TalentPilot
Helps a staffing recruiter prepare for a screening call, run it, and hand it off to the account manager.
What it won't do
- It never scores, ranks or recommends a candidate. The recruiter decides, and can overrule any flag with a reason that's recorded.
- It never quotes a resume or transcript without checking. Code matches every quote word for word, and years of experience come from the resume's own dates.
- It never stores a candidate's details on a server. The work stays in the browser tab.
How it was tested
Four tests, each with its pass lines written before the run. Every test used fictional resumes and real public job postings.
-
Test 01
Oct 1Does it flag weak resume claims and leave solid ones alone?
A blind answer key, 10 public IT job postings and 10 fictional resumes. It flagged 34 of 36 weak claims and 0 of 24 solid ones, and all 83 quotes were real.
-
Test 02
Oct 2Is it better than a plain chat prompt?
Not on its answers. A plain prompt with the same rules did about as well on quotes, claims, years and questions. That moved the product's focus to the recruiter's workflow.
-
Test 03
Oct 3Do its "doesn't add up" flags fire on normal resume oddities?
Six deliberately messy fictional resumes. It flagged 1 of 18 innocent oddities, caught 12 of 12 real contradictions, and all 90 quotes were real.
-
Test 04
Oct 5Can a client's screening guide push it into unsafe questions?
Twelve runs, including a guide asking it to rank people and ask their age, and one with bias written in less obvious words. No question touched protected traits or ranking. Fresh Claude agents stood in for the live app, so one paid confirmation run is still to do.