FAQ

How does Trovinto Work?

Trovinto was founded to achieve consistent, bias free interviews with expert candidates.

Before Trovinto, hiring managers and recruiters would have to use generic questions from static libraries, but today they can use Trovinto to get tailor made interview questions from job descriptions.

Trovinto is built on the science of Natural Language Understanding and Machine Learning. The artificial intelligence supporting Trovinto has read and understood large parts of the internet and learned what text means, not just what it represents. The artificial intelligence can grasp the meaning of a job description, it understands the role in detail and reasons about which questions to ask a candidate.

The same artificial intelligence is used to reason about the recommended answer for an interview question, and to compare that to the answer that a candidate gives.

Trovinto is built on Artificial Intelligence models which have mesmerised the world. We have partnered with Open AI who are leaders in the industry.

We want to empower our customers with expert knowledge that can help build trust with candidates and hiring managers.

How does scoring work?

Trovinto will help you score and evaluate candidates immediately, so you can select the most competent candidates for the role.

When you have created an interview script with Trovinto, you can share it as a questionnaire for candidates, either in direct communication or in your job ad. When candidates respond to the questionnaire you will immediately see an AI generated report on your dashboard

How we do unbiased and fair candidate evaluations

When a candidate responds to a questionnaire, the answers are compared with the example answer from the interview script. The answers are broken down into their semantic meaning and the abstract distance in meaning is measured in a high dimensional space, built up from a large language model that our partners with OpenAI have compiled from their artificial intelligence with human language understanding.

The answers are not evaluated by humans, and the candidate's background is not used in the evaluations. This approach drives our unbiased and fair evaluations which are able to produce immediate scores and rankings.

Synonyms, different phrasings, and different approaches to the same answer will all produce great results. The model understands the meaning of a phrase, and reasons about the meaning rather than the text used to capture the phrase. The artificial intelligence does not depend on matching keywords, which makes this solution a much better fit for candidates who express themselves differently but are equally competent.

The score is calibrated to indicate a 0-5 star rating, where 0 means the answer is off topic, a 3 is a competent answer and 5 is an expert answer.