This chapter first outlines the services that the Greek Public Employment Service (DYPA) currently offers to vulnerable jobseekers and identifies the main challenges it faces in doing so. It then reviews how public employment services across OECD countries use digital tools and data analytics to strengthen support for vulnerable jobseekers and explores how these approaches could be leveraged by DYPA. The chapter maps the core functionalities of these tools – from outreach and profiling through to referrals, skills assessment, job matching and employer services – and illustrates them with international examples. Finally, it contrasts this landscape with DYPA’s evolving digital infrastructure, highlighting both recent progress and remaining gaps, particularly in the identification of vulnerable clients and the targeting of support to them.
Strengthening Individualised Support for Jobseekers Furthest from the Labour Market in Greece
Abstract
Public employment services across the OECD are increasingly harnessing the potential of digitalisation. Digital applications now span nearly all areas of PES operations, including profiling tools, job-matching and career recommendation systems, skills-mapping tools, online career services and chatbots that facilitate information sharing and counselling.
This chapter first outlines the services currently offered to people furthest from the labour market, describes the digital tools that the Greek Public Employment Service (DYPA) already employs for all its clients, and identifies the main challenges DYPA faces in serving vulnerable clients effectively.1 It then provides an overview of modern digital tools used by public employment services in OECD countries, with a particular focus on tools using data analytics and their added value in four areas of special relevance for vulnerable clients: outreach, needs assessment, referral to appropriate services, and job matching and employer services. For each area, the chapter considers the applicability of these tools to DYPA. Given the potential benefits of a modern statistical profiling tool for identifying vulnerable jobseekers, the chapter places particular emphasis on the review of profiling tools used by PES across OECD countries, including the risks associated with their introduction.
This section examines how DYPA supports the integration of vulnerable groups into the Greek labour market. Section 2.2.1 discusses the definition of vulnerable groups used by DYPA and adopted also in this note and provides figures on the size and composition of vulnerable groups registered with DYPA. Section 2.2.2 briefly outlines the initiatives and programmes DYPA has implemented for these groups. Section 2.2.3 gives a concise overview of the tools used by DYPA, with a particular focus on those relevant to vulnerable clients. Finally, Section 2.2.4 explores the challenges DYPA faces in providing services to vulnerable groups.
2.2.1. Vulnerable clients in Greece face diverse challenges
The definition of vulnerable social groups adopted by DYPA is outlined in paragraph 8 of Article 2 of Law 4430/2016. According to this law, vulnerable social groups are those whose inclusion in social and economic life is hindered by physical or mental causes, or due to delinquent behaviour. This broad definition encompasses people with disabilities (both physical and mental), individuals with addiction problems, young offenders, prisoners, and former prisoners. The law further specifies nine special groups that are considered to be particularly disadvantaged in terms of access to the labour market: victims of gender and domestic violence, victims of illegal trafficking and human trafficking, the homeless, people living in poverty, economic migrants, refugees and asylum seekers with pending asylum requests, single parents, people belonging to marginalised cultural groups (e.g. Roma), the long-term unemployed under the age of 25 or over the age of 50, transgender individuals, and individuals who reside in Child Protection and Care Units after reaching adulthood. In this document, the term “vulnerable groups” includes both vulnerable social groups and special groups.
Jobseekers belonging to any of these vulnerable groups can register with DYPA in several ways. First, they may register directly with DYPA. First-time registration can be completed online. In a few regions, DYPA has dedicated local offices (called KPA2 offices) for vulnerable groups, where clients can register in person with the support of a DYPA employee. During the registration process, jobseekers are asked whether they belong to a vulnerable group. If so, additional information is collected (e.g. a valid disability certificate). Second, they can be referred to DYPA by other public bodies with responsibilities for vulnerable groups, such as the Ministry of Labour and Social Security (YEKA), the Ministry of Social Cohesion and Family, the General Secretariat for Equality, Human Rights, and the General Secretariat for Anti-Crime Policy. Third, they may be referred to DYPA by other governmental authorities, non-governmental organisations (NGOs) and other civil society organisations that represent the interests of these groups. For example, the Ministry of Labour and Social Security, in co‑operation with the European Social Fund Co‑ordination and Monitoring Authority has established a dense network of Community Centres across Greek municipalities, providing comprehensive support on social, employment, and family-related issues and often representing the first point of contact for vulnerable clients. While some Community Centres offer direct employment counselling through their own job coaches, in many cases, clients are referred to DYPA for more in-depth counselling and assistance (e.g. wage subsidies and other Active Labour Market Policies – ALMPs) in finding employment.
2.2.2. DYPA has invested in internal processes and programmes targeted to vulnerable groups
Three different departments and one directorate at DYPA oversee initiatives and programmes targeted at vulnerable groups. The Special and Vulnerable Groups Equality Department under the Directorate of Active Labour Market Policies (ALMPs) promotes integration and reintegration into the labour market. The Special and Vulnerable Groups Equality Department, within the Directorate of Programmes, Teaching Methods and Communication, provides specialised education and training. The KEK (Vocational Training Centres) Organisation and Administration Department, within the Vocational Training Directorate, manages schools for people with disabilities. Finally, the Directorate of Insurance and Social Benefits is responsible for implementing social policy actions.
To enhance the labour market integration of vulnerable clients, DYPA has launched a number of initiatives. While some of these initiatives are implemented as special ALMPs, others have become established practices within DYPA. As part of its established practices, DYPA offers two training programmes for counsellors focussed on vulnerable groups. Both programmes involve 35 hours of training: one programme is aimed at employer counsellors and focusses on raising awareness among employers and promoting the hiring of vulnerable jobseekers, while the other is aimed at jobseeker counsellors and provides them with a toolkit for supporting vulnerable clients. In 2023, 45 employer counsellors and 93 jobseeker counsellors participated in these programmes, working across 73 of the 120 local offices (KPA2) throughout Greece. These training programmes contribute to building counsellor capacity to support vulnerable groups effectively. Indeed, intensive, regular, and targeted counselling is crucial for the labour market integration of vulnerable clients (OECD, 2021[1]). DYPA has recently launched a collaboration with the University of Athens to train 21 new specialised counsellors and to enhance group counselling practices for vulnerable groups. Furthermore, DYPA has been able to hire significantly more counsellors over the past years. On the background of decreased numbers of registered unemployed, the caseloads have been decreasing as well and DYPA is able to provide better support to its clients. However, in international comparison, the caseloads are still high (averaging 665 clients per counsellor in May 2024),2 and delivering targeted and personalised support remains a significant challenge.
To provide more targeted support to vulnerable groups, DYPA launched the new model office for Special Social Groups (EKO Offices) on 28 May 2024 (DYPA, 2024[2]). Currently, there are two EKO offices located in Athens and Thessaloniki. DYPA counsellors working in EKO offices are specialised in counselling vulnerable clients and the new model office features full accessibility, including the possibility for remote interpreting services (including sign language) in every counselling room. EKO offices are likely to improve DYPA’s capacity to reach out to vulnerable groups by making employment services more accessible to those in need. However, its geographical reach remains limited to the Athens and Thessaloniki areas. In addition to the EKO offices, DYPA operates two facilities for special groups: the Athens Vocational Training School for Persons with Disabilities (PWDs) in Galatsi and the Thessaloniki Vocational Education and Training Centre for PWDs. These facilities focus on vocational training in various trades (e.g. carpentry, pottery, iconography) for unemployed persons with disabilities, with the aim of enhancing their employability across different sectors. Participants also receive personalised psychosocial support and counselling from specialised staff, including psychologists and social workers.
In addition to facilities tailored to vulnerable clients, DYPA has also launched a number of targeted ALMPs. These include employment incentives for employers to create jobs for vulnerable groups. One such programme is a wage subsidy scheme for 3 000 unemployed individuals from vulnerable groups, lasting 24 months, with the requirement that participating companies maintain the position for an additional 8 months after the programme ends (applicable only to private companies). Another wage subsidy programme supports 7 000 unemployed individuals facing various barriers to employment, such as mothers returning to the workforce after raising children (with at least one child under the age of 8 and at least 6 months of unemployment), individuals over the age of 55, the long-term unemployed, those without compulsory education, unemployed members of single‑parent families, and unemployed Roma. Another example is the initiative “Independent-Strong-Free” involving some 35 large companies and aiming at the labour market integration of female victims of gender-based or domestic violence.
2.2.3. DYPA has significantly upgraded its IT infrastructure and adopted new digital tools
Since the COVID‑19 pandemic, DYPA has significantly levelled up its digital infrastructure and expanded the availability of its digital services, with additional digital tools currently in development. Just a few weeks after the onset of the pandemic, DYPA launched a digital channel for applying for benefits. Shortly afterwards, it introduced myDYPAlive, a platform enabling both jobseekers and employers to book and manage remote counselling sessions. Although initially developed to ensure continuity of counselling (particularly for persons with disabilities and other vulnerable groups) during the disruptions caused by the COVID‑19 pandemic, the tool has remained in use since then (Alexiou and Koitsanou, 2021[3]). The tool has received high customer satisfaction ratings, both from jobseekers (87%) and employers (88%) and is a good example of how digital solutions have helped PES overcome the challenges of the pandemic. Additionally, during the pandemic, DYPA began offering training courses in collaboration with online providers such as Google, Coursera, and Microsoft (OECD, 2022[4]).
In 2021, DYPA launched the user-friendly website prosvasis.dypa.gov,3 specifically designed for persons with disabilities in collaboration with the University of Athens. The site complies with the Web Content Accessibility Guidelines (WCAG) 2.1 (Genova and Davern, 2022[5]), which offer recommendations for making web content more accessible to a wide range of disabilities, including blindness, low vision, deafness, limited mobility, and combinations of these (World Wide Web Consortium, 2023[6]). Moreover, since 2022, DYPA has integrated a digital assistant into its website. The chatbot, Daphne, uses natural language processing to deliver targeted information about DYPA’s services. Finally, DYPA is currently developing a new VET Management Information System, expected to be implemented in 2024, which aims to enhance the accessibility and quality of information available on various training programmes.
In addition to these new tools, DYPA has been improving its existing tools, such as its profiling and job-matching tools. Since November 2018, DYPA has used a rules-based profiling tool to categorise jobseekers into five groups based on a questionnaire containing approximately 30 questions covering various labour market-relevant factors, including work experience, language skills, and digital competencies (OECD, 2024[7]). A project between DYPA, the Directorate‑General for Structural Reform Support (DG REFORM) and the World Bank, aims to revise the current profiling model and strengthen its reliance on data. DYPA is also developing an advanced job-matching tool. Currently, jobs are posted on DYPA’s website, with limited filtering options for jobseekers. The new tool is expected to use AI to enable competency-based matching between jobseekers and vacancies. Currently, a version of the tool4 is available covering both temporary and permanent positions in the tourism and hospitality sectors only, with plans for further develop the tool and extend it to other sectors.
DYPA’s investments in IT infrastructure have benefited from the support of the European Commission’s Recovery and Resilience Facility (RRF) through the “Jobs Again” act, launched in April 2022 and aiming to strengthen DYPA’s digital infrastructure and improve the effectiveness and efficiency of ALMP provision (OECD, 2024[7]; Gazette of the Government of the Hellenic Republic, 2022[8]). DYPA’s digitalisation efforts have also benefited from synergies with other national-level initiatives. A crucial example is the Central Interoperability Centre (KED), developed and managed by the Ministry of Digital Governance, which allows DYPA to exchange data such as employment records, tax information and social security details with other public authorities for administrative purposes (OECD, 2024[7]). Beyond administrative use, DYPA has started leveraging these data within its new performance management system to monitor ALMP provision (M&E System), marking a step towards a more evidence‑based approach in the design and implementation of these policies. Moreover, DYPA’s data are fed into the Diagnostic Mechanism of the unit of Experts in Employment, Social Insurance, Welfare and Social Affairs (MEKY) of the Ministry of Labour and Social Security, providing valuable data for labour market analyses.
2.2.4. DYPA still faces multiple challenges in reaching out and providing ALMPs to vulnerable groups
DYPA faces several challenges in assisting vulnerable groups. The first challenge is outreach (OECD, 2021[1]) Vulnerable groups frequently fall under the radar as they are often less likely to proactively seek DYPA’s support. One reason for this is that they may not qualify for unemployment benefits and, therefore, lack a direct financial incentive to register. Additionally, reduced internet access and low language or digital skills may limit their ability to access information about the services available through DYPA. Certain vulnerable groups, such as former prisoners, may also have lower levels of trust in the state and its services, leading them to avoid contact with public authorities. Geographical factors further exacerbate this challenge: while it may be easier to reach vulnerable groups in urban centres like Athens, providing services in rural areas is more difficult due to limited availability of targeted support in these regions.
A second challenge lies in understanding the complex needs of vulnerable groups. The multiple barriers they encounter in accessing employment (e.g. health limitations, limited experience, low skills, and discrimination) often lead to interrelated issues that may extend beyond employment to other areas such as housing. Moreover, the current categorisation used by DYPA groups together individuals facing very different barriers to employment, resulting in diverse needs for employment services.
A third and related challenge is meeting the needs of vulnerable clients with appropriate services. On their path to employment, they often require co‑ordinated support that goes beyond the provision of employment services and may not always be directly offered by DYPA (e.g. psychosocial support). To ensure effective and co‑ordinated service delivery, strong collaboration between DYPA and other organisations, such as NGOs and governmental agencies working with vulnerable groups, is essential. Furthermore, the high caseloads handled by DYPA counsellors often limit their capacity to provide the intensive and tailored support vulnerable clients require.
A fourth challenge is to boost employers’ willingness to hire vulnerable jobseekers. Employers may be hesitant to hire jobseekers from vulnerable groups due to prejudice, perceived risks, or uncertainties about their skills. In this context, DYPA can play an important role in bridging the gap between the labour demand of employers and the labour supply of jobseekers from vulnerable groups. A key aspect of this work involves providing employers with the necessary information and matching jobseekers to suitable vacancies.
As part of a more holistic approach, involving process streamlining, strengthening outreach mechanisms and consolidating collaboration and data exchange with other public agencies and NGOs (e.g. through formal co‑operation protocols), digital tools can be instrumental in tackling these challenges and helping DYPA create effective pathways to employment for vulnerable clients. The next section provides an overview of tools used by PES across OECD countries to address challenges in these four key areas particularly relevant for vulnerable clients.
Digitalisation presents major opportunities for PES to deliver higher-quality services at reduced costs. It is, therefore, not surprising that PES across the OECD have been striving to harness these benefits by implementing digitalisation initiatives in nearly every aspect of their services and operations (OECD, 2022[4]). This shift towards a more digitalised PES has been greatly accelerated by the COVID‑19 pandemic, which forced PES to make significant changes to their operations and service delivery while facing a surge in the number of clients. In addition to the pandemic, advances in AI and data analytics,5 which are increasingly unlocking new ways to use jobseeker and labour market data, are also driving the shifts towards more digitalised PES (Brioscú et al., 2024[9]).
These digitalisation initiatives have resulted in a number of tools that provide different functionalities to support PES in their activities with their different stakeholders. Such functionalities include providing information to jobseekers and employers through enhanced websites and chatbots, conducting statistical profiling to better understand jobseeker needs and matching jobseekers to suitable vacancies. In some cases, these functionalities are strengthened by the use of AI (Brioscú et al., 2024[9]). For examples, AI is used in chatbots to provide more tailored information to clients and in statistical profiling tools to enhance the predictive power of the statistical model.
Vulnerable clients often face multiple barriers to employment (see Section 2.2.4), making it essential to provide them with individualised support tailored to their complex needs. While counselling vulnerable groups still requires face‑to-face interactions due to their often limited digital skills and/or internet access (OECD, 2021[1]), digital tools can support PES in delivering their services to the most vulnerable clients in four key areas. First, they can improve outreach to vulnerable groups, helping PES locate and engage vulnerable clients who are far from the labour market (OECD, 2021[1]). Second, digital tools can help better understand the needs of these clients, who possess diverse characteristics, face different barriers to employment and have different needs. Third, digital tools can assist PES counsellors in referring vulnerable clients to suitable services, such as training or more intensive counselling. Fourth, digital tools can enhance the matching of vulnerable jobseekers with vacancies and improve employer services.
This section provides an overview of the various functionalities of digital tools used by PES across OECD countries, focussing on those functionalities that support PES core business activities and use data analytics, strengthened, in some cases, by the use of AI. These functionalities are grouped into the four broader areas mentioned above: Improving outreach to vulnerable groups, better understand the needs of vulnerable clients, support the referral of clients to appropriate services and enhance job matching and employer services. The overview draws on data from a country-level questionnaire on PES digitalisation and AI use6 conducted by the OECD in spring 2023, alongside a review of the literature on the use of digital tools in PES to provide concrete examples.
2.3.1. Digital tools can support outreach to vulnerable groups
One of the most common challenges faced by PES is the difficulty in locating and reaching out to the inactive population. This is especially true for vulnerable groups, who are often not reached through standard communication channels and may lack the digital skills to access information online. Nevertheless, effective outreach plays a crucial role, as early intervention is vital for the success of activation policies (OECD, 2016[10]; OECD, 2023[11]).
Despite this, only a modest share of PES employ digital tools using data analytics to support outreach (see Figure 2.1). Roughly 50% of PES in OECD countries have tools specifically targeted to jobseekers and individuals at risk of job loss, aimed at providing information on available services and measures, as well as their eligibility. The delivery of this information often involves the use of chatbots and conversational bots, occasionally enhanced by AI. While these technologies make large volumes of information more accessible and easier to navigate for clients, they rely on the clients’ initiative to actively seek out that information.
In addition to sharing information, PES also use digital tools to proactively identify and reach out to potential clients (38%). However, this application remains limited and the extent of proactivity in client outreach varies greatly across PES.
PES: Public Employment Service, DYPA: Greek Public Employment Service.
Note: “Share of respondents” indicates the share of PES with at least one digital tool offering the specific functionality. “DYPA: implemented” indicates that the functionality is provided by at least one existing digital tool at DYPA. “DYPA: Implementation 2024 or later” indicates that the functionality is scheduled to be implemented in 2024 or later.
Source: OECD PES digitalisation and AI use questionnaire (2023).
Linking administrative data while ensuring compliance with the General Data Protection Regulation (GDPR) can be an effective method for proactively identifying vulnerable clients who might benefit from the services provided by the PES. If the availability of administrative data is limited or coverage is insufficient, survey data can be used as a supplement (OECD, 2021[1]). An example of this approach is the Estonian Youth Support Guarantee System, a tool developed by the Ministry of Social Affairs in 2016 in collaboration with local authorities and fully implemented in 2018 (OECD, 2021[12]; OECD, 2023[11]). The tool uses data from nine registers to help local authorities identify young people aged 16‑26 who are not in education, employment or training (NEETs) and support them in continuing their education or entering the labour market. Case managers in the municipalities can then reach out to these NEETs and, if necessary, refer them to the Estonian PES (OECD, 2021[12]).
DYPA already has, or is in the process of developing, digital tools that cover both functionalities. Information about available services is provided to both jobseekers and employers through DYPA’s mobile app and website,7 which, since 2022, has featured a digital assistant, Daphne, a chatbot that uses natural language processing to offer clients targeted information about DYPA’s services. Additionally, jobseekers and employers can book virtual appointments with DYPA counsellors via the myDYPAlive tool to receive further details on available services.
2.3.2. Data analytics can help better understand the needs of vulnerable jobseekers
PES across the OECD have adopted digital tools using data analytics and providing various functionalities to better understand the needs of their clients.
Jobseeker profiling tools are widely used across PES, with 78% of PES making use of them. These tools assess the jobseekers’ prospects of finding employment, allowing to segment jobseekers in function of their distance from the labour market and their needs. This enables resources to be focussed on those who require the most support, while reducing deadweight costs associated with providing services to jobseekers who would likely have found work independently (Desiere, Langenbucher and Struyven, 2019[13]). Overall, profiling tools help to target service provision more efficiently, particularly for those PES facing large numbers of jobseekers and high caseloads for counsellors.
To better understand needs for reskilling or upskilling, most PES (59%) have tools mapping the distance of jobseekers from occupations and highlighting gaps in competencies. While these tools are most commonly targeted directly at jobseekers, they are, in some cases, also used to support the work of counsellors. One example is Canada’s Job Transition Tool8 available on Job Bank, the website of Canada’s national employment service. The Job Transition tool uses data from the Occupational and Skills Information System (OaSIS)–a comprehensive framework detailing the skills, abilities, personal attributes, knowledge, and interests typically required in over 900 different Canadian occupations – and calculates the time needed to transition from one occupation to another, based on available education and training programmes to address gaps in competencies. In addition, the tool displays occupation-specific labour market information such as wages and number of vacancies (Canada’s national employment service, 2024[14]).
Slightly less common are tools that recommend suitable career paths. These tools can follow two approaches. One approach analyses data on skills employers expect and combines these with workers’ previous career choices to recommend specific career paths (45%). The other approach recommends career paths according to jobseekers’ interests (43%). An example of the latter is Oriënt 2.0, an orientation test developed by VDAB, the Flemish PES, to help jobseekers find occupations that match their interests. Oriënt 2.0 asks a series of simple Yes/No question designed to elicit the user’s interests and preferences. The tool then displays a list of occupations that align with the user’s interests and preferences. Additionally, the platform provides information on which occupations are in high or low demand based on vacancy data. (VDAB, n.d.[15]). While the previous version of Oriënt 2.0 asked over 100 questions and required manual assessment by labour market experts, the current version makes use of AI to adapt the questions based on clients’ previous responses, reducing the average number of questions from 114 to 58 questions. The introduction of AI has reduced the average time required to complete the test by nearly 80%, from 45 to 10 minutes (Radix, n.d.[16]; OECD, 2024[17]).
Finally, digital tools for testing jobseekers’ skills are used by only 40% of PES. While these tools do not generally involve data analytics systematically, some enable jobseekers to compare their skill profiles with others or recommend occupations that align with their skills profiles. For example, the “OECD Skills Profiling Tool” (Tuccio et al., 2023[18]) allows users to assess foundational (e.g. literacy, numeracy), occupation-specific and non-cognitive skills using academically validated tests. Users can then compare their skills profiles with those of other users and receive recommendations for occupations that match their skills. These tools not only help counsellors gain a deeper understanding of their clients’ needs, but also generate valuable data that can be used by other tools. For example, skills profile data can enhance competence‑based matching of jobseekers to vacancies (OECD, 2024[17]).
Except for tools recommending suitable career paths based on jobseekers’ interests and tools to test jobseekers’ skills, DYPA is already using, or planning to implement, digital tools that cover all other functionalities. DYPA is currently developing a new VET Management Information System scheduled to launch in 2024, which will offer a range of functionalities. One key feature will be the ability to make recommendations based on the skills employers are seeking and the previous career paths of workers. The tool should also allow to map the distance between jobseekers and specific occupations, identifying gaps in their competencies.
Additionally, since November 2018, DYPA has used a rules-based profiling tool to segment jobseekers into five categories (see Section 2.2.3). This tool assists DYPA counsellors, who can reassign jobseekers to a different category if necessary. The categorisation of jobseekers has direct implications for the services they receive. For example, the most job-ready candidates receive an automated individual action plan rather than attending counselling sessions with DYPA counsellors. Given the relevance of statistical profiling when it comes to identifying vulnerable clients, the following section provides a more detailed overview of the design of profiling tools used by PES across the OECD.
2.3.3. PES across the OECD use different types of profiling tools
The literature identifies three major types of profiling (Desiere, Langenbucher and Struyven, 2019[13]). Rules-based profiling employs eligibility criteria (e.g. age) to assess a jobseeker’s distance from the labour market. Caseworker-based profiling relies on the assessment of caseworkers in determining the jobseeker’s distance from the labour market. Finally, statistical profiling uses historical jobseeker data and a statistical model to predict the distance of jobseekers from the labour market. The lower this probability, the further the jobseeker is from the labour market. In practice, many PES use a combination of these methods, for example, applying rule‑based or statistical profiling while allowing caseworkers the discretion to adjust the results if necessary.
Compared to rules-based profiling, statistical profiling leverages information contained in historic jobseeker data: the underlying statistical model learns from past jobseeker data how specific characteristics of jobseekers (such as gender, age, disability) are linked to the likelihood of finding employment, or a related indicator, such as the probability of not re‑entering unemployment. It then uses this information to estimate the likelihood of new jobseekers finding a job based on their characteristics. This allows the statistical model to simultaneously consider the effect of different individual factors on the probability of finding employment and uncover patterns in the data that might not be directly recognisable by counsellors. However, statistical profiling also presents some challenges. For example, the accuracy of statistical profiling crucially depends on data quality. In addition, statistical models, particularly those involving the use of AI, tend to be less transparent than rules-based profiling, as it is not always evident how specific jobseeker characteristics are linked to individual outcomes.
Input data: Defining labour market-relevant jobseeker data
Figure 2.1 illustrates the typical steps involved in statistical profiling. This first building block, the input, comprises data that the profiling model uses to predict outcomes. PES across the OECD employ different types of data for profiling purposes. Typically, profiling tools use data on the socio-economic characteristics of jobseekers (e.g. age, gender, migration background), motivation to seek employment (e.g. job-search behaviour, reservation wage) and job readiness (e.g. education, skills, disabilities, care responsibilities). In some cases, these data are further enriched with information on regional labour market conditions, such as regional unemployment rates or other labour market indicators (Desiere, Langenbucher and Struyven, 2019[13]; Körtner and Bonoli, 2023[19]). While much of these data are held by the PES, additional data (e.g. detailed employment histories, health data) may need to be obtained by linking PES data with records from other institutions. The use of AI in profiling tools enables the incorporation of additional data sources that traditional models are less able to handle. For example, click data can serve as a proxy for job-search behaviour (Broecke, 2023[20]).
Source: Desiere, Langenbucher and Struyven (2019[13]), “Statistical profiling in public employment services: An international comparison”, https://doi.org/10.1787/b5e5f16e-en.
The profiling tool developed by VDAB, the PES of Flanders (Belgium), uses data on more than 400 jobseeker characteristics, including socio-economic factors (e.g. age, place of residence), labour market history, administrative data collected during current and previous unemployment spells and job search preferences (e.g. desired occupation, industry, location) (Broecke, 2023[20]; Ernst, Mueller and Spinnewijn, 2024[21]). In addition, the prototype of the profiling tool even incorporated indicators of job-search behaviour based on click data generated through jobseekers’ interactions with VDAB’s online services. However, due to the incomplete nature of these data, they were excluded from the final version of the profiling tool (Brioscú et al., 2024[9]).
In the Netherlands, the Dutch PES, UWV, employs the tool “Work Profiler”. The first version of this tool was developed between 2007 and 2010 and has been in use since 2011 in certain regions, and nationwide since 2015. During its development phase, the aim was to create a parsimonious model that could accurately predict jobseekers’ probability of finding employment using as few items as possible. Initially, 550 items were identified as potentially relevant to predicting this probability. In subsequent steps, this number was gradually reduced, resulting in a final set of 20 items used for profiling (Wijnhoven et al., 2023[22]). Except for one item (age), which is collected from administrative data, all other items (such as education, experience, language skills, physical work ability, and work motivation) are gathered through an online questionnaire (Brouwer, Bakker and Schellekens, 2015[23]). The Work Profiler has recently been revised to include new predictive items (e.g. household information) in the model.
Statistical model: Linking jobseeker data with outcomes
The second building block of statistical profiling is the statistical model. PES across the OECD use various models to examine the complex association between jobseeker characteristics and the likelihood of finding employment. While many PES employ more traditional regression-based models, such as logistic regression, a small but growing number are adopting AI-based models. These AI models allow for a more flexible specification of the relationship between jobseeker characteristics and outcomes. Moreover, they are particularly well-suited to handling unstructured data, enabling the use of a broader range of information and thereby enhancing the accuracy of predictions (Traverso et al., 2019[24]).
Currently, roughly one in six PES is using AI-powered profiling tools, with more PES in the process of implementing or planning to adopt such tools (Brioscú et al., 2024[9]). The most used types of AI are gradient boosting and random forests, both subsets of machine learning. For example, “OTT”, a decision support for counsellors used by the Estonian PES (EUIF), is enhanced by AI and achieves a high level of accuracy, with 95% of predictions being correct. The tool specifically employs gradient boosting and draws on approximately 60 variables, including jobseekers’ individual characteristics (e.g. education, previous job experience, entitlement to benefits, health) and labour market data (e.g. the number and type of vacancies at the regional level) to assess the probability of jobseekers finding employment and returning into unemployment. In addition, it identifies the key factors affecting these probabilities (Leinuste, 2021[25]).
Another example is the above‑mentioned VDAB profiling tool, which is based on a random forest model. Desiere and Struyven (2020[26]) evaluate the accuracy and fairness of the prototype used by VDAB and compare it with a rules-based profiling model. While the AI-based prototype is more accurate (66% accuracy compared to 58% for the rules-based model), it also shows a greater tendency to discriminate against jobseekers of foreign origin. This indicates that although AI-based profiling models can enhance the accuracy of the profiling process, they also present additional challenges, such as reduced transparency and increased risk of discrimination. This underscores the need for ongoing monitoring and evaluation of these tools to ensure accountability.
Output: Jobseekers’ distance from the labour market
Finally, the third building block of statistical profiling is the output, that is the jobseeker outcome predicted by the profiling model. In many PES across OECD countries, the statistical profiling tool predicts the probability of long-term unemployment, although the definition of long-term unemployment varies from country to country (Desiere, Langenbucher and Struyven, 2019[13]). For example, Australia, Italy, and the Netherlands use the probability of being unemployed for more than 12 months; Belgium and Sweden for more than 6 months; and Denmark for more than 26 weeks. Other countries, such as Estonia, predict the probability of returning to unemployment within a specified period. In some cases, exiting unemployment does not necessarily indicate successful labour market integration, as some jobseekers may leave unemployment to enter inactivity rather than employment. Consequently, some countries have refined their outcomes to capture transitions into employment, in addition to transitions out of unemployment. Examples of such outcomes are the probability of unsubsidised employment (e.g. Austria) and the probability of moving from unemployment to employment within 12 months (e.g. Ireland) (ESRI/ESRI/ESRI, 2022[27]). Constructing these outcomes, however, requires detailed employment data, which are not always available if unemployment and employment registers are not systematically linked.
Profiling score and service provision
Profiling tools can only improve service provision for vulnerable groups if they are meaningfully integrated into existing PES processes and lead to appropriate policies and measures for these groups. Typically, jobseekers are divided into three or more categories based on their profiling scores, with each category directed to different service streams. For example, the Irish PES categorises jobseekers into three groups (low, medium, and high) based on their probability of exiting unemployment within 12 months (known as the PEX – probability of exit). The PEX score, along with additional eligibility criteria, determines the intensity and timing of support that jobseekers receive. Group information sessions are organised for all jobseekers within three weeks of registration, regardless of their PEX scores. Jobseekers with a low or medium PEX score (i.e. a low or medium probability of finding employment) and aged over 25 receive more intensive counselling sessions. In contrast, jobseekers under 25 receive intensive counselling regardless of their PEX score (Desiere, Langenbucher and Struyven, 2019[13]).
Another example is VDAB’s profiling tool, which segments jobseekers into four groups based on their likelihood of finding employment within six months. Jobseekers in the red group have a probability below 35%; orange between 35% and 49.9%; yellow between 50% and 64.9%; and green 65% or higher. A fifth category (black) includes all jobseekers for whom a profiling score cannot be calculated due to missing data. This segmentation is used by VDAB to prioritise interactions between counsellors and jobseekers, with those in the red and black categories receiving the highest priority, while those in the orange, yellow, and green groups are receiving support subject to capacity (OECD, 2024[17]; VDAB, 2021[28]).
2.3.4. Digital tools can support the referral of clients to appropriate services
After identifying the needs of jobseekers, many PES across the OECD employ jointly agreed individual action plans (IAPs), which outline a pathway towards employment and specify the services and measures that will help improve the jobseeker’s employability (OECD, 2023[11]). Digital tools can support this process, both in the development of IAPs and in referring clients to suitable services.
Most PES (71%) use digital tools to develop and manage IAPs, often incorporating this functionality into the main user interface used by counsellors and jobseekers. In some PES, the output from other tools is used in creating these plans. For example, France Travail, the French PES, uses profiling results as a key input for developing IAPs. Although the use of AI to develop and manage IAPs remains rare, the emergence of generative AI (e.g. ChatGPT and Microsoft Bing AI) presents new opportunities for PES to develop tools that support their staff in the creation and management IAPs (Brioscú et al., 2024[9]).
Tools that enable jobseekers and those at risk of job loss to search for suitable training options are also common across PES (63%). This functionality is usually provided either directly on the PES website or through a dedicated tool. In some cases, these tools also offer occupation-specific information, such as wage levels or whether the occupation linked to a particular training programme is in high or low demand. One example is “Berufsinformat”,9 a chatbot developed by AMS, the Austrian PES, to provide jobseekers with information on job profiles and training opportunities, along with other occupation-specific details such as wage levels. Berufsinformat uses generative AI and draws on various AMS data sources, including the AMS Occupational Information System, which contains information on over 500 occupations (AMS, 2024[29]).
Digital tools can also help in targeting ALMPs, providing counsellors with recommendations on the appropriate support for different profiles of clients. The effectiveness of ALMPs in helping jobseekers reintegrate into the labour market can vary in function of the characteristics of the jobseeker. For example, some ALMPs may be more suitable for younger workers, while others may be better suited to older ones. Taking this information into account can help counsellors determine which ALMP is best suited to a specific jobseeker (Lechner and Smith, 2007[30]). However, only a few PES (34%) have tools specifically designed to target ALMPs. Often, this functionality is implicitly integrated into tools primarily focussed on other purposes, such as profiling tools that recommend different service streams based on risk profiles, or tools that map the distance of jobseekers to occupations and identify gaps in competencies, providing recommendations for training or giving counsellors a clearer understanding of the jobseeker’s situation, thereby allowing for a better assessment of which ALMPs might be most suitable.
One example of such a tool is SEND@, which was developed by the Spanish PES in 2019-2020 and later implemented on a full scale. This tool is designed for counsellors, providing tailored statistical information on each jobseeker that can be used during counselling sessions. For example, based on past jobseeker data the tool identifies potential changes in occupation for each jobseeker that could enhance their employability, thereby offering counsellors valuable insights to help decide which ALMPs might be most suitable for a client (OECD, 2023[31]). A similar functionality is incorporated in “Mon Assistant Personnel”, the user interface used by counsellors at France Travail, the French PES. This tool employs artificial intelligence to analyse jobseekers’ CVs and provides personalised recommendations on appropriate services, including training and job opportunities, acting as a decision support tool for PES counsellors (Brioscú et al., 2024[9]).
At DYPA, all counsellors use the same IT system offering various functionalities, including the development and management of individual action plans. As discussed in Section 2.3.2, the most job-ready jobseekers receive an Automated Digital Individual Action Plan. Automating the preparation of individual action plans for these jobseekers streamlines the collection and processing of information. By allowing jobseekers to enter the relevant information directly into the system to build their professional profile, the tool reduces the need for repeated data collection and administrative exchanges between jobseekers and employment counsellors, improves the consistency and completeness of data, and allows counsellors to allocate more time to core tasks, in particular providing support to jobseekers furthest from the labour market. DYPA is currently working on a Digital Employment Counsellor, a chatbot that provides personalised counselling services to jobseekers, enabling them to access information on suitable training options. The chatbot will also support jobseekers in updating and adding new information to their profiles, so that individual action plans can be updated accordingly. While such initiatives can assist vulnerable clients in navigating available information, they are heavily dependent on the clients’ own proactivity. As a result, they can complement personalised and tailored support provided by counsellors but cannot replace it.
To enhance the targeting of ALMPs and assist counsellors in selecting appropriate training for their clients, DYPA is also developing a new VET Management Information System – scheduled to be implemented in 2024 – aimed at making information on available training options more accessible to counsellors. In this area, DYPA could further enhance the targeting of ALMPs by incorporating results from counterfactual impact evaluations (see, for example, OECD (2024[7])) on the impact of ALMPs across different client profiles, with a particular focus on vulnerable clients.
2.3.5. Digital tools help provide employer services and job matching for vulnerable groups
PES across the OECD have long recognised the importance of collaborating closely with employers to support the labour market integration of jobseekers (Dromundo Mokrani, Lauringson and Xenogiani, 2024[32]). As a result, employers have become key stakeholders for PES, which have increasingly engaged in facilitating the matching of jobseekers with vacancies. This matching process involves several steps on both the employers’ and jobseekers’ sides (Broecke, 2023[20]). Employers design and draft job postings, advertise vacancies and screen candidates. Jobseekers, in turn, create application documents (e.g. CVs), search for vacancies and submit their applications.
To support employers and jobseekers throughout these steps, PES employ various digital tools. The most common tools across OECD countries are those that assist jobseekers and individuals at risk of job loss in finding vacancies (93%), and those that help employers find suitable employees (88%). These tools typically take the form of job platforms, allowing jobseekers to browse vacancies that match their interests and skills, and enabling employers to search for appropriate candidates. Although the primary aim of these tools is to facilitate the matching of labour supply with demand, their effectiveness depends on the proactivity of both jobseekers and employers.
One example of a digital tool specifically targeted to vulnerable groups is the web-based tool “Candidate Explorer,” developed by UWV, the PES of the Netherlands, in collaboration with Dutch municipalities. The tool’s purpose is to increase the visibility of individuals with an occupational disability to employers. In 2013, the Dutch Government and social partners signed the Job Agreement, which aimed to create 125 000 jobs for people with occupational disabilities. As part of this agreement, private companies are financially incentivised to hire individuals with occupational disabilities. The “Candidate Explorer” allows companies to access relevant labour market information about candidates with occupational disabilities while keeping their identities anonymous. Companies can then contact candidates through UWV. Additionally, UWV directly engages in matching candidates with suitable companies with a profile on the “Candidate Explorer”. (Werk.nl, n.d.[33]; OECD, 2023[11]; Kampers and van der Krogt, 2022[34]).
Matching jobseekers with vacancies is a core task in many PES across the OECD. This task is particularly important for vulnerable jobseekers, who often have limited work experience and face additional barriers to employment (e.g. disabilities, caregiving responsibilities), making it more challenging for them to find suitable vacancies and employers. Moreover, advanced job-matching tools can facilitate competence‑based matching, enabling jobseekers to be matched to vacancies based on their actual skills and competencies rather than other parameters such as their profession (Dromundo Mokrani, Lauringson and Xenogiani, 2024[32]). Competence‑based matching can be particularly beneficial for vulnerable jobseekers who lack formal professional qualifications (or whose qualifications are not recognised in Greece) but have nonetheless gained the required skills, for example, through practical experience (OECD, 2024[17]).
At least 78% of PES in the OECD use digital tools to facilitate this matching process. The task involves three main components: obtaining information on jobseekers’ profiles, gathering information on the requirements of vacancies and assessing the alignment between the two. Often, job matching is made on a limited set of criteria, such as education and previous job experience. However, because this approach considers only a restricted amount of information, its performance is often limited.
AI offers new opportunities to enhance the performance of job matching tools. First, it can improve the matching process by considering a larger amount of information, thereby ensuring a better alignment between the skills and competencies of jobseekers and the requirements of vacancies. Second, AI can assist in extracting information from both jobseeker profiles and job vacancies, enabling this data to be used more effectively in the matching process and facilitating competency-based matching (Brioscú et al., 2024[9]; Broecke, 2023[20]). Currently, approximately one in five PES is using AI to improve their job matching tools.
For example, the PES in Korea started using AI technologies to enhance matching in 2018-2020 (OECD, 2024[35]). Initially, AI was implemented to collect information from public job postings and create an up-to-date database of occupations and their requirements. This database served as the foundation for the matching process. The use of AI has since been extended to the matching process itself to incorporate additional information. The new AI-powered matching tool uses click data from comparable jobseekers on the job portal to identify vacancies that a jobseeker might be interested in. Similar approaches of competency-based matching enriched with click data are also being used or considered in other OECD countries, such as Belgium (Flanders) (VDAB, 2021[28]) and Luxembourg (Baer, 2023[36]).
Other functionalities commonly provided by PES across the OECD include those that assist jobseekers in creating CVs and other application documents (78% of PES) and in applying for vacancies (71%). Additionally, 63% of PES use tools that help employers design job postings and 25% have digital tools (including chatbots) providing information and counselling to employers. These tools can be employed to promote partnerships between PES and employers, including to facilitate “job carving”. Job carving involves reorganising tasks of a specific job to make it more accessible to certain jobseekers. For example, adjustments can be made to the pace and intensity of work to better accommodate candidates with disabilities (Dromundo Mokrani, Lauringson and Xenogiani, 2024[32]).
Less common are tools that proactively identify vacancies with a high likelihood of recruitment (24% of PES), even though such tools could be particularly promising for the placement of vulnerable clients. For example, historical data on the employment pathways of vulnerable clients could help identify employers and vacancies that are particularly well-suited for vulnerable jobseekers. In France, private sector companies are required to file a Pre‑Employment Declaration before hiring an employee. In 2018, France Travail, the French PES, developed the tool “La Bonne Boîte,”10 which uses predictive AI and data from Pre‑Employment Declarations to identify firms likely to recruit within the next six months, even before they post job vacancies (Brioscú et al., 2024[9]; owalgroup, 2019[37]; World Bank, 2023[38]). The aim of the tool is to provide jobseekers with access to the “hidden market” (i.e. vacancies that are not publicly advertised), allowing them to target their unsolicited applications to firms likely to hire. An evaluation of La Bonne Boîte found that recommending the use of the tool to jobseekers has a positive effect on their job-finding rates. Moreover, the tool helps reduce labour market mismatch by directing job seekers towards occupations in high demand (Behaghel et al., 2022[39]).
Except for tools enabling jobseekers to apply directly for vacancies, DYPA has digital tools in place covering all other functionalities. Additionally, DYPA is currently developing an advanced AI-based job-matching tool aimed at enabling competency-based matching between jobseekers and vacancies based on the ESCO classification. Moreover, DYPA has introduced a separate digital tool specifically for the tourism and hospitality sectors, which facilitates job matching between employers and jobseekers. Currently, the tool myDYPAlive supports contact between DYPA and employers, allowing, to some extent, the identification of employers with high recruitment potential. This practice could be further enhanced by leveraging historic data on jobseeker placements and their employers and drawing insights from the pathways of past jobseekers – particularly those from vulnerable groups – into employment. This would help identify additional employers who have not yet utilised DYPA’s services.
This note provides an overview of digital tools using data analytics employed by PES across OECD countries, discussing the added value these tools could bring to DYPA in strengthening its capacity to serve vulnerable groups. The overview categorises the tools based on their functionalities in four areas crucial to the labour market integration of vulnerable clients: outreach, needs assessment, referral to appropriate services, and job matching/employer services.
The overview shows that many of the functionalities offered by tools used by PES across OECD countries are also provided by the digital tools employed at DYPA, mirroring the impressive progress DYPA has made in digitalisation in recent years. However, the overview also identifies areas where DYPA could further invest to further enhance its support to vulnerable groups.
References
[3] Alexiou, I. and K. Koitsanou (2021), MyDYPAlive: Tele-counselling services to jobseekers and employers, European Commission, https://ec.europa.eu/social/BlobServlet?docId=24838&langId=en.
[29] AMS (2024), AMS Berufsinformat, https://www.ams.at/arbeitsuchende/aus-und-weiterbildung/berufsinformationen/berufsinformation/berufsinfomat?open=berufsinfomat#was-kann-der-berufsinfomat (accessed on 21 August 2024).
[36] Baer, I. (2023), Algorithms to facilitate job matching in Luxembourg, Digital tools for matching jobseekers and vacancies: international workshop, http://oe.cd/5ny.
[39] Behaghel, L. et al. (2022), Encouraging and Directing Job Search: Direct and Spillover Effects in a Large Scale Experiment, https://publications.banque-france.fr/sites/default/files/medias/documents/wp900.pdf.
[9] Brioscú et al., A. (2024), “A New Dawn for Public Employment Services. Service Delivery in the Age of Artificial Intelligence”, OECD Artificial Intelligence Papers, No. 19, OECD Publishing, Paris, https://doi.org/10.1787/5dc3eb8e-en.
[20] Broecke, S. (2023), “Artificial intelligence and labour market matching”, OECD Social, Employment and Migration Working Papers, No. 284, OECD Publishing, Paris, https://doi.org/10.1787/2b440821-en.
[23] Brouwer, S., R. Bakker and J. Schellekens (2015), “Predictors for re-employment success in newly unemployed: A prospective cohort study”, Journal of Vocational Behavior, Vol. 89, pp. 32-38, https://doi.org/10.1016/j.jvb.2015.04.001.
[14] Canada’s national employment service (2024), How does the Job Transition Tool calculate transition times?, https://www.jobbank.gc.ca/support/question?qaid=215&tid=59 (accessed on 22 August 2024).
[19] Clegg, D. and N. Durazzi (eds.) (2023), Predictive algorithms in the delivery of public employment services, Edward Elgar Publishing.
[13] Desiere, S., K. Langenbucher and L. Struyven (2019), “Statistical profiling in public employment services: An international comparison”, OECD Social, Employment and Migration Working Papers, No. 224, OECD Publishing, Paris, https://doi.org/10.1787/b5e5f16e-en.
[26] Desiere, S. and L. Struyven (2020), “Using Artificial Intelligence to classify Jobseekers: The Accuracy-Equity Trade-off”, Journal of Social Policy, Vol. 50/2, pp. 367-385, https://doi.org/10.1017/s0047279420000203.
[32] Dromundo Mokrani, S., A. Lauringson and T. Xenogiani (2024), “The role of public employment services in enhancing labour market matching”, in Handbook on Labour Markets in Transition, Edward Elgar Publishing, https://doi.org/10.4337/9781839106958.00030.
[2] DYPA (2024), New DYPA model office for citizens belonging to special social groups in Athens, https://www.dypa.gov.gr/storage/news/dt-neo-protipo-ghrafio-en.pdf (accessed on 2 September 2024).
[21] Ernst, S., A. Mueller and J. Spinnewijn (2024), “Risk Scores for Long-Term Unemployment and the Assignment to Job Search Counseling”, AEA Papers and Proceedings, Vol. 114, pp. 572-576, https://doi.org/10.1257/pandp.20241092.
[27] ESRI/ESRI/ESRI (2022), Predicting the probability of long-term unemployment and recalibrating Ireland’s Statistical Profiling Model, ESRI, https://doi.org/10.26504/rs149.
[40] Eurostat (2023), Glossary: Data analytics, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Data_analytics (accessed on 7 October 2023).
[8] Gazette of the Government of the Hellenic Republic (2022), Νόμος ΥΠ’ ΑΡΙΘΜ. 4921. Δουλειές Ξανά: Αναδιοργάνωση Δημόσιας Υπηρεσίας Απασχόλησης και ψηφιοποίηση των υπηρεσιών της, αναβάθμιση δεξιοτήτων εργατικού δυναμικού και διάγνωσης των αναγκών εργασίας και άλλες διατάξεις.
[5] Genova, A. and E. Davern (2022), Practitioner toolkit on strengthening PES to improve the labour market outcomes of persons with disabilities, https://ec.europa.eu/social/main.jsp?langId=en&catId=89&furtherNews=yes&newsId=10387.
[34] Kampers, T. and R. van der Krogt (2022), The Candidate Explorer: How to make persons with a occupational disability visible for employers.
[30] Lechner, M. and J. Smith (2007), “What is the value added by caseworkers?”, Labour Economics, Vol. 14/2, pp. 135-151, https://doi.org/10.1016/j.labeco.2004.12.002.
[25] Leinuste, K. (2021), Decision support tool „OTT“ for employment counsellors in the Estonian PES, https://www.oecd.org/content/dam/oecd/en/events/2021/10/new-digital-tools-for-pes-counsellors/pes-digital-oct2021-estonia.pdf.
[7] OECD (2024), Impact Evaluation of Training and Wage Subsidies for the Unemployed in Greece, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/4b908517-en.
[17] OECD (2024), Modernising Latvia’s Public Employment Service through Digitalisation, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/07e830f5-en.
[35] OECD (2024), Strengthening Active Labour Market Policies in Korea, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/44cb97d7-en.
[31] OECD (2023), Impact Evaluation of the Digital Tool for Employment Counsellors in Spain: SEND@: Report on the Design and Implementation of an Impact Evaluation of the Digital Counselling Tool for Spain’s Public Employment Services, OECD Publishing, Paris, https://doi.org/10.1787/fe1ec3c3-en.
[11] OECD (2023), Personalised Public Services for People in Vulnerable Situations in Lithuania: Towards a More Integrated Approach, OECD Publishing, Paris, https://doi.org/10.1787/e028d183-en.
[4] OECD (2022), Harnessing digitalisation in Public Employment Services to connect people with jobs, OECD Publishing, Paris, https://doi.org/10.1787/367a51f3-en.
[1] OECD (2021), “Building inclusive labour markets: Active labour market policies for the most vulnerable groups”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/607662d9-en.
[12] OECD (2021), Improving the Provision of Active Labour Market Policies in Estonia, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/31f72c5b-en.
[10] OECD (2016), Making Integration Work: Refugees and others in need of protection, Making Integration Work, OECD Publishing, Paris, https://doi.org/10.1787/9789264251236-en.
[37] owalgroup (2019), Artificial Intelligence in Employment Services - A Mapping. Final Report, https://tem.fi/documents/1410877/15020328/Artificial+intellingence+in+employment+services+-+A+mapping/24844ede-0570-c8da-4ed3-c91ec25b8e76/Artificial+intellingence+in+employment+services+-+A+mapping.pdf.
[16] Radix (n.d.), Radix and VDAB reinvent orientation test with Artificial Intelligence, https://radix.ai/cases/flemish-orientation-test-faster-and-more-accessible-than-ever-thanks-to-collaboration-between-radix-and-vdab/ (accessed on 29 July 2024).
[24] Traverso, A. et al. (2019), “Diving Deeper into Models”, in Kubben, P., M. Dumontier and A. Dekker (eds.), Fundamentals of Clinical Data Science, Springer International Publishing, https://doi.org/10.1007/978-3-319-99713-1_9.
[18] Tuccio, M. et al. (2023), “The OECD Skills Profiling Tool: A new instrument to improve career decisions”, OECD Social, Employment and Migration Working Papers, No. 293, OECD Publishing, Paris, https://doi.org/10.1787/598ff539-en.
[28] VDAB (2021), AI & Advanced Analytics @ VDAB, https://web-archive.oecd.org/2020-11-24/570855-AI_presentation_VDAB.pdf.
[15] VDAB (n.d.), Oriënt, https://orientatie.vdab.be/ (accessed on 29 July 2024).
[33] Werk.nl (n.d.), Kandidatenverkenner banenafspraak, https://www.werk.nl/werkgevers/wervingsadvies/personeel-subsidies/kandidatenverkenner-banenafspraak/index.aspx#paragraaf1 (accessed on 23 August 2024).
[22] Wijnhoven, M. et al. (2023), “The Work Profiler: Revision and maintenance of a profiling tool for the recently unemployed in the Netherlands”, International Social Security Review, Vol. 76/2, pp. 109-134, https://doi.org/10.1111/issr.12327.
[38] World Bank (2023), The Use of Advanced Technology in Job Matching Platforms: Recent Examples from Public Agencies, https://thedocs.worldbank.org/en/doc/ceb5c5792ad0d874e9b1c3cc71362f46-0460012023/original/Digital-Job-Matching-Platforms-S4YE-Draft-Note-for-Discussion.pdf.
[6] World Wide Web Consortium (2023), Web Content Accessibility Guidelines (WCAG) 2.1, https://www.w3.org/TR/WCAG21/ (accessed on 3 September 2024).
← 1. This report refers to jobseekers “furthest from the labour market” and “vulnerable” interchangeably, signifying these jobseekers that face multiple and complex barriers to employment (for example, related to skills, experience, health or care responsibilities). The official term used by DYPA is “special and vulnerable groups”.
← 2. For the calculation of counsellor caseloads, only DYPA employees who report working exclusively as counsellors are considered. However, in some offices, DYPA staff may perform multiple roles, including that of jobseeker counsellor.
← 5. Data analytics refers to the use of technologies, techniques or software tools for analysing data to extract patterns, trends and insights to make conclusions, predictions and better decision making with the aim of improving performance (Eurostat, 2023[40]).
← 6. 39 countries and 41 PES (responses from the three regional PES in Belgium are considered separately) responded to the questionnaire: 36 OECD Member countries (all OECD Member countries except for the United Kingdom and the United States) and 3 OECD accession countries in the European Union (EU): Bulgaria, Croatia and Romania.