Bridging the AI competence gap: Mapping skills and training needs in SMEs of technology and creative sectors

Abstract: Small and medium-sized enterprises increasingly express interest in artificial intelligence, yet research still lacks a detailed understanding of their existing competencies and concrete training needs, particularly in technology and creative sectors. This study addresses this gap by examining how these enterprises understand, apply, and develop artificial intelligence skills, and what support they require. The study draws on twenty‑five thematic interviews and qualitative content analysis, guided by a multidimensional competence framework. The findings show significant variation in literacy, application skills, ethical awareness, data practices, and strategic integration. Both sectors demonstrate strong motivation to learn but struggle with identifying feasible use cases, evaluating risks, and accessing practical training. The study concludes that enterprises need holistic and hands‑on learning pathways that combine foundational understanding, domain‑specific application, and responsible use. The results inform the design of targeted training, organisational support models, and future competence assessment tools, while providing a baseline for further research.

Keywords: Artificial intelligence, competence, small and medium-sized enterprises, technology industry, creative industry, training

Authors:

Miia Kosonen, corresponding author, Research Manager, South-Eastern Finland University of Applied Sciences, Patteristonkatu 3, 50100 Mikkeli, miia.kosonen@xamk.fi

Kati Saltiola, Project Manager, South-Eastern Finland University of Applied Sciences, Patteristonkatu 3, 50100 Mikkeli, kati.saltiola@xamk.fi

1. Introduction

Artificial Intelligence (AI) has profound effects on firms across sectors, as well as society and working life overall. Indeed, AI could represent a more fundamental shift — or disruption — than the spread of the internet did in the 1990s. While large enterprises have dominated AI development and adoption due to their resources and data availability (Oldemeyer et al. 2024), small and medium-sized enterprises (SMEs) should not be overlooked in this transformation. On a societal level, SMEs represent a large share of economic activity and hold a major AI-related productivity potential.

In Finland, 24 % of firms applied AI in 2024, but among small firms, the share was only 17 %. According to a recent EK report (Elinkeinoelämän Keskusliitto 2025), 68 % of firms cite lack of competencies as the main obstacle. SMEs commonly face barriers such as insufficient knowledge, high costs, resistance to change, and inadequate infrastructure (Oldemeyer et al. 2024; Mohd Rasdi & Umar Baki 2025).

To address these challenges, prior research emphasizes the need for tailored and context-specific support practices and training to enable AI adoption in SMEs (Kramarenko 2025; Dinh et al. 2025; Sánchez et al. 2025). For local businesses, such support is crucial for regional renewal, competitiveness, and employment. Supporting management in strategic decision-making, strengthening AI skills among personnel, and addressing regulatory and ethical issues are essential. However, designing effective training requires a clear understanding of current competencies and perceived needs in the target industries.

This study responds to these gaps by:

1. Mapping current AI-related competencies and training needs;
2. Focusing on SMEs in technology and creative sectors;
3. Informing the development of tailored support practices and training.

Our research questions (RQs) are:

• RQ1. Which types of AI competencies can be identified from SMEs in the technology and creative sectors?
• RQ2. What are the key AI competence and training needs among SMEs in the technology and creative sectors?

To answer these research questions, we conducted an empirical study relying on qualitative data from 25 company interviews. Overall, the findings provide an evidence base for designing training and organizational support for sustainable and strategic AI use. This is critical as AI reshapes technical processes, creative production, management practices, and workforce skill structures. Theoretically, the study extends the discussion on AI competencies in SMEs by applying the novel AIOK framework as a lens for identifying both shared and sector-specific patterns across technology and creative firms. Practically, the findings offer a grounded basis for designing more targeted training, coaching, and support measures that reflect SMEs’ actual competence gaps, everyday use cases, and responsible AI adoption needs.

This article is organized as follows. We start by reviewing current literature on AI use, readiness, and competencies in SMEs. The literature review also introduces the AIOK framework applied in the current study. We then describe the research design and methodology in Section 3. The results are presented in Section 4. In the concluding Section 5, we summarize our most important findings and contributions, accompanied by potential themes for further research.

2. Literature review

As the use of AI becomes more common in everyday business operations, it is increasingly important to clarify what AI is and what competence means. These concepts form the foundation for understanding how technological change is reshaping the capabilities and skills required in organizations. In our study, this framing is essential for examining the current AI‑related competencies among SMEs and identifying the areas where additional competence development and training are still needed.

2.1 Definitions of artificial intelligence and competence in organizations

Artificial Intelligence refers to computational systems capable of performing functions typically associated with human intelligence, such as learning, reasoning, planning, and creating. AI systems can perceive and interpret their environment, process data, solve defined problems, and act toward specified goals. Instead of merely conducting automated tasks on a repetitive basis, a key attribute of AI is related to its (machine) learning capacity: to be labelled as AI, systems should be able to adjust their behavior based on prior outcomes and operate at least with a certain degree of autonomy. As a rapidly advancing technological domain, AI increasingly shapes how digital systems analyze information, support decision‑making, and enable new forms of automation (European Parliament 2023).

Competence, in turn, refers to a combination of knowledge, experience, skills, abilities, and other attributes that enable individuals to perform effectively in their roles and tasks (Weinert 2001). In organizations, competence management involves systematically identifying and assessing current capabilities and comparing them with future needs. Making competence visible and measurable helps clarify strengths, gaps, and development potential across the workforce. Structured practices such as competence assessments and role‑specific profiles support targeted development and ensure that capability building aligns with organizational goals. Rather than being a standalone HR activity, competence management functions as a continuous, data‑informed process that supports both organizational renewal and employees’ long‑term growth (C&Q Systems 2025).

2.2 AI adoption and competence challenges in SMEs

Existing research suggests that the main barriers for SMEs are less about access to models or tools and more about organizational and human competencies: understanding AI, preparing data, integrating AI into processes, and managing ethical and legal risks. Recent reviews emphasize that Artificial Intelligence (AI) is increasingly recognized as a key enabler of productivity and competitiveness, but AI adoption in SMEs lags behind large firms. Indeed, in their meta-analysis, Oldemeyer et al. (2025) identify 27 distinct challenges to AI implementation in SMEs. The most frequently cited barrier is a lack of knowledge and skills related to AI, mentioned in 35 of the analyzed articles. Other recurring obstacles include unclear return on investment (ROI), limited financial resources, insufficient digital infrastructure, and low data quality (Oldemeyer et al. 2025).

A qualitative multi-case study by Mohd Rasdi and Umar Baki (2025) highlights how AI integration in SMEs is constrained by both internal and external competence gaps. Internally, SMEs struggle with change management, limited awareness of AI, fear of job loss, and insufficient technical and data skills. Externally, they face workforce skills shortages, talent competition, complex regulatory and ethical requirements, and difficulties in accessing funding and infrastructure. SMEs must develop distinctive internal capabilities — such as technical competence, data analytics skills, and innovative management practices — while simultaneously navigating macro-level political, economic, and legal conditions around AI (Mohd Rasdi & Umar Baki 2025).

As mentioned in the Introduction, Finnish evidence so far points in a similar direction. Less than 17 % of Finnish SMEs currently use AI, although overall AI use places Finland among the leading EU countries. SMEs are generally positive towards AI but explicitly request support in recognizing suitable use cases and building the necessary skills and data foundations. Companies that have started to use AI report benefits such as time savings, improved decision making, enhanced service quality, and new business opportunities, but also describe practical challenges related to staff competence and data quality (Elinkeinoelämän Keskusliitto 2025).

2.3 Organizational AI readiness and competence

Several studies conceptualize “AI readiness” as a multidimensional construct that combines organizational capabilities, resources, and skills. For instance, Jöhnk et al. (2021) developed an organizational AI readiness framework based on interviews with 25 AI experts and a synthesis of adoption literature. The authors argue that organizations must continuously assess and develop their AI readiness, and that AI awareness, ethics, and cross-functional collaboration are particularly critical for realizing AI’s business value. They identify five categories with 18 factors: 

• Strategic alignment (AI business potential, customer AI readiness, top management support, AI–process fit, data-driven decision-making)
• Resources (financial budget, AI-related personnel, IT infrastructure)
• Knowledge (AI awareness, upskilling, AI ethics)
• Culture (innovativeness, collaborative work, change management) 
• Data (data availability, quality, accessibility, and data flow) (Jöhnk et al. 2021).

Building on this work, Naheed et al. (2025) propose a preliminary multidimensional AI readiness assessment model for SMEs. Their Technology-Organization-Environment-Human (TOEH) framework captures technological readiness (IT infrastructure, data availability and access, cybersecurity), organizational readiness (organizational structure and culture, leadership support, operational integration, information management), environmental readiness (competitors, regulation, government support, market, suppliers and collaborators), and a dedicated human dimension that includes skills, motivation, and the ability to implement change. The model is designed as a practical tool that assigns each element a readiness state (informal, struggling, approaching or desirable) and produces an action plan for capability development (Naheed et al. 2025).

Long and Magerko (2020), in turn, define AI literacy as “a set of competencies that enables individuals to critically evaluate AI technologies; communicate and collaborate effectively with AI; and use AI as a tool online, at home, and in the workplace”. Key competencies in their conceptual framework include recognizing AI, understanding intelligence and the distinction between narrow and general AI, grasping representations and decision making, knowing the steps of machine learning, appreciating the human role in AI systems, and developing data literacy and critical interpretation of training data. These competencies are accompanied by design considerations such as explainability, embodied interaction, and contextualization of data to support learning for non-technical audiences (Long & Magerko 2020).

In the education sector, UNESCO’s (2024) AI competency framework for teachers articulates five interlinked aspects: human-centered mindset, ethics of AI, AI foundations and applications, AI pedagogy, and AI for professional development across three progression levels (Acquire, Deepen, Create). The framework stresses that teachers should be able to operate AI tools, understand their underlying models and datasets, design learning activities with AI, critically assess “ethics by design,” and use data analytics for their own professional growth (UNESCO 2024).

Finally, the AI Skills for Business Competency Framework developed by Alan Turing Institute (2024) and UK partners transfers similar ideas to a business context. It distinguishes four learner personas (AI citizens, workers, professionals and leaders) and five dimensions of competence: privacy and stewardship (data life cycle, security, legal and regulatory understanding), specification, acquisition, engineering, architecture, storage and curation (data collection, data engineering, deployment), problem definition and communication (requirements elicitation, success criteria, stakeholder management), problem solving, analysis, modelling and visualization (statistical and AI methods, uncertainty, appropriate use of AI tools) and evaluation and reflection (project evaluation, governance, transparency and explainability, sustainability and reflective practice) (Alan Turing Institute 2024).

To sum, these frameworks converge on a multidimensional view of AI competence that includes (1) conceptual understanding of AI and data; (2) practical skills in using and developing AI tools; (3) ethical, legal, and human-centered judgment; and (4) the ability to integrate AI into domain-specific work practices.

2.4 The AIOK framework and sector-specific competence needs in technology and creative industries

The AIOK (AI-Osaamisen Kartoitusmalli in Finnish, AI Competence Assessment Framework in English) framework, developed in collaboration between the authors and Mindhive Ltd (2025), adapts this multidimensional competence perspective to the context of SMEs. The framework explicitly distinguishes between narrow AI (e.g., predictive maintenance, quality control, demand forecasting, process optimization) and generative AI (e.g., ChatGPT, Copilot, image generators, copywriting assistants), noting that managers often conflate these technologies or hold unrealistic expectations, which affects the whole organization’s ability to adopt AI (Mindhive Ltd 2025).

The AIOK framework includes five dimensions:

1. Basic skills – recognizing AI, understanding key concepts, differentiating narrow and generative AI, and forming realistic expectations (Mindhive Ltd 2025)
2. Application – using AI tools in everyday work, identifying use cases, critically evaluating AI outputs, and prompt engineering for generative AI (Mindhive Ltd 2025)
3. Ethics and responsibility – awareness of ethical issues, data protection, bias, transparency, and compliance with the EU AI Act (Mindhive Ltd 2025; UNESCO 2024)
4. Integration – linking AI to business processes, data utilization, strategic planning, and balancing skills and technology (Jöhnk et al. 2021; Mindhive Ltd 2025)
5. Development – technical ability to create AI solutions, understanding models, continuous learning, and building an organizational learning culture (Long & Magerko 2020; Mindhive Ltd 2025; Naheed et al. 2025).

Therefore, the framework draws explicitly on AI literacy research, UNESCO’s teacher framework, organizational AI readiness factors, and recent AI readiness models for SMEs. It also differentiates indicators for technology-intensive industries (e.g., IoT, computer vision, ERP integration) and creative sectors (e.g., content production, image generation, copyright and authenticity), highlighting that competence profiles and risk landscapes differ between these domains (Jöhnk et al. 2021; Long & Magerko 2020; Mindhive Ltd 2025; Naheed et al. 2025; UNESCO 2024).

3. Research design and methodology

3.1 Research approach and strategy

There is still little empirical knowledge on which types of AI competencies are already present in SMEs and what concrete training needs these firms perceive, particularly in technology and creative sectors where the business effect of AI appears significant. By combining data from thematic interviews with SMEs and the AIOK framework as an analytical lens, this study addresses gaps in AI adaptation. It aims at mapping existing AI-related competencies (RQ1) and training needs (RQ2) in technology and creative SMEs, while also generating empirical input for further AIOK framework-based survey development and building appropriate continuous learning offering for SMEs.

The study follows a pragmatist philosophical stance, as the main goal is to produce concrete and useful insights that help SMEs themselves understand their current AI competencies and better identify the types of training and support they actually need. Pragmatism suits this research well because the focus is on solving a timely, practical problem rather than testing a specific theory (Saunders et al. 2016). For this reason, the study uses an inductive and descriptive approach together with thematic interviews, which allow participants to describe their everyday experiences, challenges, and expectations related to AI in their own words. These methods help capture what is happening in practice and what kinds of competencies organizations seem to lack. The findings are then used to support the development of training content, support models, and future assessment tools, which are fully in line with the pragmatist idea that knowledge is valuable when it improves real‑world action and decision‑making.

A descriptive research strategy was chosen, as it helps to reveal concepts and constructs that are needed for building sound theories. This study relies on an inductive approach, where empirical data builds ground for identifying new patterns and insights, rather than testing an existing validated theory. The phenomena are observed and described carefully, after which they are placed in categories, in order to produce frameworks or typologies (Christensen 2006) to build ground for empirical testing through quantitative research. For the purposes of studying AI-related competencies in SMEs, this seems a valid approach, as the phenomenon in question is relatively new and not yet widely investigated. As a new tool, the AIOK framework also needs more elaboration and empirical work to serve its further development and applicability.

A single survey-based methodology was chosen to study a socially constructed phenomenon: organizational competencies and needs related to AI in SMEs. Due to the timetable constraints in the larger project of which the study is part, a longitudinal study was not conducted. Instead, the project team decided to focus on a certain moment. Qualitative data was collected through thematic interviews, to apply a method that enables people to reflect on their own experiences and perceptions related to the topic (Auerbach & Silverstein 2003).

3.2 Data collection

The study draws on semi-structured interviews with 25 SMEs in Autumn 2025 to examine AI maturity, perceived competence gaps, and training needs in areas such as AI tools, information management, cybersecurity, content creation, and ethical/legal aspects. The interviews targeted technology and creative industry firms within the same region, yet representing high diversity in size, sub-sector (e.g., manufacturing, software, digital media, event production, photography, handicraft, and design), and digitalization level. The interviewees worked as CEOs, managers, leading experts, or solo entrepreneurs. Altogether, 13 technology firms and 12 creative sector firms participated in the interviews.

Each interview (30–45 minutes) covered themes related to AI use in business, interest in AI, and prior training about the topic, current competencies, barriers, future expectations, and training needs. With the permission of each participant, interviews were automatically transcribed using the Microsoft Teams transcription tool.

Potential ethical concerns in our research setting include the protection of commercially sensitive and personal information, consent for participation, and the risk that AI-assisted transcription or analysis may reproduce inaccuracies. In our study, these concerns were addressed by explicitly requesting participants’ permission for recording and transcription, limiting the material to the purposes of the research and training project, avoiding the inclusion of sensitive customer or personal data in interviews and reporting, and manually reviewing the transcripts and AI-assisted outputs as part of the researchers’ joint quality control process.

3.3 Data analysis

Data were analyzed using qualitative content analysis. Transcripts were coded for themes related to AI maturity (no use, experimental use, integrated use), five competence areas derived from the AIOK framework as described earlier, and concrete training needs identified. For the latter, sectoral patterns (technology vs. creative) were identified to guide training design.

The analysis procedure was conducted as follows:

1. Making field notes and drawing rough mindmaps directly from online interviews.
2. Reviewing interview transcriptions and making the necessary corrections manually, simultaneously complementing field notes and mindmaps.
3. Creating a shared transcription document, one from the creative industry and one from the technology industry, resulting in datasets of 234 pages and 203 pages.
4. Conducting 1st-level coding to both datasets, focusing on embodiments of current competencies and training needs, conducted by two researchers in parallel.
5. Exporting the identified 1st-level codes to a separate document using VBA macro, resulting in datasets of 29 and 23 pages.
6. Conducting 2nd-level coding with a goal of explicitly differentiating between AIOK framework categories (B-basic skills, A-application and everyday use, E-ethics, I-integration, D-development) and training needs, while also illustrating findings with citations. Table 1 provides illustrative examples of the 1st and 2nd-level codes.
7. Complementing the analysis, cross-checking findings, and summarizing results per industry.

AI assistance (specifically M365 Copilot) was applied in four separate stages of the analysis and overall study: 1) in generating the interview transcriptions in Microsoft Teams 2) in generating the necessary VBA code in stage 5 above to export the 1st level codes to separate documents 3) in grouping the identified 2nd-level codes (stage 6) to a summary table to serve the final analysis round (stage 7), and 4) in language revision of the manuscript.

Researcher/investigator triangulation (Patton 2002) was applied in several stages of the study. Two researchers were present in each of the interviews. The transcription data were corrected and complemented in collaboration, and the 1st-level codes were identified and checked by both researchers transparently before proceeding to stages 5-7. The 2nd level coding was conducted by researcher 1. The findings were double-checked and agreed upon jointly.

4. Results

Based on the analysis, there is substantial variation in AI readiness among SMEs. Most of the participating firms were at the stage of experimental use, but there were also SMEs where AI was not yet in use at all, and those who already started incorporating AI in their renewed strategy. 

While existing research emphasizes AI adoption challenges and obstacles (e.g., Oldemeyer et al. 2021), they represent only one side of the coin. In our sample, several SMEs have long experimented with AI-driven automation or creative tools and have built their AI competencies, while it is a reality that many still lack both technical expertise and strategic understanding. There is practically no organisation in the world that could become “ready” with AI, pointing out why it is important to focus on identifying AI-related competencies and systematically developing them, not merely sticking to problems or over-emphasizing the rapid transformation of AI tools. Indeed, all participating SMEs in both industries expressed strong motivation to acquire new skills and learn collaboratively. While our setting certainly involves some bias – the most motivated SMEs are the most likely to participate in a training project and interviews – our data also indicates that interest in building new AI competencies is not limited to large firms only.

In the following, the identified competencies and training needs are described in more detail, starting from creative industry firms and followed by technology industry firms.

4.1 Results from the creative industry: Competencies and training needs

4.1.1 AI fundamentals: Uncertainty and uneven literacy

Summarizing results on AI competencies in creative-industry SMEs starts with basic skills identification. Across respondents, AI literacy appeared highly uneven, ranging from daily routine use to near-zero experience. Uncertainty is not merely technical, but it also concerns the scope of potential use: what is feasible, relevant, and appropriate for one’s own business. Indeed, some interviewees approached advances in AI more as a general force of change rather than something concerning their own everyday work, indicating a kind of alienation from organization-internal AI use.

4.1.2 Everyday application: Content work, ideation, and efficiency gains

Regarding everyday application, many respondents expressed uncertainty about potential use cases and their benefit to their own business. Several interviewees described being beginners, lacking clarity on what AI can be used for, or needing basic orientation before meaningful experimentation can begin.

The most prominent theme in the interviews was pragmatic, task-level adoption, especially in creative production and supporting often lonely knowledge work. The most common applications include drafting and adapting texts (e.g., marketing content, e-commerce product descriptions, contracts, proposals, procurement documents), translation and “plain language” rewriting, as well as image/video-related workflows (e.g., accelerating editing, generative imagery trials, or sketching visual assets).

In application competencies, it was also noteworthy that many of the participating SMEs in the creative industry already excel in differentiating between human-based tasks and AI-driven tasks. A recurrent framing was that AI functions as a “co-worker” or “sparring partner” for ideation and refinement rather than a fully autonomous producer (or a trustworthy producer, for that matter), with multiple respondents emphasizing that outputs must always be reviewed, iterated, or reworked. This requires a skilled human workforce by nature. 

Most interestingly, while a few respondents expressed slight concern about being replaced by AI, others had conducted actions representing quite the opposite side:

“When I started to realize the potential of AI, I simultaneously decided to hire an illustrator. Some might consider that a crazy choice, but I see it the opposite way. The value of hand‑made, genuinely crafted illustration increases the more generic AI‑generated content there is.” (C7)

“It already feels a bit lighter now that I don’t have to carry every single idea solely in my own head, but can instead bounce things around with the AI to some extent.” (C9)

Efficiency benefits were frequently noted. AI is described as accelerating routine drafting, reducing the cognitive load of ideation, especially for solo entrepreneurs, and enabling faster turnaround in content-heavy tasks. However, respondents also highlighted that effective use depends on the ability to steer tools (e.g., specifying language variants, controlling output style, or technical parameters), which implicitly links everyday application to prompting and domain expertise. Current competencies in prompt engineering vary, but it is evident that without a skilled workforce, there is no one to guide AI applications.

4.1.3 Ethics and responsibility: Copyright, data boundaries, and reliability risks

Concerning ethical and societal issues, an important and highly salient topic was the effect of AI on creative work overall:

“Of course, since I’m a professional photographer myself, it has been frustrating to see AI emerging — those jobs will be taken in the future.” (C12)

“I’m very aware this is something I can’t change, so I just have to go along with it and stay on the crest of the wave. So could AI perhaps be more of a friend to me than an enemy?” (C9)

Competencies related to ethical and responsible use of AI concern a cluster around three other issues. Firstly, copyright and creators’ rights are repeatedly described as central — particularly in relation to AI-generated images and the provenance of training data. Secondly, data boundaries and privacy/security shape what content organizations are willing to input into commercial tools; some respondents report restricting inputs to material that is already public or intended for publication. Thirdly, reliability and quality risks are salient. For instance, there are translation errors that are difficult to detect, mixed language varieties, and “hallucination-like” inaccuracies that were described as requiring strong critical literacy and subject-matter understanding. In addition, regulatory uncertainty is perceived as a theme that needs guidance and training.

4.1.4 Integration into business processes and strategy: Emerging but less articulated

Compared with task-level use, explicit strategy talk was less common in the interviews, which could be considered typical for smaller firms. Nevertheless, there are clear signs of process and service integration. For instance, respondents mentioned using AI to support marketing analytics, proposal writing, reporting, procurement processes, customer-facing experience concepts (e.g., conversational AI as part of exhibitions), and especially new service/product development (e.g., branded AI image banks or AI-enabled components embedded in digital products). 

“I could develop new services around AI. – We are creative people: when AI takes some part of works that we have earlier billed from our customers, we are able to develop new ones.” (C2)

“What AI enables is a wider pool of services we offer to our customers. They also get more with the same money, things we could not have produced 10 years ago.” (C3)

Some statements connected AI adoption to business constraints typical of small creative firms — limited time, limited budgets, and the need to broaden service portfolios — suggesting that integration is often opportunistic and client-driven rather than guided by formal roadmaps.

4.1.5 Continuous development and culture: Need for hands-on learning and feasible training

Referring back to Basic skills in 4.1.1, interviewees emphasized that training should open up potential use cases:

“This is where you could best help us with this training project, give us insight about how to use AI.” (C6)

A major cross-cutting theme was learning infrastructure. Respondents reported a strong interest in training but face barriers such as cost, time, and uneven accessibility of good-quality offerings. In line with the findings of prior studies on AI adoption in SMEs, many interviewees described how it is difficult to know in practice which training is worth the time, and especially the money. 

However, interviewees emphasized that training should always be practical and “hands-on,” not merely webinars with talking heads, but with sufficient time to internalize knowledge and apply tools in one’s own work rather than only hearing inspirational examples. According to the interviewees, especially peer learning — testing tools together and sharing experiences — appears as an enabling mechanism for sustained uptake.

At the organizational level, interviewees described varying degrees of internal enablement, from daily use by some staff to the inability to provide internal training for everyone, highlighting a cultural challenge of allocating time and legitimizing learning as part of work.

4.1.6 Summary from the creative industry

Overall, the interviews depicted a field where AI is increasingly normalized as one tool among others, yet adoption is constrained by uneven literacy and limited resources. The strongest immediate value lies in augmenting ideation and content workflows, while more strategic integration appears in early-stage experiments around new services and customer experiences. Ethical issues — especially copyright, data governance, and reliability — are not peripheral; they directly influence everyday choices about tool selection, input restrictions, and quality assurance. 

The creative industries’ respondents described AI as simultaneously enabling and disruptive. It can accelerate routine work and expand feasible offerings, but it also raises acute questions about rights, quality, and professional identity. Effective support would likely require tiered learning pathways (from fundamentals to applied workflows), affordable and time-feasible training formats, and an organizational culture that legitimizes experimentation while enforcing responsible-use boundaries. The results highlight specific training needs and inform the design of targeted training packages and capacity-building measures both within the project and broader AI-oriented R&D infrastructures.

4.2 Results from the technology industry: Competencies and training needs

4.2.1 AI fundamentals: Early-stage literacy and uneven awareness

Across SMEs in the technology industry, AI literacy remains a bit uneven despite growing exposure. Several respondents describe being in a starting phase of adoption, with limited systematic use beyond exploratory ChatGPT or Copilot trials. 

“We have a huge gap — practically a chasm — in the sense that some people use AI constantly, while others haven’t even opened such applications.” (T8)

“We are quite at the starting point. Other large projects have put back our AI implementation.” (T1)

While some companies have provided company-wide introductory training (e.g., training all white‑collar workers in generative AI tools), others rely on self-directed learning in the absence of structured programmes. 

4.2.2 Everyday application: Support for knowledge work, data tasks, and technical workflows

Like in the creative industry, in SMEs, a recurring theme was slight uncertainty about what AI can realistically do, which hinders idea generation and use-case identification. This seems to form a “chicken-egg” type of problem: identifying use cases would require more understanding about the potential of AI, but understanding the potential of AI calls for more concrete use cases approachable from the perspective of the business in question.

“We have a dedicated employee who follows the development of AI, and picks the most promising applications from the perspective of our business.” (T8)

AI is currently used most actively in daily knowledge‑work tasks, including summarising documents, drafting material, generating text templates, supporting translations, and assisting with internal communications. Tools like Copilot are used to process standards, extract key points, and draft or refine large bodies of content, substantially reducing routine workload. In more technical settings, AI supports coding, debugging, and test automation, helping to minimise human error and accelerate repetitive routines.

Data handling is a major emerging application area. Respondents mentioned automated reporting, data retrieval, Power BI support, and Excel‑based analytics, for instance, although many still use these capabilities at a basic level. In sales and marketing, AI is applied to offer creation, pricing, customer communication, and crafting marketing materials. Some companies are experimenting with domain‑specific AI agents, such as internal documentation retrieval agents to support technical service.

Critical evaluation seemed to form a relatively strong competence area also in technology industry SMEs. Respondents underlined the importance of critical AI literacy, as AI may produce convincing but incorrect claims or fabricated sources, necessitating systematic verification practices.

4.2.3 Ethics and responsibility: Privacy, data protection, and reliability risks

Overall, the participating SMEs in the technology industry were highly aware of the ethical issues around AI. The most pronounced ethical concerns were related to data privacy. Companies emphasise strict rules about what can be safely entered into AI tools, especially concerning customer data, internal pricing, or personal information. HR‑related uses raise heightened privacy risks due to sensitive personal data.

“I have systematically tried to avoid innovative use of AI before we understand the risks better.” (T1)

“Clearly, we need more knowledge and dialogue about data privacy and safety. How can we use AI in our business, bearing in mind the huge amount of customer data we need to deal with? — We also need to understand the environmental effects better, find guidelines about when it is reasonable to use AI overall?” (T3)

Contractual and regulatory clarity were an additional concern. Some respondents pointed out AI Act is not yet familiar enough to them, which has also motivated participating in the training project. Companies request clear guidance on what is allowed under current regulations, how to manage vendor relationships, and how to prevent accidental data leakage. Other risks include cost risks (e.g., chatbots generating unpredictable usage-based fees) and concerns about the environmental impact of AI computation. OpenAI may not care about setting measures for its environmental responsibility, but many small technology firms are interested in doing so.

4.2.4 Integration into business processes and strategy: From emerging vision to targeted use-cases

Strategic integration of AI is underway, but often still early, indicating lower competence than in the two previous sections. Some technology industry SMEs have started embedding AI into strategic plans and leadership agendas, while others describe fragmented or informal strategic direction. AI is increasingly incorporated into core business processes, particularly through automation of back-office tasks, optimization of production planning, predictive analytics in sales, and improved interoperability between information systems.

“Considering the bigger picture, AI really is an issue of leadership and management.” (T2)

“We need to understand better what is possible with AI, we need to open our eyes.” (T7)

Companies also explore customer-facing or internal AI services, such as product/service agents, offering AI-enabled tools to customers with significant data assets, or utilising external AI vendors for demonstrations and pilot projects. Vendor ecosystems play an important role: organisations rely heavily on software suppliers to build AI features into ERP, CRM, and production systems, shaping the pace and direction of adoption. A critical concern is therefore related to how the existing systems and AI could be integrated, or whether AI is something that will eventually replace existing domain-specific systems.

Therefore, the SMEs are trying to solve a problem that escapes even further with every single step. The trickiest part is how SMEs know what is worth investing in with scarce resources and uncertainty about the actual solution: 

“When we ask about AI solutions, everyone is always offering something, but they never show what it actually does, and the starting price tag is always at least 10.000 euros. So it has stalled there, because you can’t be sure whether the tool even works.” (T7)

4.2.5 Continuous development and organisational culture: Training, support, and readiness gaps

All participating SMEs in the technology sector identified training and continuous support as essential. Some have already delivered systematic AI training, while others lack any collective effort and have depended more on individual initiative or ad-hoc training. Like in creative sector SMEs, respondents emphasised the need for role‑specific, hands‑on programmes that go beyond introductory “AI basics” and move toward practical workflow-level skills — particularly in data processing, automation, and agent development. A single, uniform training model does not accommodate the varying levels and tasks within the workforce.

“We have a need to upgrade our understanding of how AI supports coding work; this is what we need more knowledge about. It is such a constantly evolving field that you need to upgrade on a continuous basis.” (T5)

A common cultural challenge is ensuring that all employees remain involved, avoiding a divide between early adopters and those falling behind. 

“I see a significant risk – both from the perspective of the company and the person – in that someone drops out from understanding the development of AI. That could backfire later.” (T2)

There were also mentions of the general lack of an AI-capable workforce in comparison to demand: 

“We do not have so many AI competencies at the moment, and it is something you can’t even buy with money. So the biggest challenge, actually, is how can we get such talent here?” (T6)

Companies noted that leadership must legitimise experimentation, provide enough dedicated time for learning and experimentation, and support behavioural change. This would support building a culture that helps to attract talent. Peer learning and inter-organisational exchange are also seen as valuable mechanisms for maturing AI competencies.

4.2.6 Summary from the technology industry

The technology sector demonstrated a process‑ and system‑driven orientation toward AI, with strong interest in automation, data‑driven decision-making, and integration into enterprise systems. At the same time, key competencies — especially safe data handling, critical evaluation of outputs, and strategic alignment — remain uneven. AI literacy gaps create a bottleneck for realising value: while some tasks are efficiently augmented, many firms are still “chatting” with AI tools instead of applying them in higher‑impact workflows.

Ethical concerns—mainly privacy and reliability—shape adoption decisions directly. Organisational policies, controlled environments, and robust training are prerequisites for expanding AI use into sensitive processes. Finally, the results imply that strategic value emerges when organisations move from isolated experiments to structured integration across functions, supported by leadership, governance, and partnerships with trusted vendors.

The interviews portrayed a sector actively exploring AI’s potential while navigating uncertainties related to competence, risk, and strategic alignment. AI is seen both as a practical tool for reducing routine work and as an enabler of more sophisticated data-driven competencies. Realising this potential, however, requires sustained organisational learning, clearer governance, and stronger bridges between exploratory use and systematic integration.

5. Discussion

To summarize our findings related to RQ1 about AI competencies, a distinct competence in the SMEs across sectors was a clear understanding of the role of AI as a sparring partner or capable assistant, demonstrating high awareness but avoiding over-reliance. In companies where experience on AI use was higher, managers and entrepreneurs had also invested in substance knowledge and workers capable of steering AI applications, and by no means replacing the workforce with AI to gain “productivity improvement”.

In general, managers and entrepreneurs were also well aware of the ethical, legal, and information security-related principles, but called for more understanding on how to execute them in practice. For instance, creative industry SMEs pondered the moral justification of constantly reusing someone else’s creative material in large language models (LLMs), and technology industry SMEs called for a better understanding of when it is reasonable to use AI and when not, because of its high environmental load. 

A typical shortcoming in AI competencies was the lack of ability to see the concrete use cases and identify new opportunities. Almost all participating SMEs underlined the need for developing more competencies in seeing the potential of AI: where it could be used to systematically support and broaden business, in line with current operations, and build competitive advantage. Here, most SMEs would benefit from sharing case insight and learning together with peers. This is possible as many of the participating SMEs were not direct competitors but rather provided complementary offerings in their market, and many have already networked with each other.

How about the concrete training needs identified, then? Regarding RQ2, SMEs call for more holistic training programs, in contrast to random offerings or overloaded “hype” around AI, which often describes certain AI vendors or applications. In creative industry SMEs, a key implication was that competency development should not be treated merely as learning how to prompt, but as a solid combination of (a) foundational AI literacy, (b) domain-specific application patterns, and (c) responsible-use practices embedded into routines (review, validation, documentation of tool use, ethics, legal, and copyright issues). In a similar vein, findings from the technology industry suggest a need for multi‑layered competence development, balancing foundational literacy with role‑specific applications and expectations.

Table 2 summarizes the key findings and differences between the two sectors.

In line with existing research (e.g., Oldemeyer et al. 2024), the participating SMEs underlined the current knowledge/resource constraints and potentially high costs in developing AI competencies. For a small company, it is often difficult to see what is worth investing in, as regards training offerings. In contrast, time was a challenge but not a key constraint: practically all participants were eager to participate and dedicate time in training, which would offer concrete benefits and help to increase the firm’s productivity. 

The study makes three key contributions:

1. Empirical mapping. The study provides a systematic analysis of AI readiness and competence gaps among SMEs in two sectors within one region, a perspective largely missing from prior research. The findings benefit regional developers, business advisors, trainers, educators, and AI solution providers, among others.

2. Theoretical and practical impact. The study applied the conceptual AIOK framework developed in collaboration with Mindhive Ltd (2025), opening up a new theoretical lens to study AI competencies in SMEs. The results directly inform tailored training content, coaching, and policy support, ensuring that interventions align with real-life company needs. By identifying sectoral differences and common strengths/obstacles, the study also supports workforce development, regional policy, and more responsible AI adoption.

3. Future baseline. The findings offer a baseline for follow-up studies on how AI training and upskilling influence strategic capabilities, operational processes, and competitiveness, thereby contributing to sustainable regional development and resilience. The study also builds ground for further application of the AIOK framework.

There are several limitations in our study, such as the focus on only one point in time in data collection, studying only two industries, plus the information bias (typically, the CEO, one manager, or entrepreneur representing the whole company in the interviews). In further research, it would be interesting to compare insights from different personnel groups in SMEs. The results may also have been biased in terms of enthusiasm and motivation, as technology and creative industries may have a stronger interest in advances in AI than other industries. For instance, in health and wellbeing, education or agriculture, the attitudes, experiences and overall competencies could appear different.

In further research, it would be beneficial to map AI competencies with the AIOK framework based on a more detailed empirical survey instrument. As the current study focused on one point in time, a valuable avenue for further research would also be a longitudinal study on how SMEs have been able to develop their AI competencies, and how the actual needs for competence development differ across firms. This opens up interesting possibilities both for technology and creative industries investigated in our study, and more widely across sectors to allow comparisons.

To sum up, the participating SMEs were well aware of the obstacles, such as a high workload, which may hinder learning, but simultaneously they demonstrated high motivation and “realistic enthusiasm” about AI. The best motivator is concrete examples from peers and experienced trainers. Indeed, AI was seen as worth investing time in learning, resulting in potential savings later.

In line with our research theme, we conclude with a near-future vision by M365 Copilot.

Prompt: “Write one catchy sentence about what the future of AI adoption and use in SMEs in creative and technology industries looks like. Do not repeat the industries, but just summarize with an anecdote or funny example what to expect.“

Copilot said: “Pretty soon, an SME’s Monday meeting will start with someone asking the AI assistant why it already prototyped three new ideas overnight – complete with mood boards, a budget, and a note saying ‘I took the liberty of improving your plan.’”

Acknowledgment

The authors gratefully acknowledge the financial support from the European Union for the Smart Moves – AI Competencies and Business Renewal project (2025-2027).

References

Alan Turing Institute (2024). AI Skills for Business Competency Framework. Version 2. London: The Alan Turing Institute. https://zenodo.org/records/11092677

Auerbach, C.F., & Silverstein, L.B. (2003). Qualitative Data. An Introduction to Coding and Analysis. New York: New York University Press.

C&Q Systems (2025). Osaamisen johtaminen ja osaamisresurssien hallinta [Competence management and management of competence resources]. https://cqsystems.fi/palvelut/osaamisen-johtaminen/

Christensen, C.M. (2006). The ongoing process of building a theory of disruption. Journal of Product Innovation Management, 23, 39–55.

Dinh, T., Vu, M-C. & Tran, G. (2025). Artificial Intelligence in SMEs: Enhancing Business Functions Through Technologies and Applications. Information, 16(5), 415.

Elinkeinoelämän Keskusliitto (2025). Ensiaskeleet tekoälyn käyttöönottoon liiketoiminnassa – opas pk-yrityksille [First steps in adopting artificial intelligence in business – A guide for SMEs]. Helsinki: EK. https://ek.fi/wp-content/uploads/2025/03/EK_Ensiaskeleet-tekoalyn-kayttoonottoon_FINAL.pdf

European Parliament (2023). What is artificial intelligence and how is it used? https://www.europarl.europa.eu/topics/en/article/20200827STO85804/what-is-artificial-intelligence-and-how-is-it-used

Jöhnk, J., Weißert, M., & Wyrtki, K. (2021). Ready or Not, AI Comes—An Interview Study of Organizational AI Readiness Factors. Business & Information Systems Engineering, 63(1), 5–20. https://link.springer.com/article/10.1007/s12599-020-00676-7

Kramarenko, A. (2025). Artificial intelligence for small and medium business: Perspectives and challenges. Journal of Engineering Management and Competitiveness (JEMC), 15(1), 43-56.

Long, D., & Magerko, B. (2020). What is AI Literacy? Competencies and Design Considerations. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI ’20). https://dl.acm.org/doi/epdf/10.1145/3313831.3376727

Mindhive Ltd (2025). AIOK-malli. Tekoälyn osaamistarvekartoituksen toteutussuunnitelma. [AIOK framework. Implementation plan for AI competence assessment.]

Mohd Rasdi, R., & Umar Baki, N. (2025). Navigating the AI Landscape in SMEs: Overcoming Internal Challenges and External Obstacles for Effective Integration. PLOS One, 20(5). https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0323249

Naheed, S., Pinto, R., & Pirola, F. (2025). A Preliminary Multidimensional AI Readiness Assessment Model for Small and Medium-Sized Enterprises. Procedia Computer Science, 253, 774–783. https://www.sciencedirect.com/science/article/pii/S1877050925001474

Oldemeyer, L., Jede, A., & Teuteberg, F. (2024). Investigation of Artificial Intelligence in SMEs: A Systematic Review of the State of the Art and the Main Implementation Challenges. Management Review Quarterly, 75, 1185–1227. https://link.springer.com/article/10.1007/s11301-024-00405-4

Patton, M.Q. (2002). Qualitative Research & Evaluation Methods. Thousand Oaks: Sage Publications.

Sánchez, E., Calderon, R. & Herrera, F. (2025). Artificial Intelligence Adoption in SMEs: Survey Based on TOE–DOI Framework, Primary Methodology and Challenges. Applied Sciences, 15(12), 6465.

Saunders, M.N.K., Lewis, P., & Thornhill, A. (2016). Research Methods for Business Students. Pearson.

UNESCO (2024). AI Competency Framework for Teachers. Paris: UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000391104

Weinert, F.E. (2001). Concept of competence: A conceptual clarification. In D.S. Rychen & L.H. Salganik (Eds.), Defining and Selecting Key Competencies, 45–65. Hogrefe & Huber Publishers.

URN: http://urn.fi/urn:nbn:fi:jamk-issn-2341-9938-96