PRODUCT NOTES / CVSIFT
CVSift candidate-review walkthrough
A closer look at the candidate-review prototype: what it demonstrates, where it needs human checking and how a pilot could be evaluated.
Prototype walkthrough. This is an internal product demonstration using fictional candidates. It is not a client case study, and no measured client outcomes are claimed.
The problem we are exploring.
Reviewing several CVs against the same role criteria involves repeated searching, note-taking and checking. Relevant evidence can be phrased differently or spread across a document. A useful review tool should help a person find and inspect that evidence without treating the presence of a phrase as proof of suitability.
CVSift explores that narrower task: organise the evidence, make it available for checking and keep the reviewer responsible for their judgement.
What the demonstration shows.
Define the role and criteria
Start with the role and check the criteria that will guide the review. A clear criterion gives the reviewer something specific to look for.
Compare phrase coverage
Compare the demonstration CVs to see where relevant phrases appear. Coverage is a navigation aid, not a validated measure of competence or a hiring recommendation.
Inspect the evidence
Open each candidate’s evidence, check it against the source CV and record your own findings. Missing matches need manual review: different wording can describe relevant experience.
Export the review
Export matches or a shortlist that the reviewer has checked. The demonstration also includes an optional passphrase-encrypted workspace for continuing later; this feature alone does not establish that a production deployment is secure or compliant.
What this prototype does not prove.
The recording does not establish hiring accuracy, fairness, time savings or a return on investment. Phrase matching can miss equivalent wording and can find a term without understanding its context. Reviewers must check both matches and missing evidence.
This website does not accept CV uploads. For an initial discussion, use fictional examples. Any future deployment processing real candidate data needs an agreed data flow, appropriate permissions, security testing, retention arrangements and assessment of the hiring use case.
Read the data and human-review FAQs or read the website privacy information.
How we would evaluate a pilot.
These are proposed measures for a future pilot, not results we have achieved. Agree a baseline, representative examples and acceptance criteria before testing.
| Question | What to compare |
|---|---|
| Does it reduce review effort? | Time to complete the same review manually and with the prototype, including source checking and corrections. |
| Does it surface useful evidence? | Relevant evidence found, irrelevant matches and relevant evidence missed, checked against a human-reviewed reference. |
| Can reviewers explain their findings? | Whether recorded findings can be traced back to the source CV and the agreed role criterion. |
| Where does it struggle? | Different document formats, equivalent wording, incomplete information and cases requiring manual handling. |
A publishable client case study would need a completed project, documented baseline and results, limitations, and permission to share. There are no verified client results on this page.
Discuss a CVSift pilot.
Tell us about your review process, typical volume and what you want to learn. CVSift scope and pricing are agreed individually; the website’s custom automation prices are not a CVSift subscription offer.
Email hello@futuresai.org about CVSift or read about scope and costs.
Start with your workflow.
Tell us what repeats, which tools you use and where the process gets stuck. Use fictional or redacted examples.
Discuss your workflow