Executive summary

Context

VDAB, the Flemish public employment service, worked together with Superlinear to update and streamline its orientation test by making use of machine learning.

Goal of the Project

Reduce the time it takes job seekers to complete the orientation test, improve the user experience, and simplify analysis of the responses to result in better career recommendations.

Solution

Orient 2.0, a digital and AI-supported tool that creates a personalized orientation test and uses machine learning to predict professional interests.

Case study

The challenge

VDAB is the Flemish public employment service. The institution is a partner for 4 million active citizens and 250,000 job seekers in Flanders today and is active on a wide range of levels. It assists job seekers who are looking for a rewarding job, but also provides training programs, offers recruitment and training support for employers, and collects and analyses data on the labor market.

Helping people figure out which jobs best match their interests and skills is a large part of this social challenge.

To do so, VDAB used to rely on a classic orientation test in which participants were required to complete an extensive survey of over 100 questions. Afterward, labor experts needed to manually create scored relationships between specific questions and professions. The process was time-consuming for both job seekers and VDAB staff.

VDAB saw great potential in machine learning to improve the test. The employment service committed to a ‘digital first’ approach to bring its orientation test in line with three of its strategic goals:

• Offer perspective to all active citizens when it comes to their career: develop a modern and user-friendly tool to help them discover which direction they want to take their professional life.

• Activation: inspire job seekers through flexible tools to consider new and different sectors or job opportunities.

• Data director: integrating machine learning and artificial intelligence in the orientation test and other digital tools suits the purpose of reinforcing VDAB as a trusted advisor and data director on the Flemish labor market.

"We were looking for a partner to work closely with our artificial intelligence team to create a new, improved orientation test. We felt that machine learning could make our orientation test shorter and could provide job seekers with more accurate recommendations. As it turns out, the COVID-19 pandemic and the subsequent economic crisis further increased the need for innovative software solutions that guide citizens to rewarding jobs. Superlinear was a good match for this challenge."

Michael De Blauwe, Project leader AI at VDAB

The briefing

Superlinear and VDAB’s AI team joined forces to reinvent the orientation tool by using the latest developments in artificial intelligence and machine learning.

The objectives of the new, improved orientation tool were clear:

• Improve the user interface of the orientation test and personalize it, automatically tailoring it to the individual taking the test.

• Shorten the time it takes to complete the test.

• Make it easier for VDAB to maintain the test by creating automatically generated relationships between specific questions and professions through machine learning algorithms.

• Improve the result of the test with better job suggestions, in line with the current labor market.

• Make it easier to take the test for citizens whose native language is not Dutch.

• Be user agnostic to lead to fair, unbiased results.

In short: VDAB wanted to shorten and simplify its orientation test, while at the same time improving the user experience and the quality of the results.

Superlinear worked together with VDAB’s AI, design and data teams to improve the effectiveness, UI and UX of VDAB’s orientation tool. This multidisciplinary approach led to the creation of a streamlined, digital orientation test: Orient 2.0.

The solution

By implementing Artificial Intelligence, Superlinear and VDAB managed to improve the orientation test on multiple levels. The time to complete the questionnaire was reduced from 45 to 10 minutes – a time saving of almost 80%. The average number of questions was reduced from 114 to 58, due to the extensive personalization of the test. Last but not least, the use of machine learning resulted in more accurate recommendations for the people taking the test.

The technology behind these improvements uses AI to its full potential. The machine learning model Superlinear and VDAB implemented in Orient 2.0 picks up on relationships between job seekers’ responses and professions, asks specific follow-up questions to complete its analysis of the person filling in the questionnaire, calculates the probability that person will like a certain job, and finally delivers data-backed job recommendations tailored to the individual as well as the current job market.

So how did we get this result?

The successful implementation of Orient 2.0 owes much to the close collaboration between VDAB and Superlinear. The AI teams of Superlinear and VDAB consulted on a near-daily basis during key phases of the development. Thanks to this teamwork, it took the renewed Orient only four months to go from conception to its beta launch. The COVID-19 pandemic and the lockdown did not influence this timeline, another testament to the speed and flexibility of both teams.

Both VDAB and Superlinear followed a strict ‘user and digital-first’ approach. The developers listened to direct feedback from the users and analyzed their behavior while completing the test. By doing so, the team was able to get a detailed and reliable idea of what VDAB’s target group is truly looking for. This allowed quick adjustments to the tool to be made and will also allow for continuous fine-tuning over the coming months and years.

Last but not least, the development followed a tried-and-true process of experiment, explore, execute.

In the first phase, VDAB and Superlinear experimented with different prototypes and ideas to test whether or not they were viable. Afterward, VDAB and Superlinear further explored and improved their AI algorithm, its user-friendliness, etc. This also meant testing the solution with actual users. For the initial beta test run of Orient, the developers collected the input of 3,000 test users. After successful user testing, the solution was ready to be launched and executed.

The streamlined Orient turned out to be a massive success among users. Less than two weeks after the launch, the test had already attracted 10,000 users.

"It is clear that there is a market demand in Flanders for performant orientation tests, but also for user-friendly, digital tools that create tangible added value for job seekers. Given the popularity of Orient 2.0 since its launch and the massive improvements we were able to make thanks to Superlinear and machine learning, I’m confident we’ll help thousands of active citizens get a better idea of where to take their career. Our close collaboration with Superlinear also means we’re in a good position to further improve our labour market tool in the coming years."

Sylvia Mathijs - Project Leader Orientation at VDAB

We were more than excited when VDAB approached us for this project. I speak for our entire team when I say that we’re proud to help improve the Flemish job market in any way we can. From day one, we saw how AI could improve the orientation test and understood the step forward VDAB wanted to take. Likewise, we felt that VDAB’s leadership and AI experts truly understand the importance of smart data use and are fully committed to digitizing their offering to make the Flemish job market future-proof.

Davio Larnout - CEO at Superlinear

The future

VDAB and Superlinear will continue to work together to seize new AI opportunities that can bring innovation to the labor market. Their collaboration has already resulted in tools such as the digital e-learning platform ‘LeerInUwKot’.

Both organisations share the ambition to blend different digital solutions in order to offer users a seamless experience. In the long term, VDAB envisions a labor market service that covers job orientation, training, and job search by integrated, AI-supported tools.

In addition, both VDAB and Superlinear attach great importance to ethical AI and embedding fairness in machine learning tools. The partners will closely monitor Orient 2.0 – and other co-developed job market tools – to ensure that they do not unwillingly replicate societal biases, like under- or overrepresenting specific professions.

In the short term, Superlinear will continue to be involved in the monitoring and improvement of VDAB’s Orient test.

A key part of AI integration is continuous improvement. The labour market is constantly evolving, so AI needs to keep up with market trends and user preferences. The initial success of Orient shows that we made a real impact on the people using it. Our challenge is to now build on this success and to further implement AI in the Flemish labour market and make VDAB a leader in digital, user-friendly labour solutions.

Davio Larnout - CEO at Superlinear