Manufacturing firms are increasingly pursuing AI-led transformations (e.g., predictive maintenance, quality inspection, and autonomous production), yet implementation often stalls after pilot projects. Prior literature links this “pilot-to-scale” gap to disorganised data and legacy IT (Clemens et al., 2023; Plathottam et al., 2023; Rauh et al., 2022), misaligned investment logics and budgeting routines (Hadhri et al., 2025), and capability deficits spanning technical skills and digital leadership (Mqoqi et al., 2026; Obi et al., 2025). At the same time, responsible AI introduces additional coordination challenges because accountability, transparency, and compliance remain difficult to operationalise across organisational layers and heterogeneous use cases (Besinger et al., 2025; Eng-ström et al., 2025). Based on previous studies (Engström et al., 2025), we argue that AI adoption is a socio-technical change where value arises from combining technology (models, data pipelines, and platforms) with decision-making rights, incentives, workflows, skills, and accountability. If this combination is weak, organisations might fall into “pilotism,” resulting in many projects but little learning or integration. This study asks: What organisational and governance mechanisms are required in manufacturing firms to convert AI aspirations into scalable and beneficial implementations? To answer our research question, a qualitative approach combining a focused literature review with co-creation workshops involving managers and specialists from six companies (A-F) was employed (Ahmed & Asraf, 2018) to identify AI goals, perceived blockers, and scaling mechanisms through discussions and group tasks (Ørngreen & Levinsen, 2017). The empirical material (field notes, audio recordings, and secondary data) was analysed using content analysis to find recurring patterns (Bengtsson, 2016). Across firms, evidence confirms literature: AI remains stuck in fragmented pilots because strategy, budgeting, data/legacy IT, skills, and governance are misaligned. More specifically, AI remains stuck because key organisational elements are misaligned. Firms often run many promising Proofs of Concept (PoCs), but the organisational system required for scaling (strategy → funding → data/IT → people → governance) does not “fit together”, preventing pilots from being industrialised into scalable, accountable deployments. There is a notable lack of shared direction, which results in isolated initiatives that fail to contribute to overall enterprise learning or effective scaling. The current funding model adheres to traditional ROI metrics, which create barriers to securing resources for scaling successful pilots, as they often fail to meet conventional thresholds or timelines. Many pilots are developed under conditions that do not generalise well, leading to challenges in scalability. Companies lack robust pipelines and deployable infrastructure necessary for broader implementation. There is an over-reliance on a limited number of individuals, as roles and learning routines are not well-defined. This situation hinders the establishment of standardised practices across teams. The absence of clear accountability and risk control mechanisms has blocked the operationalisation of successful pilot initiatives. On the other side, companies are working to tackle challenges by implementing a “governance spine” that merges centralised direction with decentralised execution. This approach features a dedicated AI strategy, clear decision rights, and local ownership through AI ambassadors and cross-functional teams. Key elements include a structured use-case pipeline (idea → business case → pilot → adoption), standardised tracking of ROI and KPIs, and role clarity with capability-building infrastructures for ongoing learning. Strong data governance ensures security and compliance. Successful scaling is more about organisational change than technology, requiring effective leadership and adaptive resourcing to align initiatives with operational needs.

Purpose: This paper examines how gender shapes employees' emotional reactions to workplace AI, and whether these differences are associated with epistemic legitimacy, explained by perceived AI knowledge and conditioned by psychological safety. Design/methodology/approach: We analyse survey data from 104 employees using a conditional process framework. Multiple regression models test direct gender effects on cognitive and emotional reactions to AI, mediation via perceived AI knowledge, and moderation by team-level psychological safety. Findings: Women report less favourable cognitive and emotional reactions to AI than men. These differences are largely attributable to lower perceived AI knowledge. Psychological safety attenuates the direct gender effect: in high-safety climates, gender gaps in reactions diminish; in low-safety climates, they re-emerge. Overall, the pattern supports a contextualised partial-mediation model in which perceived knowledge is pivotal, but its explanatory power depends on climate. Research limitations/implications: The cross-sectional design limits causal inference and the generalisability is bounded by the organisational context studied. Future research should use longitudinal or experimental designs, examine additional inequality dimensions (e.g. age, role), and unpack how AI literacy interventions reshape appraisal dynamics over time.Practical implicationsAI initiatives should build employees' perceived understanding (how AI works, limits, and human-AI complementarity) and foster psychological safety so that questions and uncertainty are acceptable. Monitoring gendered participation and confidence during roll-out helps prevent AI from amplifying existing inequalities. Social implications: Managing AI adoption as an inclusion challenge-rather than solely a technical one-can reduce uneven emotional costs of digital transformation and support fairer access to AI-enabled opportunities. Originality/value: The study integrates gender, appraisal (perceived AI knowledge), and climate (psychological safety) into a single framework explaining both cognitive and emotional reactions to AI. It reframes gender gaps as contextual, highlighting levers - literacy and climate - that organisations can use to enable more equitable AI adoption.

AI is viewed as the next major technological breakthrough for organizations. The range of areas, professions, and practices that can be improved with AI assistance or automation is overwhelming. However, this wide array of possibilities also brings a variety of expectations about how AI will change organizations and employees’ everyday work. Considering that voiced expectations influence adoption processes by both reflecting and shaping certain relational, belief-driven dynamics, we can learn a great deal about AI adoption by studying the organizational plurality of AI expectations. Therefore, this study examined AI expectations held by employees in three organizations currently adopting AI for use in the workplace. The study is based on a thematic analysis of empirical material form 15 focus groups in three Swedish AI-adopting organizations and shows how AI expectations shape the following: (1) a growing desire to move from exploring AI to establishing AI routines and regulations, (2) emerging dilemmas related to both the violation and fulfillment of AI promises, and (3) how dynamics and unpredictability in the AI field require organizations to adapt to shifting trends and innovations.

Purpose

Artificial intelligence (AI) has the potential to be a disruptive technology instigating paradigm shifts and significantly impacting both operational and strategic processes while presenting organizations with contradictory demands. This paper aims to explore tensions in the early stages of AI adoption in manufacturing firms.

Design/methodology/approach

Drawing on paradox theory and 23 focus groups with 112 white-collar employees, we identify 9 themes categorized into 2 overarching dimensions: (1) temporal tensions between current capabilities and anticipated futures and (2) relational tensions between human and machines.

Findings

Temporal tensions unfold along emergent trajectories where progress is uncertain, nonlinear and continuously negotiated. Relational tensions are marked by ethical reflexivity, as employees navigate shifting expectations around trust, control and human identity. Our findings show how these tensions reinforce one another and contribute to organizational hesitation and strategic drift.

Research limitations/implications

The study focuses on large multinational manufacturing firms, limiting the generalizability of findings to other industries or small and medium-sized enterprises. Additionally, the study captures only the early stages of AI implementation, necessitating further longitudinal research to examine long-term organizational adaptations. Future research should explore cross-industry comparisons and investigate how different leadership styles influence AI-driven transformations.

Practical implications

This research provides actionable insights for managers navigating AI integration. Addressing tensions proactively – through strategic workforce upskilling, clear AI governance frameworks and fostering human-AI collaboration – can mitigate resistance and maximize AI’s benefits. Organizations should embrace paradoxical thinking, leveraging AI to enhance productivity while supporting employees in transitioning to evolving roles.

Social implications

The study highlights the societal impact of AI adoption in the workplace, particularly concerning job displacement fears and the evolving human-machine relationship. Managing AI transformation responsibly requires ethical considerations, workforce empowerment and inclusive policies to ensure that AI-driven innovations benefit employees and organizations alike.

Originality/value

By conceptualizing AI adoption as a process shaped by interwoven temporal and relational paradoxes, the study offers new insights into the emotional, ethical and structural dimensions of integrating AI into established organizational settings.

This paper explores Industrial doctoral students’ experiences of interpersonal interactions and how they contribute to learning in context of development of digital production. For this we interviewed six IDSs working on developing digital technologies for production systems in six distinct manufacturing companies. The qualitative data collected was analysed using Gioia method. The findings include the characteristics of interpersonal interactions, conditions and potentials for learning. Overall, the study found that interpersonal interactions, especially in physical environments are valuable for the IDSs and perceived to contribute to learning at individual, group and organizational levels.

Knowledge capability and innovation performance are essential drivers of organizational success, with managerial work playing a crucial role in fostering these factors. This paper examines planned and unplanned managerial work from a knowledge creation perspective, analyzing 21 short interviews and 21 days of shadowing seven leaders, alongside interactive analyses with management teams in four manufacturing SMEs. A four-step analysis of 2125 activities revealed four categories of managerial work, with unplanned activities accounting for 52% of work time, highlighting the complexity and ad-hoc nature of managerial work. Internally initiated unplanned managerial work was crucial for tacit knowledge conversion. If policymakers want to support lifelong learning initiatives and strengthen innovation capabilities in manufacturing SMEs, they need to encourage the unplanned aspect of managerial work, as it is crucial for knowledge creation, fostering the evolutionary aspect of organizational learning and enhancing innovation capabilities.

Purpose

The purpose of this study is to explore the process of initial sensemaking that organizational members activate when they reflect on AI adoption in their work settings, and how the perceived features of AI technologies trigger sensemaking processes which in turn have the potential to influence workplace learning modes and trajectories.

Design/methodology/approach

We adopted an explorative qualitative and interactive approach to capture free fantasies and imaginative ideas of AI among people within the industry. We adopt a conceptual perspective that combines theories on initial sensemaking and workplace learning as a theoretical lens to analyze data collected during 23 focus groups held at four large Swedish manufacturing companies. The data were analyzed using the Gioia method.

Findings

Two aggregated dimensions were defined and led to the development of an integrated conceptualization of the initial sensemaking of AI technology adoption. Specifically, sensemaking triggered by abstract features of AI technology mainly pointed to an exploitative learning path. Sensemaking triggered by concrete features of the technology mainly pointed to explorative paths, where socio-technical processes appear to be crucial in the process of AI adoption.

Originality/value

This is one of the first studies that attempts to explore and conceptualize how organizations make sense of prospective workplace learning in the context of AI adoption.

The interest in collaborative research projects between industry and academia is growing, but the understanding of knowledge creation and how work-integrated learning takes place between these arenas, requires further attention. A deeper understanding of collaborative research projects and the learning processes is needed since interactions between industry and academy show both complicated and complex patterns. The study is based on a qualitative, explorative approach. Seven company representatives were interviewed concerning their experiences of collaboration with the academy. The results shows that the company participants’ competence level is crucial concerning motivation and abilities to contribute to the collaborative results. The communication between the academics and the practitioners is mainly performed by top-down processes, which gives advantages in the learning process to practitioners with high academic knowledge. A model is presented showing possible ways to increase the learning process, from top-down processes to bottom-up, that will include all the participants regardless of competence. To attract academics and company partners in participating in collaborative projects, this paper shows that it is essential that the project management is knowledgeable about individual participants’ knowledge level and take this into account when jointly designing the project setup.

Purpose The purpose of this paper is to shed light on different types of knowledge created and how this links to the project design, process, and content. Design/methodology/approach In this paper the authors investigate participants' experiences from a three-year interactive research project, designed to trigger reflection among the participants. They apply a knowledge creation perspective on experiences expressed by participants as a result of different research project activities. Findings The study resulted in five categories of insights with potential for sustainable influence on the participating organizations: an understanding of concepts and theories; an understanding of the impacts of collaborative, reflective work processes; an understanding of the meaning of one's own organizational context; an understanding of the importance of increased organizational self-awareness; and an understanding of the potential for human interaction and communication. Practical implications The author's findings suggest that it is possible to design a project to promote more profound and sustainable effects on a business beyond the explicit purpose of the project. They advise practitioners to make room for iterative reflection; be mindful to create a trustful and open environment in the team; challenge results with opposing views and theories; and make room for sharing experiences and giving feedback. Originality/value This study contributes to unraveling key practices which can nurture conditions for knowledge creation in interactive research projects and business projects alike.

Introduction

Moving from manual, to automated, to connected AI operations systems implies a significant transformation in the organisation of work (European Parliament, 2015:8) (Brock & von Wangenheim). To understand these “realistic AI” processes, to build competence for certain tasks. it is crucial to understand what organisational competencies that are needed and how to organize knowledge creation processes in practice (Ellström, 2001) Schön used the concept of “knowing–in-action” is nonreflective and solving most everyday practical problems, here understood as executional learning (Engström & Wikner, 2017). Thus, this knowing, according to Schön (1983), is not enough to meet more complex situations. To be aware of tacit knowledge, we need to distance ourselves and learn to reflect. More complex, uncertain and unclear tasks require “knowing-on-action” and collaboration between several competences to create new knowledge or to reach a new solution here understood as developmental learning (Engström&Wikner).. Anton et al. (2020) state that in many organisations there is a lack of AI-related competencies that prevent development of the full AI potential. For the development of the field, it is important to study the dynamic interplay between advanced technology and the social side of work from a learning and competence perspective. Therefore, this paper aims to explore how industrial organisations understand their competencies in relation to AI transformation from a knowledge creation perspective.

Research method

The study was part of a collaborative research project with an interdisciplinary research team and representatives from five industrial partners. In four-month cycles the industrial partners engaged in “homework” presented, analysed and discussed in common workshops. For this study, the homework was guided by the DIGITAL approach (Brock & von Wangenheim, 2019) and based on the explanatory model (Anton et al., 2020). The industrial partners studied how resources and competencies related to specific organisational tasks in their own organisations could be identified and defined. To aid the data collection (that was done by the industrial partners themselves) a framework capturing Anton et al.’s (2020) 13 dimensions of competencies (Leadership, Communication, Customer-focused decision making, Business development, Data science/STEM, Agile software development, Initiative and engagement, AI technology, Programming, Digital analysis tools, Data and network technology, Digital competencies, and Data management) was used. For each dimension the partners assesses the competence level: Competence central to the process; Competence exists internally; Competence partly exists internally; Competence does not exist internally; Competence can be gained by development internally; Competence needs to be sourced externally. These were in line with Brock & von Wangenheim’s (2019) logic that managers when starting AI project should do “internal resources check”. The data was analysed in four steps. First, focus group data was analysed by the facilitators at each industrial partner. Second, the competence mapping was analysed by the “working groups” at each industrial partner. Third, the transcribed data from the two industrial partners used in this paper were reviewed individually by t he authors. Fourth, the cross-disciplinary group of authors from both academia and industrial partners gathered for a common analysis session. This session primarily focused on the data from the competence mapping but also cross-checked with the input from the cross-functional focus groups to triangulate the outcome. During the common analysis the conceptual framework presented in the discussion section was developed through iterations between the theoretical framework based on the findings by Anton et al. (2020), and the data from the project.

Findings

The preliminary findings show differences among the industrial partners in how they view their own competencies. For some organisations organisational structures are in place, e.g., dedicated AI Labs, where the work with understanding the benefits and usage of the technology is ongoing on a rather advanced level. In other organisations the work has just been initiated. Overall, all representatives stress the importance of top management support and the need for dedicated forums. Among the organisations that have come the farthest in their AI transformation the structure given by the proposed framework is not enough. They emphasise the need to further frame it into also understanding what the competence is associated with and why it is needed. They view the leadership as almost having to have an evangelistic approach to it, where it does not seem to be enough with “only” technical experts. A conceptual framework, consisting of the relationship between the two dimensions: the managerial competencies and the technical competencies, is developed (Figure 1). The managerial competencies dimension concerns organisation and organising. The technical competencies dimension on the other hand captures the complexity level of the technology that is needed, the system of systems. The diagonal illustrates the relationship between these two dimensions, that is, the relation between technological complexity and organisational ability. The lower part of the diagonal captures isolated, simple processes (presumably internal) while the upper part of the diagonal captures integrated, complex processes (presumably primarily related to external parts and/or actors).. For high levels of technical complexity that requires high levels of technical competencies within the organisation the organisation also needs to advance the managerial competencies and the developmental learning processes. However, while in the long-term perspective we suggest that going off the diagonal will be inefficient and ineffective, hence, waste, it might be needed to do that temporarily, as the organisation develops. We believe that this developmentcan be either technology-driven or organisation-driven.

The proposed conceptual framework is intended to help organisations plot their own current position based on the two dimensions and identify what changes are needed to reach the diagonal. It can also be used to define where on the diagonal the organisation ultimately wants to end up. It is not relevant for all companies or even for all sectors overall to be at the top right side. We believe that AI transformation cannot be approached as either technologydriven or managerial-driven, but as an e interdependent process of both dimensions.