The Responsible Organisation

Emprego en Galicia is the Public Employment Service of the Xunta de Galicia (Spain) and is dedicated to the development of the organisational and workforce fabric of the region, focusing on information dissemination on job vacancies and supporting training activities and active labour market policies. Among other objectives, the mission of the agency is to: 

• Create spaces that generate synergies and resources that enhance the agility, flexibility, and capacity for adaptation and innovation of the workforce in the face of changing situations in the labour market, increasing the competitiveness of organisations and labour well-being. 

• Generate change in the current work environment, improving people’s skills to build resilient organisations. 

The problem

The Emprego en Galicia lacked a comprehensive approach to analyse the competencies and skills offered by job seekers and the skills demand by employers, leading to inefficient job matching on the market. Such a mismatch between labour demand and supply, driven mainly by asymmetrical information on the type of skills offered and requested is also worsened by the lack of structured data. Career counselling and decision-making for both job seekers and companies is therefore not leveraged to its full potential without a data-driven and competency-based approach.

The Solution and its implementation

To enhance its job matching public service, Emprego en Galicia has implemented “EMi – Emprego Intelixent”, an AI-based platform available in all public employment offices in the region of Galicia that assists job seekers in their career pathway through a range of tools. 

Technical features 

EMi platform implements multiple AI-powered tools to support the agency’s employment services.  

• First, EMi uses advanced natural language processing (NLP) techniques to automatically classify job offers received and gathered from online regional job portals. Each job offer is assigned to a specific occupational standard code called CNO, which allows to have a comprehensive and structured view of all job offers with easily searchable and machine-readable codes. At the same time, skills are evaluated and classified similarly. As of today, the taxonomy covers over 13,000 skills and training competences, which are assigned automatically by the tool. 

• Two other different machine learning models are included in the platform to predict labour market evolutions in the region (based on Facebook’s Prophet, which is known for time series analysis) and to carry out an overall applicant employability assessment (based on XGBoost, an open-source library for implementing predictive modelling). 

• Lastly, EMi contains a training search engine that analyses an extensive corpus in Spanish and Galician language to improve the accessibility to training content offered through the platform, matching the missing skills of job seekers with helpful content. 

Overall, these models are integrated with one another to aid the job seekers with suggested career paths based on their future employability rate. This allows Emprego en Galicia’s technicians to design a tailored insertion pathway for each jobseeker in a more effective way. 

EMI’s adoption  

The adoption of EMi has been gradual and participative, involving professionals from the Galician Public Employment Service from design to implementation. Four activities have been carried out to assure an effective and smooth adoption:  

• Working groups with the Emprego en Galicia staff were organised to design and validate the usability and functionalities of the solution. 

• Incremental phasing of the solution with partial releases to test usability and functionality was adopted. During development, pilots of partial releases were conducted in three representative employment offices (urban, rural, etc.) to evaluate and improve the solution before extending it across the region. 

• Involvement of external suppliers with subject matter expertise was sought. Through the Regional Ministry of Employment, five external suppliers and local SMEs have been involved in the technological development, in the data gathering and structuring and employment specialists for the preparation of the taxonomies. 

• End-user support was provided with an ad-hoc change management team. Over 50 expert users were extensively trained to act as change agents, ensuring smooth implementation and overseeing usability design aligned with the needs and expectations of the end users, namely job seekers and companies. Not only they have been involved in the whole design and piloting process, but now end users have access to a support service for training, functional issues, and identifying potential technical problems, permitting regular monitoring and feedback now that the tool is “live” as well. 

Currently, the tool is in use in the production environment in the 53 employment offices in Galicia as well as in 128 collaborating entities, adding up to a total of more than 400 public employees. 

Future plans 

• A work plan aims to guide more than 70,000 job seekers using EMi by September 2025, while at the same time improving as a technical solution, with the development of new features, including: 

• A prediction model for job openings 6 months in advance, powered by Generative AI (see below); 

• A “Talent Map” that will allow to have a spatial overview of the worker’s skills in Galicia at the local council level, skills that employers are looking for and the future trends detected based on historical data. 

• A “Training Needs Map” that will show the competencies that are lacking to inform training offers that each local council organise. An alert system will also be introduced to the platform. 

• A machine learning -based model that will support in the analyse patterns and decisions made by job counsellors when helping job seekers, evaluating their efficacy and hence recommending them similar actions for other comparable job seekers. 

There is a plan to add these new features above and expand them even more with Generative AI, as the development of a chatbots to assist job seekers with self-profiling and soft skills testing, reducing the workload of employment offices and giving job seekers more control over their job search process. It is also planned to offer a chatbot to the companies that need to share their vacancies. Moreover, the integration into the platform of generative AI will also enable the extraction and association between skills from heterogeneous and unstructured information sources, such as job offers and curricula to make even better job offers and job skills classification than today.  

Expected benefits 

The tool has produced and is expected to cause further benefits: 

• Enhanced quality of public service, particularly in guiding job counsellors, job seekers and employers; 

• This new algorithm is helping the organisation to update its employability model, wherein job skills demanded and supplied in the region are now at the centre of the matching process. The  goal is to improve workers’ adaptability and self-awareness to speed up their comeback to work, hence diminishing their unemployment spell. This skill-based model powered by AI provides career counsellors with more comprehensive information to improve their recommendations. 

• Algorithms allow job counsellors to have a more comprehensive view of job seekers’ profiles leading to more efficient career path suggestions and tailored insertion pathways;

• Improved transparency and policy evaluation mechanisms, offering live data for policy makers to study active employment policies in the region. In fact, the solution also aimed to provide better monitoring and evaluation of Active Employment Policies to improve their effectiveness and efficiency; 

• The deployment of generative AI will allow to have even more tailored results and ad-hoc virtual assistants will assist directly job seekers; 

• Scalability and replicability of the solution is another potential benefit, as it uses occupational standard codes (CNO) for job skills and offers and it also relies on could architectures. 

Main challenges 

Despite the many benefits, some challenges are expected to be faced in the future: 

• Ensuring compliance with the EU AI Act provisions might be a challenging task, given the novelty of the legislation and the ongoing learning it requires. If, for instance, this system might be categorised as of High-Risk, additional legal requirements (such as record-keeping, transparent documentation, among others) might be requested in the future, posing organisational and technical challenges to the organisation; 

• AI hallucinations and biases in the future might pose some challenges to the deployment of more advanced generative AI models to support the platform. 

For more information on the solution, please contact Boquete Sanluis (santiago.boquete.sanluis@xunta.gal)