A conceptual study published in Industrial Artificial Intelligence by researcher Kwabena Abrokwah-Larbi establishes a theoretical framework grounded in dynamic capabilities theory to examine how generative artificial intelligence contributes to customer personalisation development within small and medium-sized enterprises. Drawing on an extensive literature synthesis across artificial intelligence, machine learning, data architecture, and marketing systems, the author models customer personalisation as the primary dependent variable influenced directly and indirectly by generative artificial intelligence as the independent variable. The investigation formulates eight core research propositions that delineate the structural mechanics through which generative models interact with operational digital resources. Within this theoretical architecture, deep learning algorithms and smart data act as vital mediating mechanisms that translate generative computation into tailored customer interactions, while the Internet of Things serves as an essential moderating force that governs the relationship between generative intelligence and smart data acquisition. The conceptual model formalises how resource-constrained enterprises can systematically configure digital capabilities to achieve context-oriented customer personalisation.
Dynamic capabilities theory provides the foundational conceptual lens for evaluating the transformative potential of generative systems within smaller business structures. As conceptualised in management scholarship by Warner and Wäger, dynamic capabilities represent innovation-focused organisational abilities that empower enterprises to systematically generate, extend, and adapt their existing resource bases in volatile commercial environments. The paper posits that the disruptive characteristics of generative artificial intelligence necessitate comprehensive digital capability reconfiguration so that small and medium-sized enterprises can effectively exploit evolving marketing possibilities. In modern consumer markets, customer demand has increasingly shifted toward rapid, context-specific interactions, compelling smaller businesses to restructure their internal digital workflows. By implementing generative artificial intelligence, enterprises can dynamically modify their communicative capabilities to create diverse formats of original marketing content, encompassing automated text composition, synthetic voice responses, tailored photographic elements, and customized video assets. This capability transformation enables firms to realign their market scanning competencies and establish resilient digital infrastructures capable of continuously supporting customer personalisation.
The technical differentiation of generative artificial intelligence forms the basis of the framework’s early propositions regarding algorithmic integration. In sharp contrast to conventional artificial intelligence architectures that depend primarily on controlled training paradigms, generative models employ uncontrolled learning processes where algorithms identify latent patterns and high-level representations directly from unstructured datasets without explicit supervision. Deep learning functions as a multilayer neural network architecture and a specialized branch of representation learning that enables generative models to parse, interpret, and generate natural language that mirrors human conversation. The study proposes that deep learning directly influences generative artificial intelligence toward customer personalisation, designated as Proposition 1, because foundational techniques like natural language processing provide the underlying computational infrastructure for real-time customer query handling and dialogue automation. Concurrently, Proposition 2 establishes that deep learning represents an influential element in customer personalisation, as advanced neural networks evaluate complex behavioural histories to determine customer preferences with heightened predictive accuracy across commercial channels.
The operational reality of deep learning applications in commercial personalisation is prominently manifested through recommendation engines, which have shifted from discretionary enhancements into standard operational expectations across modern digital services. The author highlights empirical precedents from major technology platforms, including Google’s deployment of wide and deep learning algorithms to predict advertising click-through counts and Twitter’s utilization of deep neural architectures to forecast customer purchasing decisions, alongside algorithmic recommendation frameworks implemented by Amazon, YouTube, and Netflix. Despite these advancements, the framework acknowledges significant operational hurdles identified in prior literature by Schmitt, which restrict deep learning adoption within smaller business environments. These barriers include acute computational complexity, inadequate data processing infrastructure, shortages of specialised technical personnel, and the persistent opacity surrounding the causal reasoning of deep learning outputs. Furthermore, because deep learning algorithms require expansive datasets to optimize performance, smaller enterprises must deliberately address data volume constraints to fully leverage neural networks for customer personalisation.
Smart data functions as an indispensable mediating asset that bridges technical computation and practical marketing execution within the proposed model. Citing foundational definitions by Triguero and colleagues, the study defines smart data as specialized technologies capable of processing immense volumes of unstructured information to extract verifiable economic and operational value. Proposition 3 states that deep learning serves as an influential element in smart data analysis, reflecting the capacity of recurrent neural networks and long short-term memory architectures to extract sophisticated consumer concepts from raw information streams. Proposition 4 asserts that generative artificial intelligence is an influential element in the generation of customer-focused smart data through systematic knowledge discovery and smart data mining. To transform raw operational records into actionable smart data, generative systems process information across the 5Vs characteristics identified by Yau: accelerating data velocity, accommodating large volumes, extracting verified value, harmonizing diverse data varieties, and rigorously maintaining data veracity.
The integration of smart data directly shapes downstream personalisation initiatives, establishing the empirical necessity for Proposition 5, which posits that smart data is an influential element in customer personalisation. High-quality smart data structures consolidate granular consumer attributes, including historical purchasing trajectories, demographic profiles, online browsing behaviors, real-time social platform interactions, and affective sentiment assessments. By extracting actionable intelligence from these structured repositories, generative artificial intelligence enables small and medium-sized enterprises to move beyond reactive promotional adjustments toward proactive customer engagement. Rather than merely responding to completed consumer transactions, predictive algorithms analyze synthesized smart data patterns to anticipate future customer desires, automate individualized shopping itineraries, and curate tailored value propositions. As supported by marketing scholarship from Afshar and Aheleroff, the systematic application of smart data elevates commercial communication above standardized advertising barrages, reducing information saturation for consumers while simultaneously fostering elevated brand trust, customer engagement, and long-term brand loyalty.
The Internet of Things operates as a pivotal moderating construct within the theoretical model, specifically governing the foundational link between generative artificial intelligence and smart data acquisition. Defined as an interconnected network of physical devices capable of gathering, exchanging, and communicating real-time information, Internet of Things ecosystems supply continuous sensory data regarding physical customer actions and environmental contexts. Proposition 6 formally specifies that the Internet of Things moderates the relationship between generative artificial intelligence and smart data. The framework outlines two critical operational procedures that generative models execute upon incoming connected device feeds: automated data description, where artificial intelligence identifies and contextualizes inbound sensory streams for deep learning analysis, and compression and synthesis, wherein critical customer insights are distilled from immense unstructured datasets into compact, archivable formats. Supported by technologies such as radio frequency identification, wireless sensor networks, Bluetooth, cellular connectivity, and cloud interfaces, connected networks provide the raw empirical velocity required for sustainable smart data generation.
The structural mediation mechanisms of the conceptual framework are formally expressed in Proposition 7 and Proposition 8, which link generative inputs to commercial personalisation outcomes. Proposition 7 affirms that deep learning mediates the relationship between generative artificial intelligence and customer personalisation, demonstrating that generative platforms depend entirely on trained neural network layers to evaluate continuous consumer datasets and output high-fidelity marketing artifacts. The continuous training of deep learning models on specialized enterprise datasets allows generative tools to generate individualized product descriptions, dynamic email marketing campaigns, and interactive conversational engagements that directly reflect customer context. Proposition 8 establishes that smart data mediates the relationship between generative artificial intelligence and customer personalisation. Through an ongoing iterative cycle, generative artificial intelligence aggregates fragmented touchpoints across e-commerce storefronts and social platforms, evaluates fine-grained customer records at a granular level, and establishes unified customer profiles that guide real-time marketing customization.
The downstream consequences of successful generative personalization yield profound strategic and managerial implications for small and medium-sized enterprises competing against larger market incumbents. The conceptual framework details three paramount marketing outcomes stemming from context-oriented personalisation: interactive marketing capabilities, customer value co-creation, and enduring consumer loyalty. As enterprises transition from passive promotional dissemination to active customer dialogues, personalized communications encourage consumers to participate directly in product refinement and bespoke service design. This collaborative co-creation process allows products and services to evolve from purely functional utilities into offerings imbued with distinct aesthetic and emotional value tailored to specific customer preferences. To realize these operational benefits, however, the author emphasizes that small business leadership must commit targeted financial capital toward modernizing internal Internet of Things infrastructure, adopting scalable cloud analytics, and upskilling staff to manage generative workflows alongside established marketing competencies.
Notwithstanding its extensive conceptual contributions, the investigation explicitly notes several theoretical boundaries and methodological limitations that define future research trajectories. Because the published paper is strictly theoretical and proposition-based, it contains no empirical datasets, quantitative statistical testing, or qualitative field validations measuring actual enterprise performance metrics. The author explicitly underscores that future scholarly inquiries must implement rigorous empirical methodologies, utilizing survey instruments, multi-case qualitative studies, or longitudinal econometric analyses across small business sectors to validate the proposed mediating and moderating relationships. Additionally, the paper encourages scholars to investigate how generative customer personalisation architectures function within emergent immersive computing environments, most notably metaverse platforms where customer data streams and real-time interaction paradigms diverge significantly from conventional digital storefronts. The resulting theoretical architecture establishes a comprehensive, structured foundation for forthcoming empirical investigations into artificial intelligence adoption among small and medium enterprises.
Subject of Research: Theoretical framework of generative artificial intelligence and customer personalisation in small and medium-sized enterprises
Article Title: The role of generative artificial intelligence (GAI) in customer personalisation (CP) development in SMEs: a theoretical framework and research propositions
Article References: Abrokwah-Larbi, K. (2023). The role of generative artificial intelligence (GAI) in customer personalisation (CP) development in SMEs: a theoretical framework and research propositions. Industrial Artificial Intelligence, 1(1), Article 11. https://doi.org/10.1007/s44244-023-00012-4
Image Credits: AI Generated
DOI: 10.1007/s44244-023-00012-4
Keywords: generative artificial intelligence, customer personalisation, small and medium-sized enterprises, dynamic capabilities theory, deep learning, smart data, internet of things, interactive marketing, value co-creation, customer loyalty, recommender systems, uncontrolled learning, smart data mining, knowledge discovery, representation learning, customer profiling, 5Vs characteristics, digital transformation
Cite Scienmag News
Blake Davidson. (November 19, 2026). Theoretical Framework Models Generative Artificial Intelligence for Small Enterprise Customer Personalisation. Scienmag. https://scienmag.com/theoretical-framework-models-generative-artificial-intelligence-for-small-enterprise-customer-personalisation/
Blake Davidson. "Theoretical Framework Models Generative Artificial Intelligence for Small Enterprise Customer Personalisation." Scienmag, 19 November 2026, https://scienmag.com/theoretical-framework-models-generative-artificial-intelligence-for-small-enterprise-customer-personalisation/. Accessed 19 November 2026.
Blake Davidson. "Theoretical Framework Models Generative Artificial Intelligence for Small Enterprise Customer Personalisation." Scienmag. November 19, 2026. https://scienmag.com/theoretical-framework-models-generative-artificial-intelligence-for-small-enterprise-customer-personalisation/








