Manufacturing Stakeholders’ Perceptions of Factors That Promote and Inhibit Advanced Technology Adoption

Abstract

This study explores factors promoting and inhibiting advanced technology adoption in small- and medium-sized manufacturing firms (SMEs). With AI’s rapid advancement impacting productivity and efficiency across industries, understanding the challenges that SMEs face to remain competitive is crucial. Utilizing the Unified Theory of Acceptance and Use of Technology (UTAUT) model as a theoretical framework, we analyzed managers, engineers, and line workers’ observations on workforce challenges, training needs, and opportunities faced by SMEs to provide insights into their smart manufacturing deployment experiences. Our findings highlight social influence’s role in promoting technology adoption, emphasizing community, shared experiences, and collaborative networks. Conversely, effort expectancy emerged as the largest inhibitor, with concerns about the complexity, time, and resources required for implementation. Individuals were also influenced by factors of facilitating conditions (organizational buy-in, infrastructure, etc.) and performance expectancy on their propensity to adopt advanced technology. By fostering positive organizational environments and communities that share success stories and challenges, we suggest this can mitigate the perceived effort expected to implement new technology. In turn, SMEs can better leverage AI and other advanced technologies to maintain global competitiveness. The research contributes to understanding technology adoption dynamics in manufacturing, providing a foundation for future workforce development and policy initiatives.

1. Introduction

Advancements in artificial intelligence (AI) have gripped society’s attention and the US industry’s investments. Historically, AI has undergone multiple periods of both booming and receding levels of investment, research, and development built on hope and promise [1,2]. But now, the optimistic perspective that AI can outmatch human labor, boost productivity, and lead to scientific breakthroughs fuels investment to new heights. Industry efforts continue to lead research and development, and companies across all sectors are seeking AI-skilled employees to meet the market demands [3]. In the next decade of digital growth, it is predicted that global investment in AI will reach USD 200 billion by 2025 and contribute USD 15 trillion to the global economy by 2030 [4,5]. That is a short turnaround, though—the market investment is likely to continue growing at a pace outmatching substantial human productivity boosts and human adoption [5]. Across the nationwide industry, individual organizations, and small project teams, AI’s technical implementation and sociotechnical infrastructure are developed and governed by humans in various contexts [6,7].

The individuals that make up these industries have different propensities toward accepting, adopting, and using advanced AI technologies. Whether it is a fear of AI or a general lack of knowledge about AI, the adoption of this technology does not match the rate of demand that is required to maintain global economic competitiveness. For instance, increased levels of knowledge about AI have been shown to enhance AI adoption, whereas rudimentary knowledge can inhibit AI adoption [8]. Adopting AI across a variety of people at a firm can lead to less of a bottleneck in efficiency gains across an organization and enhance decision-making for key stakeholders [9]. In a human-in-the-loop approach, having employees interact with machines, facilitate an interface from machines to analytics, and contribute expert knowledge for AI systems to learn from can support successful digital transformation, too [10]. But these recent advancements in technology evolve so rapidly for small- and medium-sized enterprise (SME) manufacturing firms that they struggle to get their foot in the door. Between robotics, tiny machine learning (ML) edge-sensor deployment, and predictive analytics tools for procedural improvement, SMEs face a lot of new technology to assess and integrate. Large and multi-national corporations have extensive resources to develop and source technology; they have room to fail and a broad pool of talent [11]. On the other hand, constrained by a lack of resources, time, and expertise, these firms’ employees struggle to meet the demands required to stay competitive [11,12].

Recent national attention toward manufacturing includes President Biden’s bipartisan CHIPS and Science Act [13], the Department of Commerce Economic Development Administration’s 31 Tech Hubs [14], and expansive manufacturing initiatives at globally recognized R1 universities. As we turn to the challenges and opportunities for local small- and medium-sized manufacturing firms, we must understand where we are heading in light of recent national attention, considering that SME firms are the backbone of US economic growth [11]. Beyond economic growth, the adoption of smart manufacturing and advanced technology in SME manufacturing firms will also ease sustainable production practices. With increased operational efficiency, firms can optimize resource allocation, reduce resource waste, improve energy efficiency, and support cleaner environments for employee health [15,16,17]. As such, the adoption of smart manufacturing is crucial not only for national economic competitiveness but also for firm-level production, employee health, and environmental impacts.

However, SMEs may lack the resources and visibility to participate sufficiently in the research community, thus being underrepresented in the holistic understanding of technology adoption in manufacturing. Therefore, this study focuses on describing the experiences and perceptions of individuals in SME manufacturing firms. To our knowledge, an investigation into the perceptions of these individuals does not yet exist, and understanding the needs of this population will inform workforce development strategies and models to sustain the currently globally competitive workforce.

This qualitative study reports on the workforce challenges faced, training needs, and recognized opportunities of small- and medium-sized manufacturing firms for increased technology adoption.

This study is motivated by an increased demand for AI and yet increasing challenges to adopt AI. The acceptance, adoption, or use of technology in society, across multiple sectors and industries, has long been studied under different theories. One prominent theory that has been shown to be reliable and widely applicable [18] is the Unified Theory of Acceptance and Use of Technology [18,19]. We will investigate how individuals within SMEs perceive workforce challenges, training needs, and opportunities for technology adoption. The Unified Theory of Acceptance and Use of Technology (UTAUT), at its core, is a human action theory that broadly characterizes attempts to explain “why we do what we do” by a number of factors or variables [20].

This objective to explain why some people adopt AI, and some do not, is often understood more broadly under diffusion of innovation theories, which help understand how and why new ideas and technologies spread [21]. For instance, in the advent of AI, how or why does AI spread and diffuse into some communities, cultures, and contexts over others? This occurs over time among individuals in a social system, and as such, we need a theory to understand what factors determine if an individual adopts technology (i.e., AI). Consequently, UTAUT helps us characterize factors that lead to adoption or explain why adoption may not occur.

This study investigates the adoption of technology qualitatively. We seek to understand individuals’ experiences and perceptions as they relate to new technology being integrated into their work. This qualitative paradigm affords us the unique capacity to make reasons for human action perceivable [22]. This paradigm can bridge concrete—real—experiences of individuals to meaningful abstract insights under acceptance theories of the world, such as UTAUT [23]. Whereas quantitative research can let researchers predict certain outcomes to a degree of generalizable confidence, qualitative research reaches into the perceptions of people to inform future action. “Qualitative data are useful when one needs to supplement, validate, or illuminate quantitative data gathered from the same setting” [24]. As such, we leverage UTAUT as an accepted theory of technology adoption and as a guiding framework to qualitatively explore technology adoption for a critical population. The theory’s constructs of performance expectancy, effort expectancy, social influence, and facilitating conditions [19] provide theoretically meaningful interpretations of individuals’ raw perceptions of AI within their SME manufacturing context.

What follows is a brief overview of Industry 4.0 and manufacturing, challenges in adopting technology, and then more detail on the UTAUT framework and its operational use in this study. We conducted a qualitative case study in a natural context to consequently drive insight into the perceptions of technology adoption for local stakeholders. Our study of SME manufacturing specifies professionals and experts who are motivated to transform their operations with advanced technology. These findings provide insight into what may be promoting and inhibiting technology adoption for SME manufacturing firms in Indiana. In turn, this can inform business leaders, project managers, policymakers, and engineers moving forward in the attempt to upskill manufacturing with AI systems.

As such, the research questions are as follows:

• RQ1. How can SME manufacturing professionals’ naturalistic responses be interpreted under the UTAUT framework for understanding technology adoption?
• RQ2. What are factors that can promote or inhibit technology adoption?

2. Background Literature

2.1. Technology and Industry 4.0

Across the scholarly reports on the industry, the term Industry 4.0 is often used to embody mass production. And in the manufacturing sector, this represents the integration of several technologies like AI, the Internet of Things (IoT), cloud computing, and cyber-physical systems [25]. Depending on the context, Industry 4.0 is closely connected to mentions of “smart manufacturing” or AI use within manufacturing. Smart manufacturing can be seen as the second generation of “intelligent manufacturing”, which was originally conceptualized as the intersection of AI and manufacturing. In the past, AI used to be more reliant on human knowledge and structured production systems like if–then decision trees. Now, in the advance of smart manufacturing, AI is data-driven [26]. This encapsulates the use of AI across data capture at the shop floor to optimize supply chain decisions, along with a variety of other use cases to optimize business output [27]. The terms “Industry 4.0” and “smart manufacturing” may be too close in definition to distinguish [11], though perhaps Industry 4.0 connotes more focus on the social effects of technical systems. Nevertheless, both of these terms will be used to explain the digital transformation of manufacturing.

In the past few years, Indiana manufacturing firms have increasingly understood that Industry 4.0 is a good investment. From 2020 to 2022, Indiana manufacturing firms’ perceptions that Industry 4.0 is a “good investment for growth” grew 82%. Moreover, similar levels of growth were present for questions about the success of implementing new technology and decreasing uncertainty about Industry 4.0 technologies. As reported in 2022, the motivations for implementing these technologies include increased efficiency (decreasing cost), improved quality (increasing consistency), increased production speed, and eased difficulties in the amount of skilled labor necessary [28]. Moreover, Indiana’s experience with Industry 4.0 adoption is not unique. Rockwell Automation reported similar metrics in a global survey on manufacturers’ needs for technology adoption. For instance, a 50% increase year-over-year in smart manufacturing adoption and a total of 60% of all companies have fully integrated smart manufacturing [29].

The technologies being implemented help firms at the system level (total throughput and system planning), machine level (predictive maintenance monitoring and optimizing conditions), and material processing level (optimizing material choice and properties). Moreover, at the parts level, AI can optimize materials and diagnose faults and errors. And in supply chain and line process engineering, AI can predict outputs, optimize production, and reduce operating costs. Industrial IoT devices can be applied across all these levels, helping to collect data for downstream analytics [30]. For instance, Conexus reports that Indiana manufacturing firms use Industry 4.0 technologies to support additive manufacturing, cybersecurity measures, machine vision and modeling, and general big data analytics [28]. These are not cut-and-paste deployments, though. Each firm faces challenges of costly implementation and resource allocation, lack of skilled labor to deploy the technology, difficulty integrating with legacy systems, and employee resistance. Moreover, employers struggle to find skilled employees to circumvent these needs and train their current employees [29].

2.2. Manufacturing and AI Technology Adoption

Challenges will be similar for small- and medium-sized enterprises (SMEs), but there is also less awareness of Industry 4.0 for SMEs, more fear of displaced employment, and less robust infrastructure to implement these technologies [31]. Moreover, trying to implement AI introduces unique challenges. For instance, one major challenge for AI is risk management. Companies may not understand how to handle the data they collect; they may worry about data breaches and losing an “edge” over competitors. Also, the nascency and complexity of AI systems make it harder to control for unplanned errors, waste, and abuse. It is often warranted to minimize that potential risk and keep the current—well-known—systems in place [32].

Ghobakhloo et al. [33] reviewed factors for success and failure in information technology (IT) adoption in SME firms. They found that SMEs need to reduce (i) internal factors, such as organizational and resource mismanagement, and (ii) external factors, such as product complexity, labor unavailability, and competitive pressure, to adopt new technology successfully. The characteristics of the “new technology” also impact adoption; Venkatesh [34] applied his UTAUT theory to examine how AI tools uniquely impact the propensity of firms to adopt this new technology. The findings demonstrate that UTAUT is still a useful theory to understand the adoption of new AI technology and its unique characteristics beyond traditional computing. Then, in an empirical paper, the social culture surrounding AI perception (e.g., aversion to AI) is a significant factor in determining technology adoption [35]. Similarly, empirical results from Hasija and Esper [36] find that AI tools need internal trust at the firm. In other words, employees need to trust the technology and build a culture that views AI as an assistant to their employment rather than a replacement. To determine the readiness of an un-transformed company to adopt smart manufacturing practices, Nimawat and Das Gidwani [37] find that companies should begin to implement new practices in small chunks at first to adapt to AI. Then, in later stages, undergo a full digital transformation. They also provide a 16-part “readiness checklist” that can help SMEs determine how far along they are on their transformation toward Industry 4.0.

Diffusion of innovation theories can help us understand and explain how, why, and to what extent new ideas and technologies spread. At its core, the theory of diffusion of innovation distinguishes between four main factors: the innovation itself, the communication channels through which people or media transmit it, the temporal placement of the innovation, and the social system it inhabits [21,38]. Moreover, there is a depth to these characteristics that can be explored and explained by additional theories. One theory that builds on this core is the UTAUT model, which helps understand the determinants of usage intention (performance expectancy, effort expectancy, social influence, and facilitating conditions [19]). The UTAUT theory and its constructs will be discussed later in this paper, as the authors used this as a guiding framework for this study.

3. Theoretical Framework

The UTAUT model integrates multiple theories to unify an understanding of what determines the acceptance and use of information technology. The theory explains the individual intention to use new technology. It posits four primary constructs that account for the most variance in usage intention: performance expectancy, effort expectancy, social influence, and facilitating conditions. These four constructs can be moderated by variables such as gender, age, experience, and voluntariness of use, but those constructs are beyond the scope of the present research study [19]. As the four primary constructs help explain why an individual adopts—or does not adopt—a technology, this theoretical framework structures our investigation into the promoting and inhibiting factors of small- and medium-sized manufacturing firms adopting advanced technology or not.

We utilize the UTAUT model to provide empirically backed constructs as we interpret participant responses. As such, this framework helps explain the qualitative findings of our study. It frames them in light of the validated UTAUT model—bridging the subjective experience of our participants to established literature on the topic.

In this section, we will conceptually define each of the four constructs with central conceptual definitions and root constructs in their conception. Later, we will operationally define them based on the context and language resulting in our dataset.

First, performance expectancy is the degree to which an individual believes using a new technology will help them attain greater job performance. In the original Venkatesh et al. [19] paper that empirically validated these constructs, performance expectancy was the strongest predictor of usage intention. Affiliated constructs (similar in definition) that make up performance expectancy include how an individual perceives the usefulness of technology, whether a technology is a good job fit if they will attain an advantage relative to the replaced technology, and the outcomes they will expect from the technology.

Second, effort expectancy approaches the general question: how easy will it be to use or implement the new technology? Understanding this concept involves two adjacent constructs: the perceived (or actual) ease of use and the complexity of new technology. The degree to which an individual believes that using a technology will be easy or difficult, combined with the technology’s complexity, significantly predicts usage intention. Moreover, as Venkatesh et al. [19] found, effort expectancy is only significant during the first period of technological adoption; as time with the technology progressed, effort expectancy became a negligible predictor of the overall intention to adopt new behavior to use technology.

Third, social influence considers the individual’s network of peers, community, or other social factors that can persuade an individual’s behavioral intention. This is defined as the degree to which an individual perceives that peers important to them believe they should use the new technology. The central idea here is that an individual will be influenced (positively or negatively) by how they believe others will view them after using the technology. The construct’s ability to predict usage intention is contingent on multiple factors, such as whether the individual is in a mandatory or voluntary system, the compliance mechanisms in place in a social system, and the subjective pressure they experience over time that can increase or decrease [19].

Fourth, facilitating conditions capture the organizational and technical context in which organizations implement technology. This is the degree to which an individual believes there is support in the organization for supporting the use of new technology and whether the necessary technical infrastructure exists to support it. Adjacent constructs such as compatibility and perceived behavioral control are part of facilitating conditions. These constructs help us understand how a new technology aligns with existing values and an individual’s self-efficacy has over their ability to implement and use a new technology. For instance, the degree to which an individual knows they have the technical resources available and organizational support to use and adopt the new technology [19].

Moreover, factors such as the age and gender of an individual impact each of these constructs uniquely, as well as the experience an individual has and the voluntariness under which they will adopt the technology (rather than a mandate). These are not within the scope of the present study. Also, as with other theories that predict and explain behavior, many other factors can be used, such as the degree of habit alignment, pleasure from adoption, moral agreement, profit expectancy, etc. Nevertheless, researchers identified the four constructs listed above as the direct determinants of the acceptance and use of technology; thus, other constructs are not within the scope of this study.

4. Materials and Methods

This study followed a qualitative research paradigm with the intent to provide insight into the minds of a subset of the population who, from empirical records, struggle to adopt AI technology. Qualitative research aims to make meaning out of the subjective experiences of a population of interest. In turn, the “meaning-making” is largely informed from a theoretical perspective about what knowledge means in context [39]. Qualitative research has been widely accepted as a valid approach to investigating sustainability. For instance, in [40], researchers used qualitative thematic analysis in interviews with 17 master’s students to investigate knowledge management strategies for research innovation within universities. Similarly, in reference [41], the investigators conducted interviews with 15 stakeholders to document their perspectives on industry needs as related to competencies needed from graduates in the sector. Although findings from such qualitative studies were subjective, they were valuable in each corresponding study because the findings were detailed and highly contextualized into each particular setting or issue.

In the current study, as described, we used qualitative research methods to apply the UTAUT framework to explain the adoption and acceptance of technology in real contexts, such as the experiences described by our participants. Although UTAUT has been primarily applied in via survey studies, it has also been used as a conceptual framework to analyze and interpret participants’ perspectives on their adoption of technology. For instance, in [42], researchers used qualitative methods to investigate the experiences of 15 researchers in their adoption of bibliometric management tools, providing unique insight to the population and informing training programs to develop in future work. Our study is positioned with the same intent, to provide rich insight to our population to complement existing development opportunities.

To conduct our investigation of UTAUT, our study engaged in a retrospective design. The authors were part of organizing a stakeholder engagement meeting for local manufacturing professionals, and after the meeting, we recognized the research salience and practical implications of studying the meeting itself. The conversations during the meeting itself were not motivated or influenced by an ulterior research motive. As such, these conversations took place naturally. As the research team was present in the meeting, we believe that our privileged—or elevated—research perspective has a unique capacity to do justice for this population. We recognize inherent bias and hidden influences that may exist—after all, all human conversation is influenced by hidden objectives—and constantly recognize this in analysis. In an earnest review of our positions, we evaluated this research project as worthy, sincere, significant, and coherent in its conception and conclusion [43].

4.1. Context and Participants

The context of this study was a stakeholder engagement meeting conducted in August 2023 that brought manufacturing industry professionals, community and government partners, and research faculty and students in order to promote the adoption of advanced technologies to support digital transformation. The event was part of a manufacturing university–industry partnership program at a large midwestern R1 research university. A total of 74 participants attended the workshop, and 32 of them were manufacturers, including managers, engineers, line workers, and floor supervisors from 23 different companies. The rest of the participants were community representatives, staff from the state senator’s offices, and faculty, researchers, and graduate students from three different universities. Attendance was voluntary as the meeting intended to engage with the community of local manufacturers and share knowledge about their successes, challenges, and opportunities that they see between themselves in this network of manufacturers. The research we report on in this study was a byproduct of the meeting and was not an organizing structure or motivation for the meeting. Thus, participants were not compensated for their time. Moreover, before the research team received the participant workbooks, a non-research facilitator provided anonymized copies to protect participants’ privacy and confidentiality. Consequently, we are unable to provide detailed participant profiles for our data.

4.2. Data Collection and Preparation

We collected data at a stakeholder engagement meeting with 74 participants, including industry professionals, university faculty and affiliates, and state organization members. During this meeting, four local manufacturers working in a university–industry partnership with the research team presented their experience with recent smart manufacturing deployment. These manufacturers were part of a larger effort to upskill the region with AI technology. They were situated in a larger ecosystem where academia, national and local policymakers, and dedicated technology providers are playing a part in the adoption of AI technology for the sustained competitive advantage of U.S. manufacturing. As such, the professionals and experts in the meeting were at different stages in their digital maturity with AI and presented on their progress, challenges, and future opportunities.

After these main presentations, AI-in-manufacturing experts facilitated discussions, and small groups discussed challenges and shared recommendations for implementing advanced technologies. During these events, participants filled out workbooks that included questions such as: What are your fears, concerns, or barriers to the implementation of digital solutions? What are training needs at various levels? What do you see as future opportunities for the creation of this community of manufacturers? While some answers were complete and thorough, answers often took the form of free-form notes, incomplete sentences, and bulleted lists. Thus, as data were prepared for analysis, the research team had to balance maintaining participants’ response integrity and transforming the data to fit into an analyzable structure (the content of the participants’ responses was never changed; only the structure of how they wrote them was). These responses were transcribed manually and then imported into ATLAS.ti (version 23.4.0) [44] to facilitate reading responses, coding data, conducting analysis, and organizing findings.

While preparing the data for analysis, we identified certain data that did not meet our criteria for analysis, ending with 36 qualitatively sufficient participant workbooks. To explain, from the 74 participants who attended this workshop, we expected to receive 63 workbooks, but only 51 were collected (81%). That is, not all participants were expected to fill out their workbooks completely, as, for example, university faculty and students might not have direct experience or access to the manufacturers’ operation and installations. When reading the data and preparing it for analysis, we found some responses were sporadic, incoherent, illegible, incomplete, or irrelevant to the topic at hand. Thus, considering our qualitative scope, 36 were determined to be sufficient data for analysis.

4.3. Analysis Approach and Design Rationale

Qualitative analysis procedures can range from highly inductive and emerging, where data are disassociated from accepted theories, to highly deductive, where data are considered in close conjunction with accepted theories of the world. Regardless, the qualitative analysis paradigm we accept follows reference [24]’s view, where all data are condensed into manageable units, then displayed (i.e., organized) to examine relationships and patterns in the data, and finally interpreted in the bigger picture to draw conclusions [24].

Our methodological procedure for qualitative analysis primarily followed deductive analysis under our theoretical framework and research questions [45]. First, as we identified the UTAUT framework as theoretically meaningful to our population and problem context, we constructed four qualitative categories correspondent to UTAUT. Second, we began to initialize the data for analysis [46]. We read our data with consideration of our research questions and constructed a dual-coding framework where a positive or negative label was applied to each qualitative code. We chose to do this since we could capture that participants were speaking about some aspects of advanced technology as proponents of adoption and some aspects as inhibitors of technology adoption. Third, under this deductive coding framework, we began to parse each participant record carefully, taking reflective notes along the way as is best practice [24]. During the coding procedure, the research team began to stabilize the knowledge and generate operational definitions of how each theoretical construct was present in the dataset [46]. Then, once all coding was complete, we had an organized and structured dataset of qualitative codes corresponding to our theoretical framework and research questions and in close alignment with our data. This qualitative procedure was conducted by trained researchers who had unique positional expertise to understand challenges and opportunities for technology adoption.

As a result, we conducted final data analysis by a code co-occurrence analysis that analyzed the relationship between each of the UTAUT constructs and how they were referenced as positive or negative factors for technology adoption. In turn, analyzing the UTAUT constructs in this way gave us unique insight into workforce training demands and closed the gap between understanding AI demand and workforce supply. This analytical procedure lets us draw implications for training programs and organizational policy for SME manufacturing firms, which are later discussed in the discussion.

4.4. Team’s Research Positionalities

The team’s research positionality followed a pragmatic approach as opposed to a positivist approach. While the positivist approach emphasizes objectivity, quantification, and hypothesis testing, pragmatic approaches are more flexible in their methods. A pragmatic approach to research was deemed necessary because it prioritizes practical outcomes and the applicability of research findings to real-world problems. By doing so, this approach allowed us to integrate both qualitative and quantitative methods. In our study, this integration occurred by quantizing the qualitative findings for visualization.

The team composition was also important for this work. Specifically, the team was composed of two senior faculty members, one expert in manufacturing and another expert in education and training. The team also included two graduate students pursuing their doctoral degrees in computer and information technology. The expert in manufacturing was part of the organizing committee and took a leadership role in facilitating the event. He facilitated the access and anonymization of the data before the analysis. The expert in education and training and the graduate students attended the event as participants. The expert in education and training led the research study. One graduate student took a lead role in analyzing the data, and the second graduate student supported the analysis and validation of the findings. The expert in manufacturing was not involved in the analysis of the data but assisted in the interpretation of the findings and their practical applicability. These roles were critical not only to bring diverse expertise but also to ensure participants’ privacy and confidentiality.

5. Results

We found each construct of social influence, effort expectancy, facilitating conditions, and performance expectancy displayed across the dataset. As the participants wrote about the strengths, weaknesses, challenges, opportunities, and other views of new technologies in manufacturing, they also expressed deeper beliefs with respect to the UTAUT model. Their responses contained elements of each construct. As we qualitatively analyzed the data, we pieced together how manufacturers unknowingly treat each construct—unique to the contexts of the manufacturing industry itself. While our data cannot predict whether manufacturers will adopt new technology, like the function of the original UTAUT model, we framed each response by each construct to better understand participants’ beliefs and ideas about how certain factors impede or promote their adoption of new technology.

The structure for subsequent sub-sections in Section 5 first discusses the high-level synthesized findings from participants’ responses. Then, it highlights specific participant quotes about each construct. Note, however, that participants’ responses are pretty fragmented as these workbooks for data collection were utilized in quick fashion—to take short notes and jot ideas—rather than expecting complete sentences from the participants. Due to this fragmented nature of responses, listed after participant quotes, we occasionally included an interpretation as a first-person subject with respect to the UTUAT model to assist in understanding and framing responses with extra context from the workshop. Please refer to Figure 1 for a sample showing how participants occasionally wrote in bullet-point and sentence form and freely structured their responses as they saw fit.

5.1. Social Influence

In this study, various statements around sharing stories of successful implementation, challenges during implementation, and communication within a community of manufacturers highlighted the social influence construct of the UTAUT model—this construct, as defined by Venkatesh et al. [19], involves how individuals’ perceptions of influential social groups can impact technology adoption decisions. We found this construct present in 75% of participants. The influence of organizational culture, both within a firm and across firms, also emerged as a significant factor, recognized by participants as a critical driver for technological adoption. The sub-construct of social factors was prominently expressed in participants’ responses, reflecting the internalization of cultural norms and interpersonal agreements between peers and leadership. Although other sub-constructs, such as image (can using this technology enhance my social image?) and subjective norm (do peers close to me suggest using this technology?) were present in the responses, they were not as dominant as a reference to culture and general social factors within the firm and across firms. For example, participants frequently noted a desire to share stories amongst a community; they expressed undertones that state, “I use the system because of the proportion of coworkers who use the system” [19]. As a result, participants who responded along variations in this sentiment were coded under the social influence construct. Moreover, the majority of social influence expressions carried a positive sentiment that this is a promoter of technological adoption.

Positive and Negative Participant Remarks

When discussed as a promoter (87.5% of the time), the following remarks were expressed:

• “What is success: get away to meet other manufacturers to meet regularly and talk in detail on solution”.
• “Community could speed process of starting up and integrating”.
• “Collaborative network to share industry knowledge”.
• “Share: ROI stories, DIY work, vendor feedback”. [Sharing stories of success and receiving feedback from the community are social components that promote adopting new technology].

These featured quotes represent only a fraction of the responses but are generally clear and indicative of expressing factors of social influence. Participants consistently linked the adoption of technology with social and community factors. They frequently expressed a desire to establish a sense of community, share personal experiences, and receive feedback from various stakeholders within the industry.

When discussed as an inhibitor (12.5% of the time), the following remarks were expressed:

• “Threats: perception and fear of job loss”.
• “Still missing employees; fear”. [Employees share a fear of new technology, inhibiting the adoption of new technology].
• “[University partners] thinking they’re experts in everything with closed minds and constantly flapping lips”. [Increased collaboration might impede my firm’s tacit knowledge and cause friction to adoption].

In these responses, participants discussed the social influence construct with factors that may be considered inhibitors to technological adoption. These types of responses comprised an eighth of the total and typically centered around human factors. Fear, threats, and perceptions of job loss might impact employees’ shared beliefs about whether it is advantageous for them to adopt the new technology. This does not mean that “community is bad” but rather that participants viewed these human elements of fear, threat, and perception as a stronger negative influence than that of a positive social influence from the community.

5.2. Effort Expectancy

Effort expectancy was primarily discussed among the participants in two ways: either the new technology was perceived as too complex, and the current, simpler methods were preferred, or adopting the technology took away too much time from employees and too many resources across the firm. We found this construct present in 75% of participants. Overall, participants rarely discussed scenarios where adopting advanced technology would inversely reduce effort and extend resources in the firm. For instance, in the UTAUT model, the perception of effort can go both ways, where an individual can perceive the ease of use of the technology. In contrast, others perceive that technology takes too much time away from regular duties. Moreover, as this construct is more salient at the early stages (or pre-stages) of adoption, negative references to effort expectancy might be prevalent due to the majority of these manufacturers having yet to implement technology. Meaning, effort expectancy becomes less relevant as technology adoption progresses [19]. Regardless, effort expectancy was dominant during our data collection period as an inhibitor of technology adoption across our participants.

Positive and Negative Participant Remarks

When discussed as a promoter (13.9% of the time), the following remarks were expressed:

• “Take away difficult and boring tasks; Improved productivity, profit leads to growth”.
• “Showing how technology can make their job easier or better”. [I think technological adoption can expect to reduce the effort required of employees].
• “More technology improvement to wireless (to make it more flexible, easier, and more robust)”.

Positive references to effort expectancy and technological adoption were rare. These highlighted quotes, however, interpreted technological adoption not as a large barrier to overcome, but one to conquer to solve other technical, business, or human barriers. Moreover, these responses indicate that new technology aids human processes; it can take away cognitive load or reduce physical strain.

When discussed as an inhibitor (86.1% of the time), the following remarks were expressed:

• “Change is difficult, relies on trust. Takes time. Iterations”.
• “Who will do the work. It takes someone to drive this, and everyone is busy”.
• “We are not lean we are anorexic. Why and what do I get for doing? Perception of complexity of implementation”.

The effort required to implement new technologies was seen as an inhibitor of technological adoption. New technology is too complex, takes too much time, and pivots the firms’ priorities away from bottom-line factors. Participants cited the difficulty of change, sometimes in conjunction with social factors like cultural change, and that change takes time and iteration. Moreover, change may never happen due to the lack of resources spread across time—no one to lead the work and no one to do the work. As the last quote expresses, these firms might be operating on the bare minimum, and even figuring out what the new technology is a challenge. (Still, it is worth mentioning that all of these participants voluntarily attended our workshop on technological advancements in manufacturing.).

5.3. Facilitating Conditions

As the most represented construct in this study, found in 83.3% of participants’ responses, participants perceived the organizational structure as a factor in technological adoption, even if they believed they could adopt new technology. Participants stated three main topics. First, many participants centered their response around organizational buy-in, where buy-in referred more to the latent agreement of upper-level personnel in the firm. The participants (commonly managers themselves) viewed management as not having “bought-in” to the technology yet; not seeing the benefit to the technology yet. Second, financial factors were cited as impeding technological adoption. These mentions included not having the resources within the firm to support new technology or not seeing the return on investment from adopting new technology. Third, the training requirements were perceived as a significant factor. Participants frequently noted that, upon adoption, training at all levels will need to occur to use, operate, and manage new technology properly. Thus, participants frequently observed that the “training infrastructure” was not in place within the firm to support new technology. Note that some participants expressed positive opinions about education and training efforts. In contrast, others expressed the lack of it as a negative.

Moreover, in the UTAUT model, facilitating conditions encapsulate factors from both the technological and organizational environments [19]. As such, a few participants expressed facilitating conditions as having the proper technological infrastructure—ensuring that data storage, security, and sharing are appropriately handled—in contrast to just needing the proper organizational infrastructure. For firm-wide adoption, it is crucial to consider the facilitating conditions of human factors like training, business factors like organizational buy-in and communication lines, and technological factors like having suitable technical systems to integrate new technology.

Positive and Negative Participant Remarks

When discussed as a promoter (20.5% of the time), the following remarks were expressed:

• “We are working on a model to educate manufacturers on how to drive their own digital transformation and surround them with independent coaches and mentors”.
• “Strengths: including operators converts them easily”. [I think that flattening the organizational structure would support adoption].

These samples of quotes show how facilitating conditions can be a positive factor for technological adoption. For instance, an organizational structure that supports training and education needs can not only lead to technological adoption but also firm-wide growth through a support system. Moreover, including floor workers in the adoption process can help convert (train) them on new technologies—leading to facilitated growth.

When discussed as an inhibitor (79.5% of the time), the following remarks were expressed:

• “What is the challenge? management”.
• “We have a legacy system that tracks uptime, counts, fault codes, but doesn’t lend itself to deeper insights like predictive maintenance or energy efficiency”.
• “Need buy-in from companies to change their culture so [new technology adoption] can happen”.
• “Executive-level training to form buy-in”. [I think that lack of training is causing resistance to buy into this new technology].

While facilitating conditions might be recognized as a positive factor in adoption by some, this construct was most commonly discussed as a challenge, barrier, or inhibitor for adopting new technology. As seen in the first highlighted quote, one participant simply expressed their challenges in one word: management. This challenge prevailed across multiple participants, who expressed difficulty converting the organizational structure to facilitate technological adoption. Moreover, technological systems already in place at the firms were a negative-facilitating condition. A few participants referenced that their legacy systems make it hard to integrate technological adoption into the organization (overlapping with the effort expectancy construct). These difficulties also all express the need for buy-in; when the organization is on board with the technology, conditions within the firm may fall in place to support adoption.

5.4. Performance Expectancy

In this study, participants expected both micro-level performance improvements and macro-level performance improvements from adopting new technology. For instance, many participants spent their time writing about data analytics and machine optimization improvements. In some cases, viewing the new technology is the chance to better predict machine failure, downtime, and optimize preventative maintenance windows. We found this construct present in 63.9% of participants. Moreover, overall equipment effectiveness (OEE) improvements were a metric cited alongside the performance expectations of the new technology. From a macro-level perspective, participants expressed performance expectancy in terms of the firm. For instance, participants viewed the firm’s return on investment as contingent on the expected performance of new systems. Moreover, participants perceived that other employees (from floor to executive level) did not see the benefits of new technology, like being blinded by needing to overcome the initial adoption phase.

Out of the four UTAUT constructs, performance expectancy was referenced in the most balanced nature of being a promoter or inhibitor to technological adoption. Some said that they could clearly see the performance improvements (driving adoption) while. In contrast, others did not see the same improvements—acting as an inhibitor to adoption. In the UTAUT model itself, Venkatesh et al. [19] found that performance expectancy was the strongest predictor of adoption. However, in this study, this was discussed the least among respondents. Of course, we cannot extrapolate the participants’ choice of responses as predictors for technological adoption; rather, the fact this was least coded might imply that participants viewed other parts of technological adoption as warranting the most space in a conversation. As discussed at the start of Section 5, Results, facilitating conditions was the most expressed construct, whereas performance expectancy was the least.

Positive and Negative Participant Remarks

When discussed as a promoter (55.5% of the time), the following remarks were expressed:

• “Strengths: incredible value when implemented”.
• “Strengths: predict failure and prevent downtime”.
• “Strengths: seeing labor and imaging quality and decreasing cost”.

These highlighted quotes indicate that the expected performance-related outcomes are positive factors that contribute to technological adoption. As discussed above, these are either discussed through a macro lens, such as stating there is “incredible value” of the technology (that drives motivation to adopt) or that long-term costs will be decreased. In other responses, participants focused on microlenses. They viewed the advanced technology as improving machine metrics like OEE, predictive failure, downtime dashboards, and day-to-day data analytics for operators. While participants did not often go into detail about performance improvements and kept their statements vague, they were not uninformed. They did not lack comprehension of the benefits.

When discussed as an inhibitor (44.5% of the time), the following remarks were expressed:

• “Data sharing is a huge barrier. No one gets fired for going with IBM. Theoretical ROI is fine, but real data is really the thing to help get the board of directors on board”. [I cannot convince the board of directors of expected performance improvement before adoption].
• “Why am I doing it. What am I getting back in return”. [I don’t perceive enough expected performance].

While many participants recognized how expected improvements to the firm’s performance drive technology adoption, many (about half) still cited performance expectancy as inhibiting adoption. For instance, a participant may understand the performance benefits but not deem them large enough to warrant implementation. As the first highlighted quote suggests, this might be because the return on investment from the technology is still theoretical—it is not yet real enough to convince upper-level management. Likewise, some participants do not expect performance improvements to have practical career benefits. The second highlighted quote suggests that participants lack a personal return. In other words, “Yes, I know the new technology is the future, but why am I doing it?”

5.5. Overlapping Constructs

Note that effort expectancy frequently co-occurred with other constructs like facilitating conditions and performance expectancy. For instance, as participants expressed technology adoption being difficult due to the lack of excessive time and resources that advanced technology requires, they may also be referencing that their firm’s facilitating conditions do not allow for flexibility, training, or the budget to implement new technology. In some cases, these constructs go hand-in-hand—the firm’s lack of adoption creates a perceived difficulty in using technology, which supports the firm’s resistance to creating an organizational structure to adopt the technology.

One manufacturer said, “we are working on a model to educate manufacturers on how to drive their own digital transformation and surround them with independent coaches and mentors” and finished their responses by brainstorming both university–industry and industry–industry collaboration. In that response, both social influences and facilitating conditions can lead to promoting factors for technological adoption.

6. Discussion and Implications

Both manufacturing firms and cross-sector strategic initiatives to support AI in manufacturing can use these results to design and structure advanced technology deployment. The results show that each of the UTAUT constructs factors into participants’ propensity to adopt or reject technology adoption. By designing environments to support positive factors for adoption and minimizing the effects of negative factors, employees may transition into using new technology more smoothly with less resistance. Refer to Figure 2 and Figure 3 to see the distribution of positive and negative factors for technology adoption. As shown, the results indicate that social influence factors are most dominant in positive factors for adoption. This finding is consistent with the literature on extending the UTAUT model for advanced technologies and AI adoption. Jain et al. [35] and Hasija and Esper [36] found that adopting AI technology is dominantly impacted by cultural perception and peer trust. Inversely, the effort required to adopt and use new technology is the most mentioned inhibitor for technology adoption. This observation is consistent with findings from manufacturing surveys conducted in the past few years [28,29]. Naturally, these converging findings implicate researchers with salient future directions to pinpoint operational techniques—both in business and policy—that can reduce the effort and cognitive load to use AI and engage local communities to facilitate peer trust and normative influence.

Also shown in Figure 2 and Figure 3, performance expectancy is the least referenced UTAUT construct (albeit just marginally less than social influence). This low reference level may seem odd, considering the well-recognized potential of AI to improve performance. The research team speculates that since people are already aware of the benefits of technology adoption for performance, they do not discuss it extensively. However, we suggest future work to dig deeper into the dynamic of how the expected performance of AI systems impacts the salient considerations in practitioners’ minds. This has, in part, been explored in a study that utilized the UTAUT model to characterize blog technology adoption. Pardamean and Susanto [47] found that “both social influence and performance expectancy had a significant relationship with behavioral intention, whereas effort expectancy did not” [47]. While the results in our study do not infer or predict behavioral intention based on UTAUT constructs (instead, we describe participant perceptions), the findings from our study suggest that participants do not perceive performance expectancy as needing the same space in conversations as do other constructs. Our qualitative findings can be considered alongside other quantitative studies, though. Moreover, as stated, we did not collect age from the participants in our workshop. In the UTAUT model, we recognize these are moderating variables and should be considered as underlying effects when participants express perceptions of technology adoption. Similarly, cultural orientation would be inferred to impact how the influence of peers impacts technology adoption and should be considered in future work.

Other empirical work to explore challenges faced by SME manufacturers complements our findings. Kopp et al. [48] conducted a survey of 89 German companies and found that if management and workers have different perspectives and opinions toward technology integration, there is a larger challenge in adopting the technology. In our study, we found that facilitating conditions like lack of executive buy-in, lack of sufficient financial resources, and inadequate employee training are much-cited factors that impede the adoption of advanced AI technologies. Kopp et al.’s [48] results share our study’s immediate implications. Organizational change should take place where managers and floor workers talk about AI implementation, share their perspectives, and work toward shared goals that benefit all employees. This is perhaps the most concerning implication for AI practitioners and business leaders. Engaging participation and stakeholder deliberation is a widely recognized technique for improving the efficacy and efficiency of tools for societal use [49]. In AI development, this public participation has been proposed to take the form of online websites where people can evaluate the decisions of AI/ML models to refine and improve algorithms [50]. We suggest this tool could be used locally within manufacturing firms to easily source firm feedback through a centralized and standardized source. However, considering that employees expressed concerns with the level of organizational buy-in, we need to recognize accessible methods for leadership teams that have practical benefits down the organizational chain. Moreover, practitioners would need time and expertise to sift through public comments since the variance in opinions on AI makes it challenging for policymakers and leadership to generate effective AI strategies [51]. Nevertheless, improving the facilitating conditions around AI adoption is crucial for successful implementations.

Furthermore, Rossini et al. [52] found that companies with higher implementation of lean production techniques have higher levels of Industry 4.0 adoption, and both levels of adoption lead to higher operational performance. Relevant to our study, they suggest that if companies aim to have successful technology adoption, they should concurrently consider implementing lean production techniques to complement the process [52]. Lean production is a broad production paradigm that aims to reduce waste and optimize value-adding activities [53]. As manufacturers are pressured to optimize their operations and minimize environmental impacts, lean production has been a natural integration [16]. Empirical investigations of how smart manufacturing, Industry 4.0, and lean production have shown that the more lean production and smart manufacturing a firm can implement, the more likely they are to survive [16]. In this, we found that SMEs face similar challenges to lean production. SMEs report that a lack of time, lack of management support and buy-in, and lack of resources are inhibiting factors to lean production [54]. In our study, we reported similar characteristics along the UTAUT constructs of facilitating conditions and effort expectancy.

In turn, we hypothesize that by fostering the social influences of technology adoption (and, as a consequence, mitigating effort expectancy), SME manufacturing firms will simultaneously contribute to sustainability goals while achieving key operational metrics for their firm’s success. As smart manufacturing continues to require employee participation and expertise across the firm, there should be complementarily facilitating conditions to support peer learning in any small or large manufacturing operation [17]. By adopting sustainable practices (as a consequence of smart manufacturing and lean production), SMEs will enhance their regional and global competitiveness and appeal particularly to environmentally conscious consumers [55]. As a result, working to align technology adoption with the UTAUT constructs while implementing smart manufacturing will give SMEs a strategic advantage.

To reduce effort expectancy and potentially facilitate the facilitating conditions, we suggest engaging in university–industry partnerships where students in advanced classes partner with local companies to solve real problems. First, mediated through experiential learning, students engage in higher-order learning about AI, ML, and Industry 4.0 by actively experimenting at local manufacturing firms. Second, the firms directly benefit from students’ work—reducing the effort required to jump-start new digital transformation work [56]. And, as an added benefit, these university–industry partnerships can provide local manufacturers with local experts to field their uncertainty around new technology. These partnerships require strong connections, which are mutually beneficial for all parties involved but can reduce the effort to adopt technology and form more robust organizational conditions. Thus, in future work, working with both management and factory workers to strategize how to overcome these challenges can be a direct in-road to begin improving operational performance. At the least, it can be a start to implementing AI technologies, and as minor as the start might be, it is worthwhile toward improved operational performance [37].

The practical implications of this study relate to the establishment of regional initiatives that promote advanced technology adoption among SMEs. The workshop detailed in this study was also the basis for the establishment of ManuFuture Today Network [57], which leverages a collaborative peer community to centralize discussions, knowledge sharing, and idea creation to maintain the global competitiveness of its members. Additionally, follow-up meetings were conducted with attendees to inform another community meeting about how SMEs can conceptualize ROI and business cases for their digital transformation. Through iterative cycles of information gathering and assessment of regional SMEs, ManuFuture Today Network identified that the focus of the initiatives should center on (1) people, by engaging employees at multiple levels of the organization, such as floor operators, engineers, and managers; (2) devices and data, supporting the adoption of low-cost technologies and sharing of data; and (3) socialization, by facilitating peer-to-peer learning, coaching, regular community events, and training.

Moreover, we recommend manufacturing firms identify networks of professionals to host training workshops and short courses to help ensure a skilled workforce of their own and maintain their region’s competitiveness. For instance, Fiock et al. [58] detail a case of a short course that engaged local SMEs and evaluated its effectiveness. Engaging stakeholders in a network of manufacturers is one attempt to facilitate larger engagement for AI adoption—hopefully driving the social influence in the community, easing the burden of leadership, and sharing strategies for reducing the effort it takes to adopt new and novel technology.

7. Conclusions, Limitations, and Future Work

Findings from this study emphasize the levels of relevance of social influence, performance expectancy, facilitating conditions, and effort expectancy regarding how SME Indiana manufacturers perceive the deployment and implementation of advanced technologies. From qualitative analysis, findings indicate that participants view social factors as the most critical aspect in promoting technology adoption and the expected effort to adopt technology as the most inhibiting factor. Moreover, factors such as management and organizational structure were viewed as negative factors for technology adoption. However, we recognize that the qualitative nature of our study inherently may pose some limitations to generalizability and replication. Thus, our findings are highly contextualized to the particular problem of manufacturing adoption of technology and the setting of SMEs in Indiana. Also, given the naturalistic setting of the data collection, the participants were left to decide the interpretation of their level of engagement with technology, whether they talked about general technology, automation technology, or AI. Conversely, given that the data were collected from a representative sample of manufacturing stakeholders in the region, we believe that the study provides some in-depth insights into the challenges that the stakeholders are currently experiencing. Furthermore, we believe that our findings may be insightful not only to manufacturing professionals but also to researchers and academics who currently engage in partnerships to support technology innovation. Our findings are connected with recent literature and further emphasize the needed support for the investment of SMEs in manufacturing firms to help facilitate the deployment of advanced technologies. Our future work will continue to document technology adoption processes and highlight the successes and challenges experienced by manufacturing stakeholders in the process.