Sieve

How Sieve works

Sieve shortlists. Humans decide. Here is exactly how that works.

1. You upload CVs

Recruiters upload candidate CVs (PDF, DOCX, or TXT) into a job. Sieve extracts the text, parses it into structured fields (name, experience, skills, work history), and stores the original file and parsed profile.

2. AI evaluates against the role

Sieve uses Claude (an AI model from Anthropic) to score each candidate against the job description, must-haves, nice-to-haves, and your company's hiring profile (stage, industry, tech stack, values). Scoring is split across five weighted dimensions:

  • Technical fit (35%) — depth and relevance of skills to the role
  • Culture fit (20%) — alignment with your company values and working style
  • Leadership (15%) — ownership, team growth, and delivery
  • Domain match (15%) — industry or problem space experience
  • Experience level (15%) — seniority match for the role

3. You see the reasoning

Every candidate gets an overall score (0–100), a breakdown across the five dimensions, and plain-English strengths and gaps. You can re-evaluate any candidate at any time if the context changes.

4. A human makes the decision

Sieve never automatically rejects or progresses a candidate. Every stage change (Shortlisted, Interview, Offer, Rejected) is an explicit click by a human recruiter. AI scores are a starting point, not a verdict.

This is a deliberate design choice. It means Sieve is not a "solely automated decision" system under GDPR Article 22 and supports compliance with the EU AI Act human-oversight requirements for hiring tools.

5. Candidates keep their rights

Candidates are entitled to:

  • Know that AI was used in screening
  • Request a human review of any significant decision
  • Receive an explanation of how their CV was scored
  • Request access, correction, or deletion of their personal data

These requests go to the hiring company (the data controller). Sieve (the data processor) supports companies to fulfil them.

6. What Sieve does not do

  • Automatically reject, interview, or hire candidates
  • Train AI models on your candidate data
  • Score based on protected characteristics (race, gender, religion, age, disability)
  • Share your candidate data with other Sieve customers

Questions

Email hello@sievecv.com or read the full Privacy Policy.