AI Candidate Screening, Without Leaving Your Inbox

The Screening Workflow, Step by Step

1. 1. Let Applications Land Where They Already Land

   Most hiring outside of large ATS deployments never starts with a form. Candidates email a CV to a careers address, reply to a job post, or get forwarded by a colleague. An email-first screening workflow keeps that intact: you forward the thread (or point a dedicated inbox at it) and the pipeline starts there. Nothing for the candidate to fill in, nothing for you to migrate.

2. 2. Parse Every Attachment on Arrival

   Screening only scales when parsing is automatic. Each incoming message is read end to end — the body copy, the PDF or DOCX attachment, and the signature — and turned into structured fields: name, contact details, years of experience, roles, skills, education, and links. This is the step teams usually fake with a spreadsheet and manual copy-paste, and it is the step that quietly eats the most time.

3. 3. Score Against the Job Description, With Reasoning

   Ranking accuracy matters more than raw speed. A useful AI screening pass compares the parsed profile against your actual JD — must-haves, nice-to-haves, seniority, location, and notice period — and returns a score with the reasoning attached. You should always be able to see why someone ranked where they did, and overrule it. A score you cannot audit is worse than no score.

4. 4. Keep One Ranked, Current Shortlist

   Screening efficiency collapses when the shortlist lives in three places. Keep a single ranked view that updates as new applications arrive, supports filtering by role and stage, and exports cleanly to a spreadsheet when someone outside the loop asks for it. The export is a report, not the system of record.

5. 5. Reply From the Same Thread

   The last mile of screening is communication. Because the whole pipeline started in email, rejections, scheduling, and follow-up questions go back out on the original thread — with the candidate's history in context. No copy-pasting between an ATS and a mail client, and no candidate wondering which address to reply to.

What to Measure

• Time to shortlist — hours between the first application and a ranked list you would act on.
• Coverage — the share of applicants actually read. Manual triage quietly drops the tail; automated parsing does not.
• Override rate — how often you disagree with the ranking. A steady, low rate means the JD criteria are tuned.
• Response time — how long candidates wait for a reply, which is the part they judge you on.

Frequently Asked Questions

Is AI candidate screening accurate enough to trust?

Treat it as ranking, not deciding. Scores with visible reasoning let you skim the top of the list quickly and spot-check the middle. The accuracy gain comes from consistency — every CV is evaluated against the same JD criteria, which is rarely true of manual triage across a busy week.

How is this different from a full ATS?

An ATS assumes candidates enter through your careers portal and that your team lives inside a new tool. An email-first AI powered hiring platform assumes candidates already email you, and that your team lives in the inbox. It is a lighter setup for small teams and agencies hiring a handful of roles at a time.

Do I need automation tooling like n8n or Zapier?

No. Wiring inbox parsing through automation flows means maintaining the flow, the credentials, and the token costs. Forwarding should be the whole setup.

What about candidates who apply through a job board?

Job boards email you the application. That email is the input, exactly like a direct application, so board applicants flow into the same ranked shortlist.