Skip to main content
When CV review is on, Flo reads each application’s CV against the Job’s ideal candidate. The result is words and quotes, not a number.

​
Words per requirement

Each must-have, nice-to-have and red flag gets one word, a one-sentence reason and a quote from the CV. Inconclusive means no evidence. It never sorts with Poor.

​
N of M met

The list shows a count instead of an overall verdict: “4 of 5” means four of five must-haves are Good or better. Hire sorts by that count, then by who met the higher-ranked must-have. There is no overall fit word and no score.

​
How a word is reached

  1. Hire cleans the CV text, strips hidden text and removes the name, email, phone and address before the model sees it.
  2. The model gives a verdict and quotes per line. It may not return a score or a recommendation.
  3. Hire drops any quote it can’t find in the CV. A line with no surviving quote has no evidence.
  4. Three readings vote on each line. A tie goes to the lower word.
  5. Hire maps the verdicts to words and puts the person’s details back into the reasons you see.
On a benchmark of 14 synthetic CVs, Flo’s words matched hand labels on 80% of lines, and on 91% of must-haves for Good or better versus not.

​
Integrity

Each review also has an integrity level, Low, Medium or High, with reasons such as hidden text or instructions planted in the CV. It never changes a fit word or hides anyone. A person can change it with a reason.

​
A person decides

No word, count, integrity level or Rule moves or closes an Application. There is no auto-reject setting. You progress or reject each applicant, and your reasons can refine the ideal candidate.
Real candidate CVs go to the model only once a provider agreement that forbids training is in place. Until then the judge runs on synthetic CVs.