One of the biggest megatrends in the world of Public Employment Services (PES) across OECD countries is the increased deployment of digital solutions to enhance service provision and processes. This chapter explores this PES modernisation journey underway within Nordic countries, all of which have taken significant steps to enhance their digital capacities in recent years. The chapter explores the digital channels and tools being used by Nordic PES to enhance and enable service provision to jobseekers and employers and discusses how Nordic PES are utilising advanced Artificial Intelligence (AI) technologies across their core areas of PES activities. The chapter also discusses the efforts taken by Nordic PES to measure the success of digital solutions and the impacts of certain international policies and regulations on digital innovations within Nordic PES.
The Role of Public Employment Services in Promoting an Inclusive Nordic Common Labour Market
Abstract
In addition to reforms to overall active labour market policy (ALMP systems), Public Employment Services (PES) in Nordic countries are also undergoing significant transformation through their digitalisation and modernisation agendas. In line with trends across OECD countries more widely, these efforts by Nordic PES to modernise their digital infrastructure has been underway to various degrees over recent decades.
Most recently, the COVID‑19 pandemic undoubtedly acted as an accelerant for digitalisation of many OECD and Nordic PES, seeing them turn to digital tools and service channels to ensure some level of service continuity for clients during the emergency period. In addition, it highlighted the different states of digitalisation across OECD PES and for some highlighted the need for upgrades or further investments in their digital capabilities (OECD, 2022[1]). While Nordic PES were comparatively well-positioned upon the onset of the COVID‑19 pandemic, further steps were still taken to enhance their digital capabilities and services during this time (OECD, 2023[2]).
Across OECD countries, PES modernisation is taking place through two main avenues. First, PES are increasingly adopting digital solutions to aid processes and service provision, be that digital solutions or service streams. Second, PES are advancing their digital capacity as part of more largescale and systematic upgrades to their operational and IT infrastructure.
This two‑fold approach to PES digitalisation applies also in the Nordic region, with Nordic PES pursuing modernisation through both channels, including in some cases in line with wider reform PES agendas. Nordic PES have deployed both digital tools and service streams to contribute to more effective and flexible services for both jobseekers and employers. In addition, Nordic PES have undertaken more significant digital advancements to improve their digital and IT infrastructure. This includes, among others, a new customer relationship management (CRM) system for PES staff in Finland (with widespread implications for services, including profiling, monitoring job-search activity of clients and targeting of ALMPs), the new PES operational IT system Galdur (or magic) in Iceland enabling more modern and data-driven processes and services, the gradual transition from legacy systems to more modern architecture in Sweden (including several AI solutions) and the modernisation of the case handling process in Norway.
These modernisation efforts offer significant opportunities for Nordic PES to enhance their efficiency and effectiveness. However, ensuring these digital solutions and tools have their desired impacts on end-users, both PES staff and clients, will require Nordic PES to further enhance monitoring and evaluation activities. Furthermore, the adoption of Artificial Intelligence (AI) by all Nordic PES, will require proactive steps to mitigate against the associated risks and challenges these technologies bring with them.
This chapter explores the core components of the modernisation journeys of Nordic PES, including the deployment of digital channels and tools (Section 3.2), the use of AI technologies (Section 3.3), how the success or impact of digital developments is being measured (Section 3.4) and finally gives consideration to the international policies and regulations impacting modernisation efforts of Nordic PES (Section 3.5).
As part of the digitalisation trend, PES in the Nordic region are increasingly relying on digital channels and tools to aid and directly facilitate service provision. This sees the vast majority of activities and services provided to jobseekers and employers able to be conducted on a remote or digital basis. However, across Nordic PES variation exists in the degree of service digitalisation. This section explores the availability of digital channels and solutions within Nordic PES in comparison to PES in OECD countries as a whole and considers the efforts being made by Nordic PES to mitigate against the risk of digital exclusion.
3.2.1. Thanks to positive experiences during the COVID‑19 pandemic, digital engagement channels have now been made permanent in all Nordic PES
In line with trends seen across OECD countries more widely, Nordic PES now all operate hybrid or blended models of service delivery, providing core services on a digital or remote basis. Looking at the core activity of job-search support and counselling, all Nordic PES facilitate this client support through a digital channel; either through a digital PES user interface or using a digital communications software and often to higher degrees than on average across OECD countries (Table 3.1). For the Danish PES in particular, this marks a departure from the pre‑COVID policy which did not facilitate digital or remote counselling. However, positive experiences during this emergency period have resulted in the maintenance of digital counselling as a permanent component of the service offering in Denmark.
Channels used for the provision of job-search support and counselling by country
Note: This figure is based on information from 39 PES in 38 countries, with information from two responding sub-national Belgian PES reported separately.
Source: Authors’ calculation based on responses to OECD questionnaire on digitalisation and AI use in PES conducted in spring 2023.
Outside of counselling, Nordic PES also perform well in providing flexible access to core PES activities. This includes the use of digital interfaces by all Nordic PES to facilitate both registration with the PES (compared to 95% of PES in OECD countries) and to enable the monitoring of clients’ job-search activities (compared to 79% of PES in OECD countries). The prevalence of these digital and remote access solutions to core PES services in the Nordic region provides additional flexibility to PES clients to engage with the PES according to their own preferences.
In providing multiple service streams, some Nordic PES have established rules or criteria to determine the channel of engagement used. In Denmark, the first interview must be in person, after which the client is free to decide their preferred mode of engagement (in person, digitally or by phone). In Norway, a client’s eligibility for counselling provided through digital channels is based on a number of criteria, including the ability for effective self-sufficient job-search, adequate digital literacy and being aged between 30‑59 years. This allows for more intensive engagement, including through other channels, with younger jobseekers and other target groups. The basis for this approach in Norway is NAV’s Channel Strategy, first introduced in 2015, coinciding with the organisation’s gradual shift away from traditional service delivery and towards increased use of digital technologies and solutions. The aim of this transition is to improve efficiency within NAV, including by prioritising resources towards those clients in need of the most-intensive engagement and support (Løberg, 2022[3]).
For PES in some OECD countries, the decision regarding the channel of service provision is further enhanced by the use of a profiling or diagnostic tool to assess a jobseeker’s digital skills or competencies. For example, the PES in the Walloon Region of Belgium has developed a simple diagnostic tool to diagnose the digital skills and autonomy of a jobseeker, deployed during the jobseeker’s initial registration meeting (European Commission, 2024[4]). It comprises four questions; the first two assessing the jobseeker’s ability to access the internet easily, regularly, and independently, while the next two evaluate their proficiency and autonomy in using online digital tools (such as for purchases or administrative tasks) and their familiarity with digital job-search tools. The recommended channel of service provision is then jointly informed by the outcome of the digital skills assessment and the profiling score (proximity to employment is calculated based on around 20 parameters). Similarly, the Pix Emploi online tool was developed in France jointly between the PES and a non-profit organisation to assess a jobseeker’s digital skills and is recommended jobseekers following their registration with the PES (OECD, 2024[5]).
3.2.2. Nordic PES perform comparatively well in the availability of digital and online solutions for both jobseekers and employers
First, looking at digital services for jobseekers, the availability of various digital services differs across Nordic countries (Table 3.2). In the beginning of a jobseeker’s journey, registering with the PES is possible online across all Nordic countries compared to 90% of OECD and EU PES. After this, some variation exists in how jobseekers can receive information on and access services. All Nordic PES except Denmark have deployed chatbots to provide information to jobseekers on ALMPs and associated eligibility conditions, compared to just over half of all PES (Section 3.3.1). Three‑in-five Nordic PES have digital solutions to both aid jobseekers to identify potential training options (Iceland, Norway and Sweden) and directly apply for such measures (Finland, Iceland and Sweden).
Online or digital solutions available for jobseekers, people at risk of job loss and citizens in Nordic PES and the share of OECD and EU PES
Note: Tick mark signifies that a functionality is available through an online solution in a given Nordic country. Share of OECD and EU PES refers to the share of responding countries with a given functionality available online or digitally (answers received from 40 PES in 38 countries).
Source: Authors’ calculation based on responses to OECD questionnaire on digitalisation and AI use in PES conducted in Spring 2023.
Digital recommender systems to suggest suitable occupations to jobseekers and aid their career orientation are implemented in Finland and Sweden, making them slightly more prevalent in the Nordic region than on average across OECD and EU countries (at 33%). Meanwhile, solutions to allow jobseekers to identify and test their competencies and skills are seen in two‑in-four OECD and EU countries, but in only one Nordic PES (Sweden). Such tools to better understand the skillsets of jobseekers can contribute to better targeted services and measures (particularly training) and can also be used as an input to career services. Digital solutions to aid the job-search journey of jobseekers are commonplace in Nordic PES, largely through the deployment of PES vacancy portals in all Nordic countries. In two Nordic countries, this is further enhanced by using AI algorithms to directly recommend suitable vacancies to jobseekers (Section 3.3.1). Finally, in meeting the job-search requirements often part of the mutual obligations of jobseekers, all Nordic PES offer digital solutions to facilitate jobseekers to report their job-search activities; compared to only 78% of OECD and EU PES.
The use of digital tools to serve employers is also well-developed in Nordic countries (Table 3.3). Much of this activity is to aid the matching process, with digital solutions to advertise vacancies and design vacancy postings present in all Nordic PES and at higher rates than across the OECD. In addition, the aforementioned chatbots used in four‑of-five Nordic PES to provide information to jobseekers have a dual functionality, also providing information to employers on available services and measures – an approach used in only one‑third of PES in OECD countries. Similarly, digital procedures for employers to apply for measures is a feature of all Nordic PES, despite only being possible in less than three‑in-five total PES. The main differences in digital services for employers in the Nordic region emerge in the area of information reporting. Three Nordic PES have online solutions for employers to share information with the PES on the status of their vacancies (compared to 65% of OECD and EU PES) and no Nordic PES provide digital means for employers to report information associated with their use of PES measures (compared to 43% of OECD and EU PES).
Online or digital solutions available for employers in Nordic PES and the share of OECD and EU PES
Note: Tick mark signifies that a functionality is available through an online solution in a given Nordic country. Share of OECD and EU PES refers to the share of responding countries with a given functionality available online or digitally (answers received from 41 PES in 39 countries).
Source: Authors’ calculation based on responses to OECD questionnaire on digitalisation and AI use in PES conducted in Spring 2023.
3.2.3. Nordic PES provide digital skills training and maintain in-person services to mitigate the risk of digital exclusion
The deployment of digital service streams and digital tools undoubtedly provide significant opportunities and benefits for PES, including in providing end-user flexibility and efficiency gains for PES. However, in moving towards these more modern and digital-driven services, PES must ensure that nobody is left behind on this journey and engage in efforts to mitigate any potential risk of digital exclusion. This includes, in particular, clients from vulnerable backgrounds, those without sufficient digital skills and those without means of accessing digital services (OECD, 2022[1]). At the Nordic level, Nordic ministers for digitalisation have also called for inclusivity to be central in the digital transition and digital public services (The Nordic Council and the Nordic Council of Ministers, 2021[6]).
In deploying multi-channel or hybrid services, a core objective for PES should be to endeavour to help clients access services through the most effective channel for them. In the Nordic region, the most common avenue by which PES are taking steps to mitigate digital exclusion is through the maintenance of traditional in-person access to PES services, providing flexibility for clients to engage with the PES according to their own preferences and needs. Such in-person services are particularly important for clients from vulnerable backgrounds and others where more intensive support and engagement is necessary.
For those clients without or with limited digital skills, PES clients in all Nordic countries can access opportunities to engage in digital skills training. In Sweden, this training known as “Digital Me” (Digitalajag) has been developed jointly between the PES and Google Digital Academy and aims to support clients to become more confident in using digital services (Arbetsförmedlingen, n.d.[7]). The training comprises three learning tracks: security and privacy, digital communication and online job-search. In Finland, guide materials are also present to help clients use digital solutions. In addition, to aid both clients without sufficient digital skills and those without means of accessing digital services (e.g. lack of access to a suitable device or internet connectivity), the PES in Denmark, Iceland and Norway provide digital access points in PES offices for clients to use and, if needed, to receive assistance in navigating digital solutions and resources.
Nestled within wider digital advancements, Nordic PES are also experimenting with AI technologies to enhance their process and services. This section explores these AI use cases by Nordic PES across the relevant areas of PES activity and the approaches being undertaken by Nordic and OECD PES to guide and govern the use of these technologies.
3.3.1. Experimentation with AI by Nordic PES is primarily limited to a number of core areas of PES activity
AI use is more widespread among PES in the Nordic region than across OECD countries more widely. All Nordic PES now have at least one AI solution in place, compared to just over one‑in-two PES in OECD countries as a whole (Brioscú et al., 2024[8]). Across OECD countries, examples of AI use are seen in all core areas of PES activity, however AI use cases within Nordic PES are limited to comparatively fewer areas of PES activity (Figure 3.1).
Share of PES in OECD and Nordic countries using AI by area of PES activity
Note: Share of PES refers to the percentage out of the total number of responding PES, whether they use AI or not. Based on information from 41 PES in 39 OECD and EU countries, with information from the three sub-national Belgian PES reported separately.
Source: Authors’ calculation based on responses to OECD questionnaire on digitalisation and AI use in PES and adapted based on Brioscú et al. (2024[8]), A new dawn for public employment services: Service delivery in the age of artificial intelligence, https://doi.org/10.1787/5dc3eb8e-en.
AI developments among Nordic PES see different levels of experimentation, in different areas and using differing algorithm types (Table 3.4). In terms of number of tools live or in production at the time of writing, the PES in Sweden stands out as a front-runner in the Nordic region with five AI-powered solutions. The other Nordic PES are using AI to enhance activities in either two (Finland, Norway) or one area (Denmark, Iceland). With these technologies becoming increasingly accessible, the prevalence of AI use among Nordic PES can likely be expected to increase in the coming years – along with adoption of AI in PES across OECD countries more widely.
Chatbots are the most common AI tool, seen in three‑in-five Nordic PES
The PES in Finland, Iceland and Norway are among the 17% of PES in OECD countries that have deployed AI powered chatbots to provide information to jobseekers and employers (Brioscú et al., 2024[8]). These developments have undoubtedly been aided by advancements in recent years, particularly in natural language processing; enabling chatbots to more intuitively process and respond to more complex prompts. This provides a clear advantage over traditional rule‑based chatbots, which can often require the recognition of predefined keywords to respond to a query and, unlike AI-driven models, cannot learn from their interactions with users over time.
AI use by area country and area of PES activity
Note: OECD average refers to the percentage of PES in OECD countries using AI in this area as a percentage of the total number of responding PES, whether they use AI or not.
Source: Based on both responses to OECD questionnaire on digitalisation and AI use in PES and Brioscú et al. (2024[8]), A new dawn for public employment services: Service delivery in the age of artificial intelligence, https://doi.org/10.1787/5dc3eb8e-en.
In Iceland and Norway, the PES have deployed AI chatbots to respond to queries from jobseekers and employers developed by Boost.AI (Vinný and Frida respectively), a Norwegian company specialising in conversational AI solutions. The PES in Finland has deployed two AI chatbots, Tarmo to answer queries from clients relating to PES services and measures and Aino which specialises answering queries relating to working in or hiring from Finland (Table 3.4).
Both Vinný and Frida have provided essential customer service capacity to their respective PES, including during times of increased demand such as during the COVID‑19 pandemic (OECD, 2023[2]; Brioscú et al., 2024[8]). During this emergency period, Norway’s PES faced a 250% increase in queries. Frida, the PES chatbot launched in 2018, provided the extra capacity needed, handling a workload equivalent to 220 full-time employees and resolving 80% of queries without human assistance. This significantly reduced the pressure on PES staff during the crisis, ensuring clients received timely and efficient responses to their questions.
Two Nordic PES are using AI to support career guidance services
Across OECD countries, 15% of PES are using AI to assist career management and job-search orientation activities. In the Nordic region, two‑in-five PES are using AI in this domain to date.
In Sweden, the PES has rolled out an AI-based tool designed to assist jobseekers in exploring potential career opportunities. The system uses various AI techniques to suggest occupations that align with an individual’s existing skills. Beyond offering new career ideas, the tool also connects jobseekers with relevant job listings from the Swedish national employment platform. Future updates are planned to broaden the tool’s functionality, such as integrating additional data sources, offering training suggestions, and providing more personalised job recommendations for individuals with disabilities.
Meanwhile, in Denmark, the PES launched an AI-driven tool in 2020 aimed at identifying the skills in demand by employers (OECD, 2022[1]; Westh Wiencken Vizel and Opstrup Hansen, 2021[9]). By analysing job advertisements using machine learning, the tool compiles an index of the most sought-after skills at any given time. This tool serves two main purposes: first, to help PES counsellors understand labour market trends and guide clients towards suitable jobs, education, or training; and second, to support education and training providers in evaluating the market relevance of their courses.
Several Nordic PES have deployed AI solutions to assist different aspects of the matching process
One‑in-five OECD PES have implemented AI-enabled matching solutions, which directly produce recommendations of job vacancies based on the profile of a jobseeker. Traditional (non-AI) vacancy matching tools use rigid one‑to‑one comparisons for each criterion, which often leads to poor performance in identifying strong matches. This is partly because these tools can only process limited information, making them less effective in dynamic and ever-changing labour markets. In contrast, AI-assisted matching has the potential to incorporate a broader range of data sources, including user behaviour such as click data. This allows for a more comprehensive analysis of the alignment between jobseeker profiles and job descriptions, resulting in broader and more accurate matching opportunities. In the Nordic region, Finland and Sweden are the only PES currently seeking to yield the benefits of AI in aiding the matching job-matching process.
In 2022, Finland launched Job Market Finland, a digital platform designed to connect jobseekers and employers using advanced AI technology. The platform offers an AI-powered vacancy matching service that calculates a compatibility score between candidates and job openings through two methods (Job Market Finland, 2023[10]; Hirsimäki, 2023[11]; Työmarkkinatori, 2023[12]). The first method involves structural data matching, where factors such as jobseeker’s work experience, skills, education, language proficiency, and location are assessed. Points are awarded for relevant professional history and skills, while mismatches in language, education, or location can reduce the score. The second method applies natural language processing (NLP) to job postings on the platform, using neural networks to extract key terms. This allows for the analysis of job relevance to individual profiles, with the system trained on vacancies in Finnish, Swedish, and English. As a result, jobseekers receive personalised lists of suitable jobs, while employers get lists of compatible candidates. Before its development, an ex-ante evaluation of Job Market Finland was conducted, which included a literature review and analysis of Finnish PES vacancy data. The study revealed that web-based matching methods improve vacancy filling efficiency by 33%, reducing both vacancy duration and recruitment periods (Räisänen, 2023[13]).
Sweden has been working on the development of an AI-driven matching tool, which had its initial release in early 2024. This tool integrates a range of advanced techniques, including machine learning, deep learning, graph analytics, and natural language processing. It leverages data from both administrative records and vacancy listings to analyse jobseekers’ profiles and match them with relevant opportunities. In addition to its core matching function, the system offers extra features to support jobseekers in making broader career decisions and improving their overall career orientation.
In addition to directly producing job recommendations, a number of PES in OECD countries are also using AI to further support job matching, including to help aid the design of vacancy postings including occupational classification, detect illegalities in vacancies, diagnose hard to fill vacancies and proactively predict recruitment likelihood. In the Nordic region, the use of AI in these areas is limited to three examples in three PES:
Finland’s matching platform, Job Market Finland, features an integrated Skills Suggester tool that utilises natural language processing to analyse user input from both jobseekers and employers. Based on ESCO (the European Skills, Competences, and Occupations classification), it recommends relevant occupations and skills.
AI can also be used to assist employers in drafting job advertisements that they submit to the PES vacancy portal. Here, the PES in Norway has since 2023 deployed a large language model to offer suggestions to employers when they are drafting vacancy postings.
Since 2023, Sweden has employed AI to detect instances of discrimination in job vacancies submitted to the PES. This tool leverages deep learning and natural language processing to assist case managers in reviewing job advertisements posted on the PES vacancy portal.
Sweden is the only Nordic PES using AI to contribute to administrative activities and knowledge generation
While not yet widespread, a number of PES in OECD countries are also using AI to assist certain back-office activities and processes. Sweden has deployed two AI powered solutions related to these activities.
Benefit fraud refers to the illegal receipt of benefits, with no underlying entitlement or under false pretences. AI has the potential to detect such fraudulent activity, with the Swedish PES being the only example in Nordic and OECD countries currently using AI for this purpose. This system, implemented in early 2024, uses a number of technologies (machine learning, deep learning, social network analysis, and knowledge graphs) and draws on a diverse array of data sources, incorporating details about jobseekers, employees, employers (including contract information), and suppliers (partners and service providers associated with the PES). AI use in this domain carries considerable risks if implemented poorly. When these systems incorrectly identify fraud, they may issue corrective action requests – such as demands for benefit repayment, often with additional fines or interest – targeting individuals who are genuinely eligible for the benefits, leading to wrongful sanctions and significant distress for impacted individuals (Brioscú et al., 2024[8]).
AI also presents opportunities to enhance its analytical capacity, both through the production of evidence (including, for example, counterfactual impact evaluations) and generation of labour market information. At present, only a few PES in OECD countries are using AI in this realm but can be expected to increase in prevalence in the near future. In the Nordic region, the Swedish PES is utilising AI to analyse labour market trends. Natural language processing is employed to analyse job postings from the preceding two years to gain insights into the skills required, educational qualifications needed, and the task composition of various occupations. This initiative is part of a broader effort within the Swedish PES to create AI solutions that monitor and analyse labour market trends, including shifts in demand for specific occupations and competencies. Additionally, the AI-generated labour market information also acts as an input into the aforementioned career orientation and matching tools of the Swedish PES.
3.3.2. The Norwegian PES has taken formal steps to guide the use of AI and mitigate its risks
While AI brings with it many potential opportunities for PES to aid their processes and service provision and to contribute to their objective in connecting people with jobs, it also presents a number of challenges and risks (Brioscú et al., 2024[8]). These include, among others, the need to establish clear lines of responsibility for AI systems, issues surrounding transparency and explainability, risks concerning data quality and privacy, the risk of bias and discrimination, potential resistance and lack of skills among staff and clients and the need for ongoing monitoring and evaluation. Therefore, successfully seizing the benefits AI presents will require Nordic PES to be cognisant of and take steps to mitigate the associated risks.
In grappling with these risks and navigating the governance of these developments, one approach can be the development of a strategy to guide the AI transformation within the PES. While all Nordic PES recognise and acknowledge the challenges that AI presents, the PES in Norway is the first to take formal steps to tackle these concerns by developing an AI-specific strategy in 2024 (Box 3.1). Such a strategy can be useful to help define the vision, rationale and objectives for AI use within the PES and can help promote understanding and acceptance of the associated changes that AI may bring, including to the processes and tasks undertaken by staff. In the case of Norway, this new strategy is intended to be a “living document” that can be adapted based on experiences and lessons gained from further deployment of AI solutions in the coming years. Other similar initiatives across OECD countries, include the “Charter for Ethical AI” introduced by the French PES in 2022, drafted by a multi-disciplinary working group which included PES counsellors and jobseekers in its membership (Pôle emploi, 2022[14]). For other Nordic PES considering developing a strategy to guide AI developments, the OECD’s AI Principles offer a useful starting point to identify the core values that should be embedded in this work (OECD, 2024[15]).
NAV, the Norwegian Labour and Welfare Administration, has been experimenting with AI since 2017 following the creation of an “AI laboratory” (or AI lab) to explore how AI could aid various aspects of its services and processes. With a number of AI tools in production and several more in the exploratory phase, a strategy to set the vision and objectives for AI use within NAV was launched in 2024.
This AI strategy is structured around three core strands:
Ambitions – NAV’s aim is to use AI to contribute to: i) getting more people into employment, ii) more accurate benefit decisions, that are understood by the user, iii) greater inclusion and a reduced rate of people not in education, employment or training (NEET), and iv) making the working day of NAV employees easier.
Enablers – The strategy sets out a variety of key enablers that will impact NAV’s ability to achieve its AI’s ambitions. These include, among others, the need for clearly defined roles and accountability, high-quality data, competence and experience with AI at all levels of the organisation, clear and efficient processes and frameworks for AI initiatives, easy to use technology and the involvement of expertise from various disciplines to be involved in and learn from these developments.
Principles – Finally, the strategy establishes the six principles to ensure responsible use of AI within NAV: i) assess consequences (potential impacts and outcomes), ii) privacy, iii) fairness, iv) explainability, v) security, and vii) transparency.
The development of the strategy was kick-started by the AI lab, before being presented to NAV’s senior leadership. Consultations were held across the organisation to shape the draft strategy, drawing input from cross-disciplinary experts, including those from IT, employment services, and benefit administration departments. The strategy is intended to evolve and be revisited over time, with semi-regular updates, to take into account lessons learned and experiences gained as AI becomes more prevalent within the organisation.
Source: Vegard Sparre (2024[16]), “NAV’s AI strategy and use cases”, presentation at an OECD-EC-NEA international workshop on the use of AI by PES organised in the context of the joint OECD-European Commission (Directorate‑General for Structural Reform Support) project: Optimising processes and services at the National Employment Agency of Bulgaria.
Governing and promoting responsible use of AI by PES can also take to form of frameworks or structures to oversee these developments. While examples in Nordic PES are yet to be seen, prominent examples from OECD countries more widely include the Ethics Committees or Boards utilised by both the PES in France and Flanders, Belgium:
The Ethics Committee of the French PES was established in 2021 and is composed of 11 independent experts from diverse backgrounds. This includes members of the PES Board of Directors, technical, ethical, and legal experts, as well as representatives of both the unemployed and recruiters. The Committee’s role is to offer advice and recommendations on AI-related initiatives, ensuring they align with the values and public service mission of the PES (Pôle emploi, 2021[17]).
In Flanders, the PES Ethics Board was established in 2022 and consists of seven members (three internal and four external), with a mix of academic, legal, and operational expertise. The board’s role is to oversee the responsible and ethical use of AI within the PES and to offer independent advice on AI-related issues and developments (VDAB, 2022[18]).
Ensuring close co‑operation with developers of AI tools for PES, whether internal or external, is also crucially important, particularly to ensure that solutions meet the needs of the PES and end-users (Brioscú et al., 2024[8]). This is being prioritised in the Norwegian PES, where developers of digital and AI solutions are being placed within the relevant departments of the PES, so that they can work alongside the relevant experts and enhance their understanding of the context for the developments.
Just like changes to ALMP measures themselves, digital solutions (both tools and services) implemented by PES should be subject to both monitoring and evaluation efforts to assess end-user experiences and understand their impacts.
For Nordic PES, the primary approach to monitoring the success of digital tools and services is through the conduction of customer satisfaction surveys. For example, user views of digital services are sought as part of the Icelandic PES’ annual client survey (of both jobseekers and employers). In Denmark, client views are also sought in the development process of digital solutions. The Danish PES offers users of their services and systems the opportunity to participate in panels, which are then used to receive feedback on new proposals or solutions that are being developed. Take‑up and usage rates and patterns are also examined by most Nordic PES.
Regarding evaluations, such assessments of digital advancements by Nordic PES are largely not well-developed. Nordic PES note that while increased digitalisation and data-driven processes are indeed enabling better monitoring, formal evaluation still remains more difficult to operationalise. Even in those Nordic countries with strong track records in undertaking counterfactual impact evaluations of ALMPs, including through randomised control trials and piloting (for example in Denmark and Sweden in particular), systematic and rigorous efforts to measure the effectiveness of digital services and solutions are not yet developed. Nordic PES should take steps to embed channels for evidence generation into the development of digital tools, in order to understand their true impacts on end-users and on their contribution to the effectiveness of PES services. For example, in the Netherlands, the PES moved to a blended service (both in person and online) model in 2016, following a digital first approach in the years prior (Lieman, 2024[19]). To assess the impact of this transition to blended services, an evaluation was undertaken between 2017‑20. The study randomised participants into one of three groups: i) new blended services, ii) online service only and iii) online services with extra attention to enforcement. The study found that conversations with a counsellor had a positive impact on client outflow and found a positive return on investment of blended services (EUR 115 million after 30 months); resulting in multi-channel services, encompassing both digital and face‑to-face engagement methods, being retained as part of the strategy of the PES. In addition, the OECD undertook an evaluation of a digital counselling tool in Spain (Send@), finding that jobseekers counselled with the tool increased their participation in ALMPs and contributed to positive employment outcomes (OECD, 2023[20]).
In the case of AI technologies, rigorous monitoring and evaluation should also take place on a continuous basis. This should not only take place to measure the effectiveness of AI solutions, but also to detect any deterioration in the model or potential bias development (Brioscú et al., 2024[8]). Furthermore, establishing a feedback loop is just as crucial as the monitoring and evaluation process itself, ensuring that corrective actions can be taken and learnings from monitoring and evaluation activities can be fed into future updates and developments.
A number of international policies and regulations are having an impact on the digitalisation decisions and developments undertaken by Nordic PES. In addition, differences in interpretation in some cases are leading to differing outcomes in what digital services or solutions are considered.
All Nordic countries are subject to the General Data Protection Regulation (GDPR), directly for EU Members (Denmark, Finland and Sweden) and by adopting the regulation through national laws for Iceland and Norway, as part of the European Economic Area (EEA). GDPR has significant implications for PES as holders and handlers of a wealth of sensitive administrative data sources, requiring significant time and work for PES to implement. In the realm of digitalisation, all Nordic PES report GDPR as being one of the main obstacles or challenges they face. Despite the common legal basis, differences in interpretation of GDPR exist across and within European countries. The same holds true in the Nordic region, with differences in interpretation of GDPR across countries, as well as differing national legislation on data privacy, in turn resulting in differing approaches to digitalisation and the development and use of digital tools within Nordic PES. For example, despite advanced job-matching systems or tools in several PES (including in Sweden and Finland in the Nordic region), the Norwegian PES has been struggling to find a solution that satisfies the strict national interpretation of the regulation. In addition, the use of a digital communications software (such as Microsoft Teams, etc) to facilitate client engagement is not permissible in the Norwegian PES, also as a result of GDPR, despite being an available channel in Finland, Iceland and Sweden.
The use of AI by Nordic PES is also impacted by international regulation, both GDPR and the recently approved EU AI Act (European Parliament and Council of the European Union, 2024[21]). While AI is not explicitly referenced in GDPR, many of its principles are highly relevant to the field of AI, particularly those concerning fairness and transparency. GDPR also addresses automated decision-making, stipulating that individuals should not be subject to decisions based only on automated processes. This challenge is particularly acute for those Nordic PES responsible for the administration of benefits and seeking to modernise these processes. In other European countries, automating aspects of the benefit administration system has been possible under GDPR, provided human supervision takes place. For example, in Estonia, although not powered by an AI-algorithm, the PES has been for several years using automated decision-making to process unemployment benefit applications (Estonian Unemployment Insurance Fund, n.d.[22]). However, in some cases the decision of PES not to engage in the use of AI or other algorithms in such sensitive areas of activity are not only driven by regulatory restrictions, but also due ethical concerns, high-profile cases that led to poor results and the high-risk nature of this work and its impact on citizens (Brioscú et al., 2024[8]; OECD, 2024[23]).
The EU AI Act classifies AI according to its risk, defining four risk levels (unacceptable, high, limited and minimal) and establishing associated obligations for developers and deployers (OECD, 2024[24]). The Act also establishes the necessary governance structures required to oversee and supervise implementation and enforcement at both the national and European levels. Similar to the OECD’s AI Principles (OECD, 2024[15]), the AI Act underpins the essential nature of transparency, explainability and accountability in AI use and the implementation of rigorous data governance practices, including to prevent the potential for bias. These principles should be at the centre of all Nordic PES AI developments, with the development of PES AI strategies (such as those discussed in Section 3.3.2) to ensure compliance with regulations such as the AI Act and GDPR and transpose the core principles into the overall PES vision and operating processes. In addition, establishing avenues for communication and exchange between technical and legals teams across Nordic PES could help in unlocking certain challenges relating to GDPR and the EU AI Act.
All Nordic PES have also engaged in efforts to employ the European Skills, Competences, Qualifications and Occupations (ESCO) classification, most commonly to aid matching tools and associated infrastructure that processes jobseeker profiles and vacancies registered by employers. In addition, as members of EURES (EURopean Employment Services), the European initiative to facilitate the free movement of workers, the adoption of ESCO also facilitates Nordic PES to enable job mediation through the EURES system and portal (see Chapter 4 for further details on EURES). ESCO functions as a comprehensive dictionary that describes, identifies and classifies skills, competences and occupations pertinent to the EU labour market (European Commission, 2024[25]). Currently, ESCO provides definitions for 3 039 occupations and associated 13 939 skills and is available in 28 languages – all official EU languages, Icelandic, Norwegian, Ukrainian and Arabic. The aim of ESCO is to provide a common framework and understanding of occupations and skills, with the objective of contributing to labour mobility within Europe. In an increasingly digital labour market, where the vast majority of both jobseekers and employers utilise digital tools to facilitate job-search and recruitment, ESCO provides a useful framework to understand the knowledge, skills and competences associated with a given occupation. In addition, in the work of Nordic PES in promoting the common Nordic labour market and addressing skills shortages, having a common approach to occupations and skills between Nordic and European countries is certainly an asset.
For several Nordic PES, ESCO implementation was a significant and resource‑intensive multi-year undertaking; an experience also felt more widely by many European PES (OECD, 2024[5]). In addition, some report various challenges, including in manually validating and updating translations and adapting terminology to the national context, including to account for colloquial terms used for many occupations in local vernacular. However, once operational, Nordic PES report positive experiences in using ESCO, including in particular in powering job-matching tools such as Job Market Finland for example (discussed in Section 3.3.1). In addition, ESCO provides opportunities to assist Nordic PES in navigating changing labour market dynamics, including those presented by the twin green and digital transitions – something being explored at the European level by the Swedish PES (Box 3.2). In addition, opportunities exist for Nordic PES to further utilise ESCO to promote the common Nordic labour market. This includes producing analysis and labour market information to better understand skills needs, advise the provision of training and inform career guidance and job-search support provided to clients.
Within the remit of the European Network of Public Employment Services (as discussed in Chapter 4), the Swedish PES is chairing the Working Group on Taxonomies. The Working Group, in collaboration with ESCO, has developed a comprehensive methodology and a dataset for green and digital competencies, adapted to the Swedish labour market context. Based on the ESCO classification framework, this work seeks to enhance interoperability and to establish a common language to support the green and digital transition throughout the European labour market.
The Swedish PES is piloting the taxonomy dataset. The aim is to create a clearer understanding of which skills are green or digital, which down the line could help the PES in decisions surrounding the targeting of supports, including for example in assessing whether employment incentives (such as wage subsidies for example) are contributing to a more low-carbon economy and in helping counsellors guide clients towards training in green and digital competencies in demand in the labour market (see also Chapter 2).
Source: JobTech Development (2024[26]), Mapping ESCO Skills and Occupations: Testimonial from Sweden, https://jobtechdev.se/en/news/mappning-mot-esco-erfarenheter-sverige-; Arbetsförmedlingen (2023[27]), ESF-projekt Kickstart React-EU: Slutrapport, https://arbetsformedlingen.se/statistik/analyser-och-prognoser/analys-och-utvardering/2023/slutrapport-kickstart-react-eu.
References
[27] Arbetsförmedlingen (2023), ESF-projekt Kickstart React-EU: Slutrapport, https://arbetsformedlingen.se/download/18.5f0045ec18ab08d5f073797/1697803269005/Kickstart%20REACT-EU%20slutrapport.pdf (accessed on 24 October 2024).
[7] Arbetsförmedlingen (n.d.), Digital me – your training in everyday digital knowledge, https://arbetsformedlingen.se/other-languages/english-engelska/about-us/our-mission/training-material/digital-me (accessed on 14 October 2024).
[8] Brioscú, A. et al. (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.
[22] Estonian Unemployment Insurance Fund (n.d.), When does the Unemployment Insurance Fund make an automatic decision regarding you, https://www.tootukassa.ee/en/data-protection-terms/data-protection-estonian-unemployment-insurance-fund/when-does-unemployment (accessed on 21 October 2024).
[4] European Commission (2024), “PES digital services to successfully integrate vulnerable jobseekers”, European Network of Public Employment Services: Thematic Paper, https://data.europa.eu/doi/10.2767/330517.
[25] European Commission (2024), What is ESCO?, https://esco.ec.europa.eu/en/about-esco/what-esco (accessed on 24 October 2024).
[21] European Parliament and Council of the European Union (2024), “Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024”, Official Journal of the European Union, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689 (accessed on 21 October 2024).
[11] Hirsimäki, T. (2023), Job Market Finland: Meeting place for people and jobs, Digital tools for matching jobseekers and vacancies: International workshop, https://www.oecd.org/els/emp/presentation_job_matching_Finland.pdf.
[10] Job Market Finland (2023), How the matching works, https://tyomarkkinatori.fi/en/info/instructions-and-frequently-asked-questions/how-the-matching-works.
[26] JobTech Development (2024), Mapping ESCO Skills and Occupations: Testimonial from Sweden, https://jobtechdev.se/en/news/mappning-mot-esco-erfarenheter-sverige- (accessed on 24 October 2024).
[19] Lieman, N. (2024), “Effectiveness of blended Public Employment Services”, Presentation at joint WAPES, France Travail, OECD and European Commision Directorate-General for Structural Reform Support Online Workshop April 2 2024 on the topic of Reforms of Public Employment Services: Thematic Workshop 2 - Digitalisation of services.
[3] Løberg, A. (2022), “Utviklingsoptimisme kan skjule økonomiske og sosiale kostnader i den digitale forvaltningen”, Arbeid og velferd, Vol. 3 (2022), https://www.nav.no/no/nav-og-samfunn/kunnskap/analyser-fra-nav/arbeid-og-velferd/arbeid-og-velferd/arbeid-og-velferd-nr.3-2022/utviklingsoptimisme-kan-skjule-okonomiske-og-sosiale-kostnader-i-den-digitale-forvaltningen (accessed on 2024 October 17).
[24] OECD (2024), “Governing with Artificial Intelligence: Are governments ready?”, OECD Artificial Intelligence Papers, No. 20, OECD Publishing, Paris, https://doi.org/10.1787/26324bc2-en.
[23] OECD (2024), Modernising Access to Social Protection: Strategies, Technologies and Data Advances in OECD Countries, OECD Publishing, Paris, https://doi.org/10.1787/af31746d-en.
[5] OECD (2024), Modernising Latvia’s Public Employment Service through Digitalisation, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/07e830f5-en.
[15] OECD (2024), Recommendation of the Council on Artificial Intelligence, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449.
[20] 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.
[2] OECD (2023), Nordic Lessons for an Inclusive Recovery? Responses to the Impact of COVID-19 on the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/2aa7bcc1-en.
[1] OECD (2022), Harnessing digitalisation in Public Employment Services to connect people with jobs, OECD Publishing, Paris, https://www.oecd.org/en/publications/harnessing-digitalisation-in-public-employment-services-to-connect-people-with-jobs_367a51f3-en.html.
[14] Pôle emploi (2022), Pôle emploi se dote d’une charte pour une utilisation éthique de l’intelligence artificielle, https://www.francetravail.org/accueil/communiques/pole-emploi-se-dote-dune-charte-pour-une-utilisation-ethique-de-lintelligence-artificielle.html?type=article (accessed on 17 October 2024).
[17] Pôle emploi (2021), Lancement du comité d’éthique sur l’intelligence artificielle, https://www.francetravail.org/accueil/communiques/lancement-du-comite-dethique-sur-lintelligence-artificielle.html?type=article (accessed on 17 July 2024).
[13] Räisänen, H. (2023), How to approach the ex ante employment effects of the Job Market matching platform?, EC-OECD Technical Workshop: Evaluating labour market policies through the use of linked administrative data. Paris, France, https://www.oecd.org/els/emp/Session3_Job_Market_Matching_Platform.pdf.
[6] The Nordic Council and the Nordic Council of Ministers (2021), Common statement on the importance of promoting digital inclusion as a central part of the digital transformation in the Nordic-Baltic region, https://www.norden.org/en/declaration/common-statement-importance-promoting-digital-inclusion-central-part-digital (accessed on 17 October 2024).
[12] Työmarkkinatori (2023), Job Market Finland: Meeting place for people and jobs, Digital tools for matching jobseekers and vacancies: International workshop, https://www.oecd.org/els/emp/presentation_job_matching_Finland.pdf.
[18] VDAB (2022), VDAB lanceert eerste Ethische Raad voor gebruik van data en AI, https://www.vdab.be/nieuws-voor-de-pers/algemeen/vdab-lanceert-eerste-ethische-raad-voor-gebruik-van-data-en-ai (accessed on 17 July 2022).
[16] Vegard Sparre, J. (2024), “NAV’s AI strategy and use cases”, OECD-EC-NEA international workshop on the use of AI by PES organised in the context of the project "Optimising processes and services at the National Employment Agency of Bulgaria".
[9] Westh Wiencken Vizel, D. and E. Opstrup Hansen (2021), The skills tool: Developing a tool Using AI methods to measure employers demand for skills in online job posting, https://www.oecd.org/els/emp/PES-Digital-Oct2021-Denmark.pdf.