AI Adoption in B2B Markets: Exploring Drivers and Barriers through Behavioural Reasoning Theory and a Care Approach

Abstract

This study examines AI adoption intention in B2B marketing in the Iranian steel industry by integrating Behavioural Reasoning Theory with a care perspective. A cross-sectional survey of 275 industry respondents was analysed using PLS-SEM and the PROCESS macro. Reasons for adoption were positively associated with attitude and intention, whereas reasons against adoption were negatively associated with both outcomes. Care value was positively associated with reasons for adoption and negatively associated with reasons against adoption; the latter association contradicted H1b. Brand reputation did not significantly moderate the direct relationships between reasons and intention, leaving H6a and H6b unsupported. Additional analyses identified interactions predicting attitude and a positive interaction between attitude and reputation predicting intention. These findings concern reported intentions and associations rather than observed adoption or causal effects.

Keywords: Technology Adoption, Artificial Intelligence, Care Value, B2B Marketing

1. Introduction

Artificial intelligence (AI) is increasingly discussed as a means of developing market knowledge and supporting business-to-business (B2B) marketing and sales activities (Paschen et al., 2019; Paschen et al., 2020). Its potential applications include analysing customer information and supporting decisions across the sales process. Research also identifies ethical, organizational and implementation challenges that complicate adoption decisions (Dwivedi et al., 2021). Understanding both supportive and inhibiting reasoning is therefore relevant to AI adoption intention in industrial B2B settings.

Potential benefits do not ensure adoption. AI implementation involves organizational choices about task allocation, skills and the relationship between automation and human judgement (Raisch & Krakowski, 2021). Research in German firms also highlights the importance of examining AI in a specific industrial context (Rammer et al., 2022). The present study examines how respondents in the Iranian steel industry evaluate reasons for and against using AI in B2B marketing.

Prior studies have investigated AI in public services (Gesk & Leyer, 2022), consumer service encounters (Gursoy et al., 2019), AI-enabled travel agents (Lalicic & Weismayer, 2021), voice assistants (Acikgoz et al., 2023), and B2B customer relationship management (Rahman et al., 2023). These settings provide useful comparisons but do not establish how the same relationships operate in industrial B2B marketing.

Technology adoption research commonly draws on perceived usefulness, ease of use and related acceptance constructs (Davis, 1989; Venkatesh et al., 2003). Behavioural Reasoning Theory offers a complementary framework that distinguishes reasons for a behaviour from reasons against it (Westaby, 2005; Claudy et al., 2015). A systematic review documents applications of this reasoning framework (Sahu et al., 2020). The present study combines this distinction with care value and reputation to examine AI adoption intention among Iranian steel-industry respondents. This contextual application should not be interpreted as evidence that earlier acceptance theories are generally incapable of explaining AI adoption.

The study employs Behavioural Reasoning Theory to examine associations among care value, supportive and inhibiting reasons, attitude and adoption intention (Westaby, 2005). The theory provides the basis for modelling reasons for and against adoption within the same framework. The additional care perspective draws on ethical discussion of AI and its relational consequences (Cohn, 2020; Villegas-Galaviz & Martin, 2024).

Specifically, this study integrates care value into the BRT framework. Rooted in the ethics of care, care value reflects managerial concern for stakeholder well-being and the broader social consequences of technological decisions (Gilligan, 1982; Tronto, 1993). Within BRT, care value is theorized as an antecedent shaping managers’ reasoning, influencing both reasons for and reasons against AI adoption. Given the potential impact of AI on employees, trust, and organizational accountability, considering care value provides a human-centred, context-sensitive extension to traditional BRT applications in industrial AI adoption research (Dwivedi et al., 2023). In this regard, care value captures managers’ attention to issues such as employee welfare, fairness, transparency, and responsibility toward affected stakeholders. It emphasizes relational and context-dependent ethical judgments rather than abstract rule-based principles, making it particularly relevant for complex industrial decision-making. As such, care value offers a meaningful ethical lens through which managers evaluate both the anticipated benefits and potential risks of AI adoption.

This study also examines brand reputation as a proposed moderator. Reputation has been theorised as a signal of organizational standing and credibility (Fombrun & Shanley, 1990; Fombrun, 1996). In a behavioural reasoning application, reputation has been examined in relation to organic food consumption (Ryan & Casidy, 2018). These arguments motivate the proposed moderation hypotheses, but do not establish that reputation necessarily buffers AI-related barriers in B2B settings.

The remainder of this study is structured as follows: the background literature and hypothesis development are presented in the second section, followed by the research methodology, questionnaire design, and data collection in the third section. The empirical analysis is detailed in the fourth section, and the fifth section provides a discussion, summarizes theoretical and practical contributions, and outlines avenues for future research.

2. Literature Review

2.1. Artificial Intelligence in B2B Markets

In B2B marketing, AI can support market knowledge development and collaboration between human and artificial intelligence across the sales process (Paschen et al., 2019; Paschen et al., 2020). These potential applications motivate interest in adoption, while organizational implementation requires attention to the relationship between automation and human judgement (Raisch & Krakowski, 2021). The present study focuses on intentions to use AI rather than evaluating the performance of a particular AI application.

2.2. AI Adoption in Industrial Marketing

The Technology Acceptance Model links acceptance to perceived usefulness and ease of use (Davis, 1989), while UTAUT incorporates performance expectancy, effort expectancy, social influence and facilitating conditions (Venkatesh et al., 2003). Behavioural Reasoning Theory adds an explicit distinction between supportive and opposing reasons (Westaby, 2005). This distinction is useful for examining respondents who may recognise AI benefits while also expressing concerns about its use.

2.3. Behavioural Reasoning Theory

Behavioural Reasoning Theory distinguishes reasons for a behaviour from reasons against it and links these context-specific reasons to attitudes and intentions (Westaby, 2005). It has been used to examine AI-enabled travel service agents (Lalicic & Weismayer, 2021), voice assistants (Acikgoz et al., 2023), and adoption or resistance to AI in hospitality (Rasheed et al., 2023). These applications provide a rationale for examining supportive and inhibiting reasoning in B2B marketing, while the present industrial context requires its own measurement justification.

2.4 Hypothesis Development

This subsection presents the research hypotheses, which are summarized in the conceptual model shown in Figure 1.

2.4.1 Values and Care Value – Reasoning

Within Behavioural Reasoning Theory, values are conceptualized as deep-seated cognitive orientations that shape how individuals and organizations interpret situations and activate specific reasons for or against a behaviour (Westaby, 2005). Values guide evaluative judgments across contexts and influence the prioritization of certain justifications over others (Schwartz, 2012).

Among various ethical and managerial values discussed in the literature, care value emphasizes attentiveness to stakeholders, responsibility for relational outcomes, and sensitivity to potential harm arising from organizational decisions (Gilligan, 1982; Tronto, 1993). Prior research in AI and digital innovation suggests that such values are particularly salient in decisions involving emerging technologies characterized by uncertainty, responsibility, and potential societal impact (Floridi et al., 2018; Dwivedi et al., 2021).

In the context of AI adoption in B2B marketing, care value may strengthen positive reasoning by emphasizing responsible innovation, improved relational outcomes, and enhanced decision quality. Simultaneously, this value orientation may heighten awareness of potential negative consequences—such as data privacy risks, information security concerns, and adverse stakeholder impacts—thereby reinforcing reasons against adoption.

Accordingly, the following hypotheses are proposed:

H1a: Care value positively affects firms’ “reasons for” AI adoption in B2B marketing.
H1b: Care value positively affects firms’ “reasons against” AI adoption in B2B marketing.

2.4.2 Care Value – Attitude

Behavioural Reasoning Theory further suggests that values may influence attitudes directly, independent of their effect through reasoning (Westaby, 2005). Decision-making research supports this view by showing that individuals may form attitudes toward innovations based on value congruence rather than systematic cost–benefit analysis (Claudy et al., 2015).

Care value provides a theoretical basis for proposing an association between ethical orientation and attitude toward AI. Ethics-of-care scholarship emphasises attention to relationships, vulnerability and the consequences of decisions for others (Cohn, 2020; Villegas-Galaviz & Martin, 2024). Whether this proposed association is observed in the present context is an empirical question.

Based on these arguments, the following hypothesis is proposed:

H2: Care value positively affects firms’ attitude toward AI adoption in B2B marketing.

2.4.3 Reasoning – Attitude

Behavioural Reasoning Theory emphasizes that attitudes toward a behaviour are formed through context-specific justifications that decision-makers use to support or oppose adoption decisions (Westaby, 2005). Unlike intention-based acceptance models, BRT explicitly differentiates between facilitating reasons (“reasons for”) and inhibiting reasons (“reasons against”), each of which independently shapes evaluative judgments (Claudy & Peterson, 2014; Sahu et al., 2020).

The B2B AI literature discusses applications in market knowledge and the sales process that motivate potential benefits such as improved customer understanding and decision support (Paschen et al., 2019; Paschen et al., 2020). In this study, perceived improvements in customer relationships, efficiency, accuracy, decision quality, risk, cost and sales are represented as dimensions of reasons for adoption.

Perceived gains in decision support and the sales process provide a rationale for including benefit-related reasons in the proposed model (Paschen et al., 2019; Paschen et al., 2020).

These anticipated benefits motivate a positive relationship between reasons for adoption and attitude; their relevance in this sample is examined empirically.

Conversely, the literature also highlights several inhibiting considerations that weaken favourable attitudes toward AI adoption. These include lack of motivation, particularly when decision-makers perceive existing processes as adequate and do not recognize the added value of AI. Resistance to AI adoption—stemming from organizational inertia, fear of change, or concerns about job displacement—has also been shown to negatively influence attitudes toward intelligent systems (Paschen et al., 2020; Raisch & Krakowski, 2021).

Moreover, the complexity of actual B2B business processes poses a significant barrier, as AI systems may struggle to fully capture the relational, contextual, and dynamic nature of real-world B2B decision-making. Finally, information security and privacy concerns remain a critical reason against adoption, particularly in B2B environments where sensitive customer and transactional data are involved (Dwivedi et al., 2021).

Consistent with BRT, these facilitating and inhibiting reasons are expected to exert differential effects on firms’ attitudes toward AI adoption in B2B marketing.

Accordingly, the following hypotheses are proposed:

H3a: Firms’ reasons for AI adoption—such as customer relationship improvement, efficiency improvement, accuracy improvement, better decision making, risk reduction, cost reduction, and sales increase—positively affect their attitude toward AI adoption in B2B marketing.

H3b: Firms’ reasons against AI adoption—such as lack of motivation, resistance to AI adoption, complexity of actual business processes, and information security concerns—negatively affect their attitude toward AI adoption in B2B marketing.

2.4.4. Attitude and Adoption Intention

Extant research in consumer behaviour and information systems consistently identifies attitude as a key determinant of behavioural intention and actual behaviour (Bian & Forsythe, 2012; Hartmann & Apaolaza-Ibáñez, 2012). In line with this perspective, BRT is conceptually aligned with foundational behavioural frameworks such as the Theory of Reasoned Action (TRA), which posits that attitudes toward a behaviour exert a direct influence on behavioural intentions.

Within the BRT framework, attitude represents an overall evaluative judgment that reflects how favourably or unfavourably a decision-maker assesses a specific behaviour after considering underlying justifications and concerns (Westaby, 2005). Attitudes are formed through a combination of cognitive evaluations and reasoning-based assessments and function as a proximal antecedent of behaviour.

In the context of technology adoption, prior empirical evidence suggests that organizations holding favourable attitudes toward a technology are significantly more likely to adopt and implement it (Bagozzi, 1992; Venkatesh et al., 2003). This relationship has been reinforced in recent studies on artificial intelligence adoption, demonstrating that positive evaluations of AI capabilities and performance outcomes increase the likelihood of organizational adoption (Grewal et al., 2021).

In B2B contexts, where AI adoption decisions involve strategic risk, relational consequences, and substantial resource commitment, attitudes toward AI become particularly influential in shaping adoption behaviour (Paschen et al., 2019; Grewal et al., 2021).

Accordingly, the following hypothesis is proposed:

H4: A positive attitude toward AI adoption positively affects firms’ adoption of AI in B2B marketing.

2.4.5 Reasoning – AI Adoption

Beyond shaping attitudes, Behavioural Reasoning Theory posits that reasons may directly influence behavioural outcomes, independent of attitudinal evaluations (Westaby, 2005). This direct effect is particularly relevant in complex organizational decisions, where managers rely on salient justifications or concerns to simplify decision-making under uncertainty (Kahneman & Tversky, 1982; Claudy et al., 2015).

In B2B contexts, where AI adoption decisions are embedded in strategic, relational, and operational considerations, reasoning-based justifications may directly shape adoption outcomes. Strong facilitating reasons—such as efficiency gains, improved decision accuracy, cost savings, and enhanced sales performance—may prompt firms to adopt AI even when attitudes are ambivalent. Conversely, strong inhibiting reasons—such as security concerns, organizational resistance, or misalignment with complex business realities—may prevent adoption despite generally favourable attitudes toward AI technologies (Sahu et al., 2020; Gesk & Leyer, 2022).

Accordingly, the following hypotheses are proposed:

H5a: Firms’ reasons for AI adoption positively affect their intention to adopt AI in B2B marketing.

H5b: Firms’ reasons against AI adoption negatively affect their intention to adopt AI in B2B marketing.

2.4.6. Moderating Role of Brand Reputation

Although Behavioural Reasoning Theory provides a robust framework for explaining adoption decisions, moderating mechanisms remain relatively underexplored in BRT-based research (Sahu et al., 2020). Brand reputation constitutes a critical organizational asset that influences firms’ perceptions of risk, legitimacy, and strategic caution (Fombrun & Shanley, 1990; Dowling, 2016; Ryan & Casidy, 2018).

Brand reputation refers to stakeholders’ overall evaluation of a firm’s credibility, trustworthiness, quality, and competence (Fombrun & Shanley, 1990). In B2B markets, where long-term relationships, mutual trust, and perceptions of organizational capabilities are paramount, brand reputation plays a significant role in how firms evaluate emerging technologies such as AI (Helm, 2007). Firms with stronger brand reputations possess greater confidence in their organizational capacity to manage change, implement technology successfully, and control potential adverse outcomes.

The study proposes that reputation conditions the relationships between adoption-related reasoning and intention. Reputation is treated as a proposed boundary condition rather than an established protective mechanism. Evidence from an organic-food application of Behavioural Reasoning Theory provides a conceptual comparison (Ryan & Casidy, 2018), but its transfer to AI in B2B markets requires empirical testing.

A positive moderation hypothesis is proposed for the relationship between reasons for adoption and intention. The proposed mechanism is that favourable reputation strengthens the relevance of perceived benefits. This explanation motivates H6b; it does not establish that the interaction exists before the results are examined.

The study also proposes that reputation moderates the attitude–intention relationship. This is an extension tested in the present context, rather than a direct prediction demonstrated by general acceptance frameworks. Accordingly, the following moderation hypotheses are specified.

Accordingly, the following hypotheses are proposed:

H6a: Brand reputation moderates the negative relationship between reasons against AI adoption and AI adoption intention, such that the negative effect is weaker for firms with higher brand reputation.

H6b: Brand reputation moderates the positive relationship between reasons for AI adoption and AI adoption intention, such that the positive effect is stronger for firms with higher brand reputation.

H7: Brand reputation moderates the relationship between attitude toward AI adoption and AI adoption intention, such that the positive effect is stronger for firms with higher brand reputation.

3. Research Methodology

This study employs a quantitative research approach to investigate the adoption of artificial intelligence (AI) in B2B marketing within the Iranian steel industry. The research focuses on examining the effects of care value, reasoning (reasons for and reasons against), and attitude on AI adoption, as well as the moderating role of brand reputation. Given the objective of testing theory-driven hypotheses derived from Behavioural Reasoning Theory (BRT), a survey-based methodology is deemed appropriate.

3.1. Questionnaire Development

Data were collected through a structured questionnaire developed based on Behavioural Reasoning Theory (Westaby, 2005) and prior empirical research on AI adoption and B2B marketing. Questionnaire items identified in the literature were adapted and modified according to the hypotheses of the present study in order to ensure contextual relevance. The questionnaire comprised two sections. The first section gathered respondents’ demographic information, including gender, age, job position, professional experience, and educational background. The second section measured the core constructs of the study, namely care value, reasons for AI adoption, reasons against AI adoption, attitude toward AI adoption, brand reputation, and adoption intention. Measurement items were adapted from established and validated scales in the extant literature and then contextualized to reflect AI adoption decisions in B2B marketing within the steel industry. All constructs were measured using a five-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”).

Adoption intention denotes the respondent’s stated willingness to use AI in B2B marketing, while attitude denotes the respondent’s evaluation of that use. These concepts are consistent with behavioural intention frameworks (Ajzen, 1991; Venkatesh et al., 2003). Care value is interpreted through the ethics-of-care literature (Cohn, 2020; Gilligan, 1982; Tronto, 1993; Villegas-Galaviz & Martin, 2024). Reasons for adoption cover perceived benefits in customer relationships, efficiency, accuracy, decision quality, risk, cost and sales; B2B AI literature provides context for these benefits (Paschen et al., 2019; Paschen et al., 2020).

Conversely, reasons against capture context-specific barriers that may inhibit AI adoption, including lack of motivation, resistance to adoption, difficulty in selecting appropriate AI models, process complexity, fear of job displacement, resistance to learning, and concerns related to privacy, information security, and legal compliance. These items were derived from established literature on organizational and technological barriers to AI implementation (Dwivedi et al., 2021; Huang & Rust, 2018; Paschen et al., 2020; Raisch & Krakowski, 2021). Brand reputation was defined as stakeholders’ perceptions of a firm’s credibility, reliability, and ability to deliver value in relation to AI technologies in B2B markets, with items adapted from Ryan and Casidy (2018). To ensure content validity and clarity, the questionnaire was reviewed by several experts familiar with AI applications and B2B practices, and minor revisions were made to improve item wording and contextual relevance prior to final data collection.

3.2. Data Collection

The primary instrument employed in this study was a researcher-developed questionnaire comprising 54 items, each measured on a five-point Likert scale. Prior to full-scale deployment, a pilot study was carried out with a group of 30 professionals drawn from the fields of marketing and management. The feedback obtained during this preliminary phase was used to refine item wording, eliminate ambiguities, and confirm that the instrument adequately captured the constructs under investigation.

The questionnaire was reported to contain 54 items, and the final sample comprised 275 respondents recruited through snowball sampling. Recruitment targeted professionals in the Iranian steel industry. The original sample-size rationale used an observations-per-item rule of thumb rather than Cochran’s formula; this rule alone does not establish adequate statistical power for the proposed structural paths or interaction effects. The nonprobability recruitment strategy also limits population-level generalization.

Data were gathered through multiple digital channels, including Porsline—an online survey platform—as well as LinkedIn and WhatsApp, enabling broad outreach across geographically dispersed respondents. Of the questionnaires distributed, 275 were deemed complete and suitable for analysis. Informed consent was obtained electronically from all participants prior to completion of the questionnaire. Participants were informed about the purpose of the study, the voluntary nature of participation, and the confidentiality and anonymity of their responses.

To ensure the scientific rigor of the measurement instrument, both validity and reliability were systematically evaluated. Content validity was established through expert review, while face validity was confirmed by assessing the clarity and relevance of items from the respondents’ perspective. Construct validity was examined using SmartPLS-3, through which confirmatory factor analysis was conducted to verify the structural integrity of the measurement model. Reliability was assessed using Cronbach’s alpha and composite reliability (CR), both of which yielded values above the accepted thresholds, confirming the internal consistency of the scales.

4.1. Respondents’ Demographic Information

As depicted in Table 1, the sample was predominantly male, with 223 respondents (81.1%) identifying as male and 52 (18.9%) as female; no participants identified as other genders. With respect to age distribution, the majority of respondents (54.6%) were between 30 and 49 years old, while 25.5% fell within the 18–29 age group. The remaining 16.8% were aged 50 or above, with 4.4% being older than 60.

Regarding educational attainment, over 80% of the respondents held a bachelor’s degree or higher, with 35.3% holding a bachelor’s degree, 33.8% a master’s degree, and 11.3% a doctoral degree. In terms of professional position, the largest group consisted of entry-level employees (42.9%), followed by mid-level (34.9%) and senior-level professionals (22.2%). As for work experience, 29.8% of participants had between 6 and 10 years of experience, making it the most represented group, while 26.2% had 1 to 5 years, 21.8% had 11 to 15 years, 12% had more than 20 years, and 10.2% had between 16 and 20 years of experience.

4.2. Data Analysis

The partial least squares approach (PLS-SEM) has been used as one of the suitable methods for solving structural equations. PLS-SEM is a variance-based approach to structural equation modelling; sample-size adequacy and statistical uncertainty must nevertheless be assessed for the model under study (Hair et al., 2019; Sarstedt et al., 2022). Furthermore, PLS-SEM is proper for testing higher-order models (Sarstedt et al., 2022). For this purpose, SmartPLS software version 3.2.8 was used for evaluating the structural model. SPSS 27 was also used to analyze demographic data. In this study, in the first stage, validity and reliability indices were used to assess measurement items, and in the next stage, all hypotheses were tested using structural equation modeling (SEM). The effect of the moderator variable was analyzed using the PROCESS Macro v4.2 (Hayes et al., 2017) in SPSS.

4.3. Data Screening and Common Method Bias (CMB)

In order to ensure a suitable model fit and the absence of large deviations from normality, the data were screened for outliers prior to evaluating the measurement model (Hair et al., 2019). Additionally, the Variance Inflation Factor (VIF) test was implemented to check for possible multicollinearity issues among the constructs. VIF values ranged from 2.088 to 2.938, indicating no multicollinearity between the constructs (Hair et al., 2019).

In addition, the purpose of the research was explained in the questionnaire to minimize Common Method Bias (CMB), and respondents were assured of the confidentiality of their information. To screen for a dominant common factor in the collected data, Harman’s Single-Factor Test was conducted. The results of the exploratory factor analysis revealed that the first factor accounts for only 32.096% of the total variance of the variables, which is below the 50% threshold proposed by Podsakoff et al. (2003). This result does not rule out common method bias. Because the constructs were measured using the same self-report questionnaire at one time point, shared-method variance remains a potential limitation.

4.4. Test of the Measurement Model

To evaluate the conceptual model of the present study, a set of reliability and validity indicators were examined. In the first stage, the internal reliability of the constructs was assessed using Cronbach’s alpha (CA) and composite reliability (CR). According to the criteria provided by Hair et al. (2019), the value of these indicators should be at least 0.70, and very high values (above 0.95) may indicate excessive overlap among items.

The analysis results showed that all research constructs had values higher than the threshold of 0.70 and lower than the warning level of 0.95. Therefore, it can be concluded that the measurement model possesses adequate reliability and satisfactory internal consistency. The results of this section are presented in Table 2.

In the next stage, convergent validity was examined using the Average Variance Extracted (AVE) indicator. According to the criteria of Hair et al. (2019), the AVE value should be at least 0.50 so that the construct can explain more than half of the variance of its items. As shown in Table 2, all constructs have AVE values higher than 0.50. Therefore, the model is in a favorable condition in terms of convergent validity.

Item loadings ranged from 0.792 to 0.890 (Table 2), exceeding the commonly used criterion of approximately 0.70 (Hair et al., 2019). These loadings support indicator reliability but do not replace evaluation of construct validity or documentation of item provenance.

To assess discriminant validity in the measurement model, the Fornell and Larcker (1981) criterion was employed. According to this criterion, the square root of the average variance extracted for each construct should be greater than its correlation with other constructs in the model. This indicates that each construct explains more variance from its own items compared to items of other constructs, and is therefore conceptually distinct. As shown in Table 3, the square root values of AVE, displayed along the diagonal of the table, are greater than their correlations with other constructs across all constructs. Therefore, it can be concluded that discriminant validity is confirmed in the research model, and the constructs have appropriate conceptual distinction from one another.

Table 3. Fornell–Larcker matrix in two panels

Panel A. Columns CV–DMOD

Panel B. Columns BRAND–COST

4.5. Structural Model: Hypotheses Testing

The structural model was evaluated using 500 bootstrap resamples (Efron, 1992). The reported path estimates were positive for attitude to adoption intention (β = 0.323), reasons for adoption to attitude (β = 0.436) and intention (β = 0.307), and care value to reasons for adoption (β = 0.524) and attitude (β = 0.153). Negative coefficients were reported for reasons against adoption to attitude (β = −0.234) and intention (β = −0.238), and for care value to reasons against adoption (β = −0.530). Although the latter path was reported as statistically significant, its direction contradicts the positive association specified in H1b; H1b is therefore not supported. Table 4 reports the structural paths, and Figure 2 presents the model diagram.

The reported coefficients of determination were R² = 0.571 for adoption intention and R² = 0.477 for attitude. These values describe explained variance in the study sample; they do not establish causal effects, out-of-sample predictive performance, or actual adoption behaviour. The additional R² values reported for the dimensions of reasons for and against adoption must be interpreted in relation to the specified higher-order measurement model, rather than treated as independent evidence of predictive validity. The structural coefficients are summarised in Figure 2.

4.6. Moderation Analysis

To examine the moderating role of brand reputation, Hayes’ PROCESS macro-Model 8 was employed with 5000 bootstrap samples and 95% confidence intervals. Variables entering the interaction terms were mean-centred before the product terms were formed. Mean-centring does not by itself establish the absence of multicollinearity.

For the reasons “for” model, the interaction effect between reasons “for” and brand reputation on attitude was significant (β=0.2145, p=0.0003), while the interaction effect on intention was not significant (β=0.0433, p=0.4134). The index of moderated mediation was significant (Index = 0.0926, BootLLCI = 0.0432, BootULCI = 0.1439), confirming moderated mediation through attitude. The nonsignificant interaction predicting intention does not support H6b as formulated. The interaction predicting attitude and the conditional indirect association are additional findings concerning different paths.

In the reasons “against” model, the interaction effect between reasons “against” and brand reputation on attitude was significant and negative (β=−0.1994, p=0.0040), whereas the interaction effect on intention was not significant (β=−0.0467, p=0.4104). The index of moderated mediation was also significant (Index = -0.1025, BootLLCI = -0.1747, BootULCI = -0.0310), confirming moderated mediation. The nonsignificant interaction predicting intention does not support H6a as formulated. Assuming higher scores indicate stronger barriers and higher reputation, the negative interaction predicting attitude indicates that the negative slope becomes more pronounced as reputation increases, rather than weaker.

To test the moderating role of brand reputation on the attitude-intention relationship, PROCESS Model 1 was used. The interaction term was significant (β=0.1707, p=0.0022), explaining an additional 1.94% of variance in adoption intention. Conditional effects analysis revealed that the effect of attitude on intention strengthened progressively across low (0.4710), moderate (0.6075), and high (0.7440) levels of brand reputation (all p<.0001).

4.7. Model Fit Assessment

The overall fit of the final research model was evaluated using two key indices: Standardized Root Mean Square Residual (SRMR) and Goodness of Fit (GOF).

The reported SRMR was 0.058. This descriptive value should be interpreted alongside the measurement and structural assessments. The previously reported GoF calculation cannot be retained as evidence of strong fit because its stated mean AVE and mean R² inputs do not reconcile with the values reported elsewhere in the manuscript. Accordingly, no conclusion of strong overall model fit is drawn from that calculation.

5. Discussion

The findings indicate associations consistent with several predictions of Behavioural Reasoning Theory in this sample of Iranian steel-industry respondents. Attitude was positively associated with adoption intention, and reasons for and against adoption were associated with attitude and intention. These findings concern individual survey responses and do not establish firm-level adoption or causal mechanisms. H1b was contradicted by the negative coefficient, and the direct moderation hypotheses H6a and H6b were not supported.

Reasons for adoption were positively associated with attitude, whereas reasons against adoption were negatively associated with attitude. These directions are consistent with the distinction between supportive and inhibiting reasoning (Westaby, 2005). The structural paths concern the aggregate constructs; they do not establish separate significant effects for each benefit or barrier dimension.

Reasons for adoption were positively associated with intention, whereas reasons against adoption were negatively associated with intention. These estimates are consistent with the proposed distinction between supportive and inhibiting reasoning (Westaby, 2005). The cross-sectional design does not establish that changing either type of reason would change subsequent adoption behaviour.

Care value was positively associated with reasons for adoption and attitude and negatively associated with reasons against adoption. The negative association with reasons against adoption contradicted H1b, leaving this hypothesis unsupported. These estimates concern respondents’ reported ethical orientation and evaluations; they do not demonstrate that an organizational intervention would improve AI adoption.

The moderation findings were specific to the paths tested. Reputation did not significantly moderate the direct relationships between reasons and intention, so H6a and H6b were not supported. Additional analyses indicated a positive interaction between reasons for adoption and reputation in predicting attitude, and a negative interaction between reasons against adoption and reputation in predicting attitude. With higher scores representing stronger barriers and reputation, the latter corresponds to a stronger negative association at higher reputation. The positive attitude-by-reputation interaction predicting intention was consistent with H7. These findings do not support a general claim that reputation buffers adoption barriers.

Overall, the study identifies associations among reasoning, care value, reputation, attitude and AI adoption intention in the surveyed sample. The findings provide a basis for further investigation of these relationships in B2B settings, while the unsupported hypotheses, measurement uncertainties and sampling limitations restrict broader conclusions.

5.1. Theoretical Implications

This study makes four primary contributions to the technology adoption literature.

First, the study applies Behavioural Reasoning Theory to both supportive and inhibiting reasons for AI adoption intention (Westaby, 2005). The observed associations were positive for supportive reasons and negative for inhibiting reasons. No formal comparison of coefficient magnitudes was reported, so the findings do not establish equal effects or the superiority of either type of reason. Care value was associated with both types of reasoning, although its negative association with reasons against adoption contradicted H1b. The aggregate construct findings do not establish separate structural effects for every barrier dimension.

Second, the study examines adoption intention in the Iranian steel industry. The findings describe this particular sample and should not be generalized to all firms, industrial sectors or developing economies. They also do not establish that negative considerations are as influential as, or more influential than, positive considerations without a formal statistical comparison.

Third, the study suggests that reputation may condition particular associations involving attitude and intention. The additional interaction estimates concern the reasons-to-attitude paths, whereas the hypothesised moderation of the direct reasons-to-intention paths was not supported. The negative reasons-against-by-reputation interaction does not establish that reputation mitigates barriers. The positive attitude-by-reputation interaction is consistent with a stronger attitude–intention association at higher reputation.

Finally, the study considers care value and reputation alongside adoption-related reasoning. These variables were measured through individual responses, and the results should therefore be interpreted at that level. Organizational readiness and environmental conditions were not included in the tested model, so the present data do not establish their relative influence or validate the broader TOE framework.

5.2. Practical Implications

The findings suggest that practitioners should attend to both perceived benefits and concerns when discussing AI in B2B marketing. Decisions about skills, task allocation and human involvement are also relevant to the broader implementation context (Raisch & Krakowski, 2021). However, this survey did not test training, transparency policies, pilot programmes or other interventions, so their effectiveness cannot be inferred from these results.

At the institutional level, the results highlight the importance of a supportive regulatory and policy environment. Clear legal frameworks for data protection and responsible AI use can reduce uncertainty and foster trust among adopting firms. Public interventions in the form of financial incentives, shared digital platforms, and broader investment in industrial digital infrastructure may further lower structural barriers to adoption. The findings also point to the need for workforce development initiatives, including reskilling programs, digital capability enhancement, and stronger collaboration between industry and educational institutions, in order to address skill shortages and mitigate concerns related to technological displacement. In addition, innovation ecosystems that connect firms, universities, and technology ventures can accelerate knowledge transfer and promote the diffusion of effective AI practices.

For technology developers and AI solution providers, the findings suggest that adoption depends not only on technical sophistication but also on usability, transparency, and implementation support. Solutions designed with user-friendly interfaces, explainability features, and privacy-by-design principles are more likely to reduce perceived complexity and increase trust in industrial settings. Likewise, training, clear documentation, continuous technical support, and low-risk trial opportunities can facilitate organizational learning and contribute to smoother implementation processes.

5.3. Limitations and Future Research Directions

Despite its contributions, this study has several limitations that provide directions for future research. First, the cross-sectional design limits the ability to capture temporal dynamics and causal relationships in AI adoption. Future studies could employ longitudinal designs to examine how attitudes and adoption behaviours evolve over time. Second, the empirical setting is confined to the Iranian steel industry within a single national context, which may restrict the generalizability of the findings. Comparative studies across industries and countries would help assess the robustness of the proposed relationships in different institutional and industrial environments. Third, the study examines AI adoption at a general level and focuses primarily on adoption intention rather than subsequent stages of use. Future research may therefore investigate specific AI applications and extend the analysis to post-adoption processes such as continued use and organizational integration. Finally, although the model incorporates key constructs from BRT, other organizational and environmental factors were not included. Further research could integrate additional determinants—such as organizational readiness, leadership support, and technological infrastructure—to provide a more comprehensive understanding of AI adoption in B2B contexts. In addition, snowball sampling limits representativeness. The inclusion of junior employees and the absence of documented firm-level aggregation restrict organizational inferences. Self-reported intention is not evidence of actual adoption. Shared-method variance cannot be excluded. Incomplete item-level documentation and inconsistencies in the reported descriptive and model statistics limit the assessment of measurement validity and the precision of the findings.

Ethical Considerations

This study adhered to established ethical standards for research involving human participants. The research involved a voluntary and anonymous online questionnaire and did not include clinical procedures, medical interventions, or the collection of sensitive personal data. Therefore, formal ethical approval from an Institutional Review Board (IRB) or ethics committee was not required under the institutional guidelines applicable to the authors at the time of the study. Participation was entirely voluntary and conducted for academic research purposes

Informed Consent

Electronic informed consent was obtained from all participants before questionnaire completion. Participants were informed of the study purpose, the voluntary nature of participation, and the confidentiality and anonymity of their responses.

Author Contributions

Sogand Akbari Tari: Conceptualization, methodology, investigation, data curation, formal analysis, and writing—original draft. Vahid Makizadeh: Conceptualization, methodology, supervision, validation, and writing—review and editing. Both authors reviewed and approved the final manuscript and agree to be accountable for the integrity and accuracy of the work.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of Interest

The authors declare no financial or non-financial competing interests related to this research.

Acknowledgments

The authors thank the steel-industry professionals who participated in the survey and the experts who provided feedback on the questionnaire during its development and pilot testing.

Data Availability

The anonymized data supporting the findings of this study are available from the corresponding author upon reasonable request, subject to participant confidentiality and applicable institutional data-sharing requirements.