Good tests are good and increase accuracy when you hire new staff. Nevertheless, extensive research shows that almost half of all recruitments fail for one reason or another. At least at the management level.
To further improve the test's forecasting ability, we have therefore also built in AI (Artificial Intelligence) on our technical platform. Now we are starting to see amazing results!
Together with one of our customers, we have evaluated over a hundred people who have been tested by us and then hired. After a couple of months, we then had human resources managers evaluate the candidates, was it a good recruitment or not?
It turned out that the candidates recommended by our AI model have an accuracy of 76,6%! This means in principle that almost 8 out of 10 people recommended by the system also considered our customer to be a good recruitment.
Many companies talk about AI and how it will change the recruitment process. So far, it is mostly talk and the conceptual confusion is great. Most companies that claim to use AI in their recruitment have really only introduced an algorithm, a static calculation model (if you pass level 1, you can advance to level 2, etc.), into their systems.
The test tools normally allow the recruiter to analyze the results himself to later determine if he is a good or bad candidate. This results in an imprecise decision that is partly based on the results and partly on the gut feeling.
Psychometric's AI model instead analyzes the candidate's answers & results, based on previous candidates' answers & results, and recommends whether the candidate should be hired or not to simplify the decision-making process. The difference is that when real machine learning is introduced, the model is trained to become increasingly accurate.
Our model is, at least in Scandinavia, one of the first to be tested against reality. But for the AI model to succeed, a few things are required:
Basically, there must be a good and stable test, with relevant standard data (who is the candidate compared to).
Committed customers who evaluate the people who are also employed. In our model, we ask customers to assess the employees after 3-6 months. On a 4-point scale, they are allowed to assess whether the candidate has been a good or bad recruitment.
The assessment itself takes into account not only the candidate's test profile, but also everything that cannot be measured in a test. It can be, for example, co-workers, customers, life situation, etc.
The more test results and subjective evaluations that are made, the more secure our dynamic AI model will also be. In this way, the model makes it possible to improve the accuracy of the individual company, in different industries or to be able to compare different cultures.
In all likelihood, this type of AI model will take recruitment to the next level. We own both the technical platform, our AI model and our tests. It gives us unique opportunities to be flexible and create unique solutions for you as a customer. Maybe we can help you too?
Sources:
Bullies and Bastards, 2014.
The war for Talent, 2001.
Psykometrika 2021, SmartTest 3.0.