yes or no, never citizenship. Because every conversation is stored with the verbatim sentence each value came from, you can show exactly what was asked and answered.

**Does it work with Greenhouse, Lever or Workday?**

Through a webhook, yes. Gnosari posts a signed JSON payload the moment a candidate record is complete, and Zapier or Make routes it into your ATS or HRIS. There is no Gnosari app to install on either side.

**What about I-9s and onboarding paperwork?**

That is a separate agent with its own field list. It collects the document type, number, and expiration date a new hire describes, and if anything is expired, expiring soon, or inconsistent, it stops immediately and sends the case to HR compliance rather than guessing. Benefits enrollment and exit interviews are separate agents again, each with a limit stated up front.

## Five minutes from now your careers page could be typing up its own candidate records.

Build the agent, put it on your careers page, and watch the first candidate record arrive. Free to start, no card.

Full page: https://gnosari.com/for/recruiting-hr