an AI to predict the resignation of employees

In Japan, business leaders are finding it increasingly difficult to recruit. To protect against resignations which can seriously disrupt a department, a university professor has developed a resignation prediction model, based on artificial intelligence.

In Japan, labor shortages affect all sectors. A resignation can turn into a tragedy because it disorganizes teams that are already overwhelmed. So to help managers identify dissatisfied employees, Professor Naruhiko Shiratori of the University of Tokyo has developed a program that allows them to anticipate withdrawals.

The idea is to build a computer model that will make it possible to identify employees at risk, that is to say those who could soon announce their departure. Sometimes team leaders or human resources don’t see resignations coming. The researcher therefore teamed up with a start-up and developed a dedicated artificial intelligence platform.

AI predictive power

So, when we talk about artificial intelligence, it has become a little vague. Today artificial intelligence is essentially based on LLM (L(arge Language Models), large language models. These are ultra-powerful calculators, which can integrate astronomical quantities of data, and which then carry out probability calculations at phenomenal speed.

As with ChatGPT, when you type the start of a sentence “the sky is…”, the calculator will scan the entire internet, all existing texts in French, whether on Wikipedia, articles, the entire franceinfo site, books, blogs, to find the adjective which has the greatest probability of following the words “the sky is…”. So he will find “blue” and then move on to the next word.

Strategic Human Resources Efforts

Predicting employee resignations is a bit of the same thing. Instead of looking for words, the calculator will look for data on employee behavior. He will look at millions of data from the company, from its sector of activity, and compare it with the profile of those who have resigned in the past. After how long these resigning employees have left since joining the company, at what time of year, at what salary level, what age, what gender, etc. The program will also analyze the answers given during job interviews.

The program controls thousands of parameters and can then scan the profiles of current employees to identify those who are most likely to announce their departure soon. Thus human resources, or the manager, can intervene upstream to try to treat this employee.

Less labor

It’s a lot of effort to keep an employee, but you have to understand Japan’s exceptional demographic situation. We often talk about it here, the country loses almost a million inhabitants per year. Every hour, there are 100 fewer Japanese! So there are tens of thousands of jobs that are not replaced each year.

There are fewer and fewer students, therefore less labor entering the job market. Groups try to retain their employees as much as possible, because they know that a departure will be very, very difficult to compensate.