In the 21st century, individuals are increasingly making career transitions (Sullivan & Ariss, 2021), necessitating a growing number of educational and occupational decisions. In fact, making career decisions is regarded as an important life skill and refers to key competencies in the context of lifelong learning (Neary et al., 2015; Sultana, 2012). It has been argued that learning analytics (LA) can potentially support humans in lifelong learning and associated transitions (Poquet et al., 2021). So far, LA tools extensively utilizing data about learners and their environments (Ifenthaler & Yau, 2020; Papamitsiou & Economides, 2014) have been mostly used to support learning and guide students within higher education institutions (Charleer et al., 2018; de Laet et al., 2020; Guerra et al., 2019, 2020; Gutierrez et al., 2020; Iatrellis et al., 2017). With the advances of AI, LA tools can better collect and analyze data and ultimately produce more accurate results (Zawacki-Richter et al., 2019). In other words, users might be provided with personalized and adaptive career guidance based on data gathered about them and their environments (Gati & Asulin-Peretz, 2011; Guleria & Sood, 2022; Kettunen & Sampson, 2019).

However, we highlight two important gaps in the studies regarding LA and AI applications for career guidance. First, the users’ perspectives on these technologies are neglected. Second, the role of AI and LA technologies remains unclear in the context of career decisions and transitions. In response to these gaps, we investigate multiple aspects of AI-enhanced LA tool for career guidance from users’ perspective. This is done to uncover possible contributions of AI-enhanced LA to career guidance. We aim to better understand how can AI-enhanced LA tool help move through life transitions, what kind of difficulties users experience and what life circumstances prompt the use of such tool. This work contributes to better understanding user perspectives, which are critical in the design and use of LA tools (Buckingham Shum et al., 2019; Knight et al., 2016) as well as career guidance services (Plant & Haug, 2018).

Namely, this article examines benefits and challenges perceived by the users, user described use circumstances and perceived support for making career decisions. We maintain that technology acceptance model (TAM) is a suitable framework for understanding the role of AI-enhanced LA applications in career guidance. This is because the essential constructs of TAM cover perceived usefulness, ease of use and intention to use a particular technology, which matches our study aims. Further, we regard making career decision as a process of individuals considering several education alternatives and choosing one of them at certain points of their lives (Kulcsar et al., 2020). In this regard, technology could support career decisions by facilitating a systematic approach to this process. To understand how AI-enhanced LA tool supported users career decision, we relied on career decision making (CDM) model. Therefore, we drew upon two distinct theoretical frameworks, namely TAM and CDM. The novelty of this study lies in its explanations from the lens of both information-technology-based constructs (i.e., TAM) and career decision making process (i.e., CDM).

1 Theoretical framework

1.1 Technology acceptance model

Proposed and developed by Davis (1989) the TAM addresses users’ acceptance and use of a particular technological tool. In the TAM, actual technology use is determined by individuals’ behavioral intention to use a technology. Behavioral intention is affected by attitude towards that technology, and by the direct and indirect influences of perceived usefulness and perceived ease of use. Perceived ease of use refers to a user’s belief indicating the use of technology will be free of effort (Davis, 1989). Perceived usefulness is defined as the extent to which a user believes that using a technology will increase or improve their job performance and productivity (Davis, 1989). Thus, people tend to use or not to use a particular technology to the extent that they believe it will enhance their job performance. Many studies in the relevant literature provided evidence of the impact of perceived ease of use and perceived usefulness on the attitudes towards use and behavioral intention (King & He, 2006; Legris et al., 2003). Behavioral intentions are defined as motivational factors influencing behavior. Generally, stronger intention predicts greater likelihood of specified behavior (Ajzen, 1991). Research on the acceptance of information and communication technology highlights the importance of behavioral intentions in predicting actual technology use (Ajzen, 2001; Turner et al., 2010; Venkatesh et al., 2003). Several different extensions and modifications have been suggested for the original TAM, by proposing cultural and social influences for behavior intentions, as well as suggesting multiple antecedents for perceived usefulness and ease of use (Hornbæk & Hertzum, 2017; Venkatesh et al., 2003). However, perceived usefulness, ease of use and behaviour intentions remain the core of the model.

1.2 Career guidance and career decision-making model

Increasingly complex education and work environment requires individuals to make multiple career decisions and experience increasing number of career transitions during one’s life (Sullivan & Ariss, 2021). These decisions have important and lasting consequences. For instance, educational decisions were reported to affect access to further education (Thompson, 2005), access to job opportunities and employment status (Schmillen, 2019; Iannelli & Smyth, 2017; Thompson, 2005), income (Chevalier, 2011; Sullivan et al., 2018; Kim et al., 2015) and socioeconomic status (Wolniak et al., 2008).

To provide support for career decisions, career guidance draws on two key cornerstones –the content of the decisions and the process of the decisions (Gati & Kulcsar, 2021). While content is usually defined as the skills, competencies, and interests of an individual, the process orientation is concerned with the ways career decisions are made (Gati & Kulcsar, 2021). A better-quality career decision-making process is expected to lead to better-quality decisions and more favorable outcomes for an individual.

Several career decision models were suggested to facilitate the career decision-making process systematically. For instance, the Career Decision Making (CDM) model (Gati & Tal, 2008; Gati & Asher, 2001); Communication, Analysis, Synthesis, Valuing, and Execution (CASVE) cycle (Osborn et al., 2020), and the dynamic career decision-making model (Esbroeck et al., 2005). These career decision-making models are theoretically grounded, practice-oriented approaches trying to offer an optimal strategy for making career decisions by adapting decision theory to the vocational domain (for a more comprehensive overview of prescriptive models see Gati and Kulcsar (2021). They outline a pragmatic framework for a systematic career decision-making process (Gati & Tal, 2008), which could be used in different career transitions. Their main advantage is the balance between the emphasis on the utility of er choice and human limitations in decision making (Gati & Kulcsar, 2021).

While each of these models propose different steps for the individual to take, the common features are availability and combination of information about self and the world of work (1), investigation of this information (2), and prioritization of several options (3). CDM provides the most detailed description of these essential elements of information provision and processing; therefore, it might be utilized as a framework for career guidance technology development and evaluation. In this model, career alternatives are systematically narrowed down by combining information on individual and career alternatives and it consists of three elements: prescreening, in-depth exploration, and choice (Gati & Asher, 2001). The CDM is visualized in Fig. 1. The figure also visualizes the information needed for the process – information about self and information about careers, which are merged in the prescreening phase.

Fig. 1
figure 1

The CDM and the information needed to make the decision

Full size image

Prescreening of many alternatives is based on individual preferences, and various aspects of careers can be taken into consideration, such as expected income or required length of training. The aspects can be considered important by the individual in both their positive and negative values (Gati & Tal, 2008). The goal of this stage is to identify a smaller, manageable set of career alternatives. Gathering information about important and favorable aspects of careers and personal preferences is an important part of this phase. In-depth exploration consists of a thorough examination and comparison of the identified alternatives. The goal of this stage is to locate a small number of suitable alternatives based on comprehensive information. Finally, the user makes a choice based on the results of an in-depth examination of alternatives and can include several best-suited options. Individuals are encouraged to use the career decision-making model flexibly and to move back and forth between the stages, depending on their needs (Gati & Tal, 2008).

2 The use of Artificial Intelligence (AI) and Learning Analytics (LA) in career guidance

The widespread use of big data technologies allows to collect huge amounts of data from learners and their environments (Gašević et al., 2016). For the last decade, this data has been considered an asset leading to the emergence of the field of LA (Siemens, 2013). Accordingly, LA refers to the measurement, collection, analysis, and reporting of data about learners and their contexts (Siemens, 2013). LA is increasingly interconnected with AI and its subfields (Siemens, 2013), such as machine learning and natural language processing which are among the most prevalent sub-fields of AI (Celik et al., 2022). In this process, LA can rely on AI methods to better collect and analyze learners’ data, and to produce more accurate results (Ez-Zaouia & Lavoué, 2017; Zawacki-Richter et al., 2019). Meanwhile, AI is defined as “computers which perform cognitive tasks, usually associated with human minds, particularly learning and problem-solving” (Baker & Smith, 2019, p. 10). AI is not a single technology, rather it is regarded as a computer field with a variety of sub-fields. The most notable sub-fields are machine learning, natural language processing, and deep learning (Zawacki-Richter et al., 2019). In this regard, LA tools powered by AI might capture data from the learners and their environment and process it ultimately producing personal and environment-specific outputs. More specifically LA tools have the potential to present and visualize collected personal learning data about various aspects of a user’s life, such as habits and interests, further enabling individuals to improve self-knowledge through visual support for revisiting and investigating their personal history (Verbert et al., 2013, 2020). Utilizing AI for career guidance purposes has generated high hopes in the field (Brown et al., 2020) and contributed to envisioning far reaching and elaborated guidance systems of students (Khare et al., 2018; Westman et al., 2021;). AI is recognized for its potential to support human decision making, including career decisions, and increase the quality of career guidance services (Gati & Kulcsar, 2021).

Recently, LA tools have made their first steps in supporting academic advising of students (de Laet et al., 2020; Guerra et al., 2020; Gutierrez et al., 2020). Functionalities such as identifying future learning opportunities and creating personalized recommendations are often exploited for academic advising purposes, for instance by helping students choose programs and majors, select and schedule courses, make long term course plans, and to investigate their academic paths (Solis, 2016; Iatrellis et al., 2017; Guerra et al., 2019; Guleria & Sood, 2022). Similarly, career decisions can be supported by technology keeping in mind specific nature and context of the decisions such as the short-and long-term consequences, number of available alternatives, familiarity with the alternatives, and the central position of an individual to make the decision (Gati & Kulcsar, 2021).

3 The aim of the study

First, in this study we seek to explore advantages and challenges of the AI-enhanced LA tool in the context of career transitions and to provide user-centered evidence for further development. This is done through the perspective of the TAM framework which is employed as a lens for the analysis. Secondly, the study aims to investigate the perceived support the AI-enhanced LA tool provides for users’ career decision-making process. This is analyzed though the CDM framework, employing CDM as a representation of the decision-making process. More generally, the study contributes to better understanding the use and adoption of AI-based tools for career guidance and for employing emerging technologies to direct human life paths. The study focuses on the user’s perspective, which plays a key role in the development of sustainable services in LA (Buckingham Shum et al., 2019; Knight et al., 2016) and career guidance (Plant & Haug, 2018).

4 Contributions of the study

This study offers several contributions. First, it expands the understanding of LA use beyond institutional boundaries by focusing on the perspective of individual concerned with career transitions. Previous research emphasized LA tool use in higher education (Tsai et al., 2020) leaving the perspective of lifelong learning unaccounted. It has also been recognized that little is known about the type of support needed to help individuals in transitional life situation (Poquet et al., 2021). In response to this gap, the current study focusses on the users described support experiences in the context of educational transition.

Secondly, the study attempts to connect guidance technology with the theoretical guidance concepts by focusing on the decision-making process and how this process is supported by the AI-enhanced LA tool. LA tools so far have been explored in connection to learning processes and constructs such as motivation, engagement, and academic performance (Karaoglan Yilmaz & Yilmaz, 2021; Silvola et al., 2021; Kew & Tasir, 2022). Different from previous studies, in this study we address the career decision-making process, which might be regarded as another important contribution.

Thirdly, the study dives deep into user perception analysis by adopting a qualitative approach. Previous research on LA use and acceptance generally measured perceived usefulness, ease of use, and intention to use factors with pre-determined items using scales (Herodotou et al., 2019; Ali et al., 2013). These studies suggested important and valuable empirical results on LA tool acceptance based on the TAM framework (e.g., Ali et al., 2013). However, less is known about what constitutes TAM constructs such as usefulness in the context of guidance. For example, the perceived usefulness of AI-based tools may appear differently in a context of career transitions as compared to planning and monitoring progress of the studies. Thus, here we employed qualitative approaches to explore perceived usefulness, ease of use and intention to use in the context of career transitions.

And lastly, this research provides empirical evidence on the user needs for outlining further AI-enhanced LA tools development in the field of guidance. Indeed, users’ needs, and expectations are the key for developing LA (Silvola et al., 2021), which are also critical in the acceptance of emerging technologies, particularly AI-enhanced career guidance tools.

To achieve our aims, we addressed the three following research questions.

4.1 Research questions (RQs)

RQ1:

How did users describe advantages, challenges, and use purposes of the AI-enhanced LA tool for supporting their career decisions based on TAM framework?

RQ2:

What are users’ needs for further AI-enhanced LA career guidance tool development?

RQ3:

How did the AI-enhanced LA tool support the users’ career decision-making process based on CDM framework?

5 Materials and methods

5.1 Research design and participants

This research utilized qualitative research design and content analysis of open-ended survey collected from 106 users (Table 1) of the AI-enhanced LA tool. The users tried out the AI-enhanced LA career guidance tool in career guidance sessions in three vocational education and training (VET) institutions in central and northern Finland in 2019. The users were recruited by guidance counselors in these schools since all users needed career guidance and considered starting or changing education in VET or applied science universities. As the demographic information was voluntary for the participants to fill, some of them chose not to provide full demographic details, as reflected in Table 1.

Table 1 Overview of user demographic characteristics
Full size table

5.2 Instruments

User perceptions about the AI-enhanced LA career guidance tool were gathered using an online open-ended questionnaire. Three open-ended questions were asked to address use intentions and the support from the AI-enhanced LA guidance tool: (1) In what situations would you use the LA tool? (2) Could the LA tool help you apply for an education or a job? Why yes or why not? (3) How could you benefit from the information you got using the LA tool? To gather user needs for further development, we asked two additional open-ended questions: (4) What feedback could you give about the tool? (5) What functionalities would you want the LA tool to have? Questions 1–3 were used to answer RQs 1 and 3. Questions 4–5 were used to answer RQ 2.

5.3 The AI-enhanced LA career guidance tool

We tested the AI-enhanced LA career guidance tool developed in the Competence Leap project. The AI-enhanced LA career guidance tool utilized national-level Finnish education registry data and recommends relevant program choices from vocational education and training (VET) or applied science universities in Finland. The recommendation is formed based on a natural language processing technique (NLP), subfield of AI and similarity of analyzed contents (development details at: https://github.com/Opetushallitus/compleap). From the user perspective, the LA career guidance tool is deployed in the following process: Evaluating users’ previous education (Fig. 2), identifying users ‘current interests (Fig. 3), and browsing recommended study programs (Fig. 4).

Fig. 2
figure 2

Previous education section in the AI-enhanced LA career guidance tool

Full size image
Fig. 3
figure 3

Interests section in the AI-enhanced LA career guidance tool

Full size image
Fig. 4
figure 4

Recommended study programs and study places in the AI-enhanced LA career guidance tool

Full size image

Information on previously completed education (Fig. 2) is obtained from the Finnish national repository for study records for multiple levels of education entitled KOSKI (Finnish National Agency for Education, n.d., a). The users are presented with their previous education divided into study units. Study units are the parts of VET degrees describing the competence based on the definitions of the Finnish National Agency for Education (Finnish National Agency for Education, n.d., b). Due to educational repositories’ limitations imposed by Finnish legislation (Act on the National Registers of Education Records, Qualifications and Degrees 884/2017 4§) it is not possible to use user’s personal data for a proof of concept LA career guidance tool. Thus, data on users’ personal previous educational data was inaccessible and was substituted by several educational backgrounds. Further, users were asked to select the most relevant education background from the list to identify their likes and dislikes with the specific study program (Fig. 2).

The expression of current interests (Fig. 3) complements previous education information and was developed using a Finnish service for the publication and utilization of vocabularies, ontologies, and classification of general education interest areas (https://finto.fi/en/about). Approximately 150 objects were grouped under 13 top-level concepts. For example, sculpture, dance, painting, film, and literature were all grouped under the art and culture concept. This allowed the users to elaborate on their interests by selecting the most motivating topics by choosing from three to ten interests.

Recommended study programs (Fig. 4) tied together the previous two components by automating prescreening and offering education recommendations, which consist of a study program and study place in a concrete location and institution. The user’s previous education and interests were compared with educational descriptions of VET and applied science universities’ programs to find matching content. Available study programs were then presented based on the data in Studyinfo.fi, an official website containing all the information about study programs leading to a degree in Finland. The users could additionally filter study programs by region and were provided with a list of academic institutions providing recommended study programs.

5.4 Data analysis

To answer the research questions, content analysis (Bengtsson, 2016) was used on the users’ responses to five open-ended questions considering both qualitative and quantitative aspects of the data. First, two researchers analyzed the data, creating and refining the sub-categories using inductive approach, where themes arose from the data (Bengtsson, 2016). The interrater agreement as expressed in agreement percentage (Gisev et al., 2013) for the identified themes was 84.84%, as calculated for 10% of the data. A unique meaningful thought was the unit of analysis, and the coding was not limited in length.

Then, to answer first and second research questions, we analyzed user-perceived advantages, difficulties, use intentions, and further development needs regarding the AI-enhanced LA career guidance tool. To achieve this, the first author grouped sub-categories into themes and assigned them to corresponding theoretical constructs. The definitions and operationalizations of the employed TAM and user need constructs are presented in Table 2. A similar approach has been previously utilized by Ifenthaler and Volker (2013), who investigated TAM constructs qualitatively in the study of educational technology.

Table 2 Operationalization of the TAM constructs and user needs in relation to the AI-enhanced LA guidance tool
Full size table

Next, to answer the third research question, we analyzed the support the AI-enhanced LA guidance tool provided for career decisions-making process. To achieve this, the users’ 192 answers previously identified in perceived usefulness category were selected for further analysis and categorized according to the CDM framework by the first author. Definitions and operationalization of CDM constructs are presented in Table 3. The results are presented in Fig. 5 visualizing the quantitative outcomes of the analysis. The intensity of darkness of the blue color refers to the support the AI-enhanced LA tool provided to each CDM element. The darker blue color indicated that CDM element is supported stronger by the AI-enhanced LA tool, i.e., as expressed by frequency of the users’ answers.

Table 3 Operationalization of the CDM constructs in relation to the AI-enhanced LA guidance tool
Full size table

6 Results

6.1 Advantages, challenges and use purposes of the AI-enhanced LA tool

The first research question investigated how users describe advantages, challenges, and use purposes of the AI-enhanced LA tool for career guidance based on TAM framework. Results clustered according to the TAM framework concepts are presented in Table 4.

Table 4 AI-enhanced LA tool’s advantages, challenges and use purposes based on TAM from user perspective
Full size table

Users described the AI-enhanced LA tool for guidance as beneficial for five main purposes: information provision, research and analysis, diversification of ideas on career paths, providing direction and decision support, and self-reflection. Category of information provisions contains a variety of answers related to education, jobs, and career information. The category of research analysis contains answers relating to various activities done with provided information, such as searching, mapping and comparison of information, as well as being able to think about provided programs and education choices in general. Category of diversification of ideas on career paths holds answers related to diversity of provided information and career recommendations. Users described getting variety of alternative education suggestions and some claimed getting new ideas they have not though about before. Category of providing direction and decision support reflects user expectations related AI-enhanced LA tool being useful for applying to education programs as well as becoming more confident, getting assurance to make the decisions and take further steps. Finally, the category of self-reflection contained answers related to the AI-enhanced LA tool being useful to learn about own interest and skills.

Further, the challenges of using the AI-enhanced LA tool were related to usability difficulties and are described in the ease-of-use category. The category contained users’ visual and cognitive challenges of using the tool. For instance, users expressed various complaints and suggestions regarding the need to improve the tool’s graphical design and information presentation as well as the need for instructions and support during use.

Finally, users intended to use the AI-enhanced LA tool in various life situations characterized by uncertainty e.g., unemployment, dropping out of studies, choosing new field of study, or searching for further education possibilities.

6.2 Users’ needs for further AI-enhanced LA career guidance tool development

The second research question focused on the users’ needs for further development of the LA career guidance tool. Results for this investigation are presented in Table 5.

Table 5 Users described needs for further AI-enhanced LA tool development
Full size table

The users expressed number of suggestions for further development which were categories into three main themes: diversification of ideas on career paths, personalization and self-reflection, and more information. Strong emphasis was placed on the need for diversification and personalization by asking for more diverse educational options and making the recommendations more personalized and tailored to the individual personality, interests, and life situations.

6.3 The support for career decision from AI-enhanced LA tool

The third research question focused on the support AI-enhanced LA tool provided to the users’ decision-making process. The results indicate that not all steps of the decision-making process were supported equally. Figure 5 visualizes the quantitative results of content analysis with intense blue color indicating more accounts of support reported by the users.

Fig. 5
figure 5

Support for users’ decision-making process provided by the AI-enhanced LA tool

Full size image

In-depth exploration element of career decision making process was the best supported, as many users answers (f = 91) concentrated on various exploration activities enabled by the AI-enhanced LA guidance tool. Choice element, although less, was also supported by indications of users getting encouragements and confidence to make decisions and to apply to study programs (f = 54). Contrarily, information about careers (f = 33) and information about self (f = 16), ware the least supported elements. Especially small number of accounts were found regarding the AI-enhanced LA tool providing users with information useful for self-reflection. Prescreening phase was completed by the AI-enhanced LA tool and did not require any user effort nor awareness; therefore, it is not included in the Fig. 5.

7 Discussion

The current study described users’ perceptions about the AI-enhanced LA tool for career guidance. The dimensions of user perception were determined based on the TAM framework, which is suitable for investigating the acceptance and usage of emerging technologies (Oyman et al., 2022). Scherer and Teo (2019) also highlighted that most educational technology acceptance studies have failed to focus on specific technology. Without providing concrete experience to users, it might be challenging to reveal their reasons for acceptance or rejection of technology and to provide sound insights into practice (Islamoglu et al., 2021; Tao et al., 2022). In response to this gap, we aimed to explore users’ thoughts after they engaged with AI-enhanced LA tool.

As per our first research question, five main advantages of the AI-enhanced LA tool for guidance were identified, namely: provision of career information; research and analysis of provided information; diversification of ideas on possible career options; providing direction and decision support; and self-reflection. These findings provide new knowledge in the context of technology supported guidance. Previous research identified information provision, communication support and automation of interactions as important roles for guidance technology (Hooley et al., 2015). This study expands on existing understanding by describing broader scope of ways technology can contribute and support individuals. The results help to determine types of support AI and LA tools might provide for individuals in a context of lifelong learning and associated transitions, which has been previously identified gap in research (Poquet et al., 2021). User described types of support are discussed further.

First, quality information on educational opportunities, careers, and job market situation were essential for the users of the AI-enhanced LA guidance tool. This finding stays consistent with previous research which identified information provision as a key role of technology for guidance (Hooley, 2015). In addition to this, our study shows that users emphasize not only being passive receivers of the information but actively engaging with the provided information by exploring it further. This finding provides support for career guidance approaches emphasizing active learning during the process of career guidance (Krumboltz, 2009). Accordingly, AI and LA tool should create possibility to actively examine the provided information with user inquiries. Diversification of career paths by providing different educational possibilities emerged as another key advantage from the users’ perspective. In line with career guidance literature, diversification of career paths is a key element in effective career guidance (Musset & Kurekova, 2018) and an objective for social justice (Skohvus, 2016). Our study adds to this knowledge by showing that diversification of ideas about career possibilities is an important user expectation. Thus, LA career guidance tools could respond to this by helping users to re-evaluate their interests in different contexts. Further, the users recognized the AI-enhanced LA tool as beneficial for providing direction and decision support. Individuals might be rather confused and indecisive when they are about to make an important decision on their academic or work life (Selder, 1989; Schlossberg, 2007). At such moments, supporting assets that strengthen and encourage individual are important (Schlossberg, 2007). Our results indicate that users perceive AI-enhanced LA guidance tool to be such a supportive asset during life transitions and described getting assurance and confidence from the tool to make necessary decisions. And lastly, users’ self-reflection was identified as another possible benefit from the AI-enhanced LA guidance tool. For instance, users claimed that the AI-enhanced LA tool has helped them identify their interest and to understand them better in work life context. Support for self-reflection has been also previously determined as a key goal of LA in education (Greller & Drachsler, 2012).

Further, users described some challenges about using the AI-enhanced LA tool, which were evaluated in the light of ease-of-use concept from the TAM framework. Several use difficulties were identified such as graphical and structural imperfections that made use complicated. Users also expressed the need for support e.g., instructions and human guidance. Prior research has recognized that users need to get familiar with the interface when they interact with an emerging technology (Dutta et al., 2022). A considerable number of usability related answers indicates a possible gap in digital career literacy, which is a previously recognized concern (Hooley et al., 2015). Thus, clarity and simplicity of use must be emphasized in the development of digital guidance tools.

Next, users named several life situations where they intend to use the AI-enhanced LA tool. The use intentions focused on two main areas - education and work, which reflects the domains of career concept (European Lifelong Guidance Policy Network, 2014). The common features of the identified life circumstances were uncertainty and the need to make decisions among multiple options, which corresponds to the previously recognized characteristics of transitional life situations (Selder, 1989; Schlossberg, 2007). This finding further indicates that the AI-enhanced LA guidance tool was perceived as a supportive asset in career transitions and goes to shows vulnerable position of prospective users.

Our second research question focused on the user needs in connection to further AI-enhanced LA tool development. Three main needs were identified: the need for further diversification of ideas on career paths, the need for better personalization and self-reflection, and the need for further information. The identified needs directly correspond to previously identified benefits of using the AI-enhanced LA tool for guidance, highlighting the importance of these themes to the users. For instance, although diversity of recommendations was already well evaluated in the current version of the AI-enhanced LA tool, a variety of education alternatives was also stressed for further development, suggesting that diversity of recommendations is a key concept to address in guidance tools development. Supporting results were obtained by Gati et al. (2003) where variety and length of the recommended list of occupations positively correlated with users’ satisfaction (Gati et al., 2003). Integration of diverse education and work life data sources could be a way to address users’ need for diversity and provide information about study programs at various levels of education locally and abroad. Similarly, information about personal skills, competencies and interests were insufficient for some users. Current AI-enhanced LA tool limitation in this domain can be explained by limited access to the users’ data and further efforts should be taken to facilitate self-reflection if AI and LA tools strive to provide successful career decision making support. Users expressed this need by referring to more personalization, such as questions about personality and interests. Such information could lead to more accurate career recommendations and provide opportunities for improved self-reflection for the user. These findings correspond to previous research where tailored support was seen as a key opportunity to improve career guidance service (Donald et al., 2018).

Our third research question concerned the support AI-enhanced LA guidance tool provided for the users’ career decision-making process. The results showed that investigated process elements were supported unequally. On one hand the AI-enhanced LA tool helped the users to investigate a manageable number of educational options, provided direction and assurance needed to make decisions. On the other hand, important contextual information such as information about self and the world of work was lacking. Previous research has identified lack of self-knowledge as a significant struggle in making career decisions (Gati et al., 1996); therefore, self-reflection of users should be better supported by tool’s emphasis on future goals, professionally relevant skills, and interests. Another important implication of this funding is that the support of the AI-enhanced LA tool might lead to changes in the recommended decision-making process from the user perspective. The information analysis process and the information used in the process by the AI-enhanced LA guidance tool are hidden from the view of the users and the users are presented with the final list of recommendations. This might interfere with the purpose of users understanding and learning how to make career decisions – one of the key goals of using career decision-making models in career guidance (Kulcsar et al., 2020). Previous research has shown that user trust in provided career recommendations is high (Gedrimiene et al., 2023) and emphasized the importance of users being able to critically evaluate career guidance from AI tools (Gati & Kulcsar, 2021). Therefore, we suggest putting stronger emphasis on the user self-reflection and enablement of different user inquiries in the development of the AI-enhanced LA guidance tools.

7.1 Limitations and further research

This research focused on qualitatively investigating user perceived benefits, challenges, use intentions and support for decision-making process from the AI-enhanced LA tool for career guidance. Further research could for example investigate to what extent AI tools diversify user ideas about possible career paths and how could they be used successfully utilized for various contexts. It is also important to acknowledge that digital tools are not used, and career choices are not made in isolation but in a greater socioeconomical context of individual’s life. Although this article takes an instrumental approach to technology and an optimistic view to career decisions, limitations and support for career choices stem also from other sources such as social environment, individual aptitudes, finance, and health. Moreover, various forms of exclusion and biases can lead to formation of certain career paths for individuals in society and also make their way into the digital guidance world (Staunton, 2019; Elias et al., 2017). These topics are, however, beyond the scope of this article. Similarly, constrains of current personal data legislation in Finland resulted in utilization of limited number of user profiles in the AI-enhanced LA guidance tool and could have affected the user responses. A rather small sample and varying demographic characteristics limit generalization of our findings.

8 Conclusions

This study is the first attempt to provide in-depth results of users’ perceptions of the AI-enhanced LA tool for career guidance. We conclude that the users intended to use the investigated AI-enhanced LA tool as a supportive asset for career transitions characterized by uncertainty. The tool showed promise to facilitate these transitions by offering five main advantages as described by users. Specifically: providing career information, facilitating research and analysis of this information, diversifying ideas on career paths, self-reflection, and providing direction and decision support. These findings suggest possible directions of the AI-based technology development to provide value for users in the context of lifelong learning and associated transitions. And lastly, the use of the AI-enhanced LA tool for guidance led to deviation from theoretical career decision-making process which might counteract with original purpose of users learning to make career decisions.