mcoler

URL is to a light demo showing minimal functionality for a user - the recruiter or hiring manager - who is using the tool to identify their top candidates for a first interview. Paid users report that the tool identifies top candidates immediately, and specifically, "Conversa interviewed the first five people Conversa recommended. I could have hired all five."

I'm the founder, looking for feedback to make our product even more useful!

We ran into the same thing with our new recruiting tool, discovered hundreds of fake applicants for the first remote engineering job we posted. We initially noticed that a large number of job applicants were using ChatGPT responses with our virtual recruiter, and showing drastically different scoring from non-fake candidates.

We started scheduling video chats with the candidates clearly using ChatGPT with our virtual recruiter, and they all got on the call and began reading scripts, and could not answer more unique phone screen questions like: What engineering principle is important to you?

It’s the most frustrating thing, taking up so much of our time and taking time away from strong engineers who deserve our time.

I heard a good podcast about this, https://darknetdiaries.com/episode/133/

We’re testing a solution for this to keep our own product operating fast and delivering amazing candidates.

We’re also keeping a list of red flags:

- LinkedIn profile was created in the last year - GitHub link has a generic identity, 0-2 followers, very little activity - Candidate will say yes to a phone screen immediately, literally right now - Or candidate will ask if the meeting is a technical screen - Candidate will go quiet if you ask them if they are available to fly to HQ to meet the team for their final interview - Job history includes large, non-tech enterprises where employment would be hard to confirm; think CVS, The Home Depot, Best Buy, etc - LinkedIn About Statement is written in the third person - LinkedIn profile uses generic language like “cutting edge technologies” - Profile photos may be avatars, or look like stock photos

None of these things individually can indicate a fake applicant, but taken all together with language used with our virtual recruiter we’ve gotten good at identifying the fakes. We’re still doing a lot of tests to prevent discrimination on our part.

If anyone wants to chat, my CoFounder and I have become obsessed with solving this.

At Almanac we're building a documentation platform filled w/ resources for all roles in tech: Growth, Sales, Product, etc. Everything we publish is sourced from contributors, people in the field who do the work everyday. Check out our Request for Knowledge board! We'd love to feature you as a contributor on our platform.