AI-Enabled Super Apps as Complex Socio-Technical Ecosystems: A Systemic View of User Continuance

Abstract

Super apps have emerged as complex digital service ecosystems that integrate multiple heterogeneous services within a unified platform architecture. As artificial intelligence (AI) capabilities become increasingly embedded into these platforms, understanding how AI-enabled features influence user evaluations has become an important research issue. This study develops a new research model by extending the stimulus–organism–response (SOR) framework to examine the determinants of users’ continuance intention toward super apps. Specifically, performance efficacy, service efficiency, and perceived security are conceptualized as stimulus factors. Satisfaction is modeled as the organism variable; and continuance intention represents the behavioral response. In addition, this study conceptualizes AI system capability as a platform-level capability that enables the integration, adaptation, and personalization of heterogeneous services. It examines both its direct effect on user satisfaction and its moderating role in the relationships between functional affordances and satisfaction. Based on survey data collected from 614 super-app users in South Korea, the research model was analyzed using partial least squares structural equation modeling (PLS-SEM). The results reveal that performance efficacy and perceived security significantly influence user satisfaction, whereas service efficiency does not have a significant effect. Furthermore, AI system capability not only directly enhances user satisfaction but also strengthens the relationships between functional affordances and satisfaction. A multi-group analysis comparing financial and non-financial super apps shows that these effects vary depending on the service context. These findings contribute to the literature by conceptualizing AI as a system-level capability that both enables and enhances the realization of functional affordances in complex digital ecosystems.

1. Introduction

In today’s digital marketplace, super apps have emerged as an important innovation in the platform economy. Super apps integrate a wide range of services within a single application of a single brand, enabling users to perform various daily activities without switching between multiple apps [1,2]. The early forms of super apps appeared mainly in East Asia through messaging platforms such as WeChat, KakaoTalk, and LINE [3]. Over time, super-app services have expanded progressively across multiple industries. For example, financial platforms such as Paytm and Revolut integrate banking and fintech services within a unified application [4]. Beyond the financial domain, many super apps further incorporate heterogeneous services such as ride-hailing, food delivery, and digital payments within a single platform ecosystem [5]. The market outlook for super apps is highly positive and reflects these advantages. According to Research&Market [6], the global super-app market is projected to reach approximately USD 80–120 billion by 2025 and grow at an annual rate of 15–30% between 2025 and 2030.

A notable trend in super-app ecosystems is the rapid integration of artificial intelligence (AI). As super apps continuously add new services and functions, users may find it increasingly difficult to identify and access the services they need. In this context, AI can play an important role in improving user experience [7]. For example, generative AI-based conversational services can help users discover the desired services more easily. Additionally, AI technologies can analyze user behavior patterns and automatically provide personalized content and recommendations. Major platform providers have already begun to integrate AI capabilities into their super-app services. For instance, Kakao, a leading South Korean mobile platform company, introduced several AI-based functions at its developer conference in September 2025, including AI summarization, ChatGPT integration, and an AI agent function that enables users to access services such as gifting, KakaoMap (a navigation and location-based service), reservations, and Melon (a popular Korean music streaming service) without switching applications or navigating complex menus [8].

In the academic literature, prior research on super apps has primarily focused on identifying the key determinants and drivers of user adoption. In this regard, established technology acceptance frameworks, such as the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT), have been widely employed. For example, Salehi et al. [9] investigated the factors influencing the adoption of super apps by applying the TAM. Hasselwander and Weiss [10] identified five distinct user clusters and emphasized the importance of function-centered integration with UTAUT variables. Zhu et al. [11] analyzed the driving factors of mobile payment adoption within LINE, a messaging-based super app, and found that while trust and habitual use of the core application facilitate mobile payment adoption, the perceived synergy and fit between the platform and its extended services are the most critical drivers of user attitudes. More recently, this line of research has begun to extend to Western contexts. Hasselwander and Weiss [12] found that super-app adoption among German consumers is largely driven by performance expectancy and price value, while perceived risk and gender differences remain significant barriers. Despite these contributions, empirical research on super apps in Western contexts remains relatively limited. While there is a growing body of research examining platform expansion strategies—such as those observed in Uber and X—as pathways toward super-app development, studies focusing on the determinants of user adoption and usage intention remain limited [13]. In contrast, a substantial body of research has emerged in East Asia, where super apps are more mature and deeply embedded in everyday digital ecosystems. As such, insights derived from East Asian contexts may offer valuable reference points for understanding emerging super-app adoption in Western markets; however, such insights should be interpreted with caution given differences in institutional, regulatory, and cultural environments.

Nevertheless, important theoretical gaps remain in the literature. First, although previous studies have examined the factors influencing super-app adoption and continuance using various theoretical frameworks, limited attention has been paid to systematically conceptualizing the role of embedded AI within these platforms. In particular, prior research has often treated AI as a set of independent features or discrete functionalities, operationalizing it through specific attributes such as responsiveness, usefulness, or interaction quality [14]. While informative, this perspective provides a limited understanding of AI in complex platform environments, as it overlooks its role as an integrative and adaptive system-level capability that operates across heterogeneous services and shapes how users realize functional affordances. Accordingly, this study reconceptualizes AI as a system-level capability that operates across integrated services. Second, despite the multi-domain nature of super apps, comparative research across different service contexts (e.g., financial versus non-financial super apps) remains scarce, thereby limiting our understanding of contextual heterogeneity in user behavior.

To address these limitations, this study adopts the stimulus–organism–response (SOR) framework to examine user experiences in super-app environments. The SOR framework is particularly appropriate in this context because super apps integrate multiple heterogeneous services within a single platform, creating complex and multifaceted stimuli that simultaneously influence users’ cognitive and affective evaluations. Unlike traditional models that focus on direct relationships between system attributes and behavioral outcomes, the SOR framework enables a more nuanced understanding of how external platform characteristics are internally processed by users before leading to behavioral responses.

In super-app environments, these stimuli are not limited to isolated system attributes but are shaped by the platform’s underlying ability to integrate and coordinate services. Accordingly, this study conceptualizes performance efficacy, service efficiency, and perceived security as key stimulus factors reflecting distinct functional affordances of super apps, while incorporating AI system capability as a higher-order platform capability that shapes how these stimuli are experienced and translated into user evaluations. In particular, AI system capability enables users to more effectively realize functional affordances through adaptive, personalized, and integrative mechanisms, thereby influencing user satisfaction and subsequent continuance intention.

Based on this framework, the following research questions are proposed:

RQ1. Do performance efficacy, service efficiency, and perceived security influence users’ continuance intention through satisfaction?

RQ2. Does AI system capability influence user satisfaction and continuance intention by shaping how functional affordances are realized in super-app environments?

RQ3. Are there differences in the significant determinants of continuance intention between financial and non-financial super apps?

From a systems perspective, this study contributes to understanding how AI-enabled capabilities influence user responses within complex socio-technical ecosystems. In super apps, multiple heterogeneous services are integrated within a single platform infrastructure, and AI system capability functions as a coordination mechanism that shapes how users experience and evaluate system attributes. By examining these relationships, this study provides insight into how users respond to interconnected digital services within super-app ecosystems.

2. Theoretical Background

2.1. Super Apps and AI

Super apps are integrated digital platforms that consolidate multiple heterogeneous services within a single application, enabling users to access diverse functionalities through a unified interface [5]. From a service provider’s perspective, super apps offer several strategic advantages. By integrating multiple services, platforms can attract a broader user base than traditional single-purpose applications. Furthermore, an integrated ecosystem enables firms to increase user retention and strengthen long-term customer relationships [1]. From a systems perspective, super apps can be understood as complex socio-technical ecosystems in which multiple digital services, user interactions, and intelligent algorithms operate as interconnected subsystems.

Super apps improve user accessibility and convenience by integrating the functions previously provided by multiple standalone applications into a single platform. However, because super apps incorporate a wide range of services within a single environment, users may experience difficulties in locating and accessing the specific functions they need. AI plays an important role in enhancing the usability and convenience of super apps. The AI functions embedded in super apps can be categorized from several perspectives. Studies focusing on generative AI in super apps have proposed different classifications for AI functions. For example, Gupta et al. [15] categorized generative AI in super apps into integrated digital payment mechanisms, platform mediation capabilities, inclusive service provision, personalized application environments, multipurpose platforms integration, and functional optimization. In addition, Martina [16] categorized AI capabilities in super apps into conversational agents such as smart chatbots and digital assistants, personalized recommendation systems, content creation and summarization, voice-to-text services, AI-based search, and conversational interfaces. Similarly, Infosys [7] summarized the AI functions in super apps, including user behavior analysis and hyper-personalization, in-app personalization and widget optimization, intelligent search, conversational AI and virtual assistance, compliance monitoring and fraud detection, language translation and cultural adaptation, dynamic pricing and marketing optimization, and smart notifications and analytics.

While the aforementioned studies provide valuable insights into the diverse functionalities of AI in super apps, they largely focus on categorizing individual features. In complex platform environments, however, these AI functions do not operate in isolation, but collectively contribute to a higher-level system capability. Specifically, AI in super apps can be understood as an integrative and adaptive capability that enables the coordination of heterogeneous services, facilitates personalized and context-aware interactions, and supports users in efficiently navigating and utilizing platform functionalities. From this perspective, AI system capability reflects the platform’s ability to orchestrate multiple services and enhance the realization of functional affordances, rather than merely representing a collection of discrete features.

2.2. AI System Capability from an Affordance Perspective

Affordance theory offers critical insights into how technological systems shape user experiences and behaviors. An affordance refers to the action possibilities that a technological artifact offers to users, emerging from the interaction between system features and users’ goals and capabilities [17]. In digital environments, these affordances are not inherent solely in system features but are realized when users perceive and actualize them in specific contexts. Thus, system attributes influence user outcomes indirectly by enabling or constraining what users can effectively do with the system.

In super-app environments, where multiple heterogeneous services are integrated within a single platform, an affordance perspective provides a particularly appropriate lens for understanding the role of AI. From this perspective, AI operates along two interrelated dimensions. First, AI introduces new system-level functionalities that extend the platform’s capabilities, enabling users to perform tasks more effectively through features such as personalized recommendations, intelligent assistance, and adaptive interactions. Second, AI functions as an integrative and adaptive system-level capability that supports the coordination and enhancement of general system affordances by facilitating seamless interactions across services [18]. Through these dual roles, AI not only introduces new action possibilities but also strengthens users’ ability to realize functional affordances embedded in the platform.

Building on this perspective, this study conceptualizes AI system capability as a platform-level capability that reflects the system’s ability to integrate, adapt, and personalize services across heterogeneous domains. In this study, performance efficacy, service efficiency, and perceived security are conceptualized as key stimulus factors reflecting the functional affordances of super-app platforms, representing users’ ability to achieve goals, optimize processes, and manage risks. However, the extent to which these affordances are effectively realized depends on the AI system capability. By enabling users to more effectively utilize system functionalities and navigate interconnected services, AI system capability enhances the translation of these stimulus factors into user satisfaction and continuance intention. Accordingly, AI system capability is expected to influence how users interpret platform attributes, thereby moderating the relationships between stimulus factors and satisfaction.

2.3. SOR Framework

The SOR framework provides a theoretical foundation for explaining how external stimuli influence individuals’ internal states and subsequently lead to behavioral responses [19]. In this framework, a stimulus refers to environmental factors that trigger an individual’s psychological reactions [20]. These stimuli represent elements of the external environment such as product or service characteristics, interface design, and service atmosphere, which can shape an individual’s cognitive and emotional experiences [21]. The organism component represents the internal state of individuals arising from exposure to external stimuli [22]. It typically includes both cognitive and affective responses that mediate the relationship between stimuli and behavioral outcomes [23]. In service and technology contexts, user perceptions and evaluations of a product or service, such as their satisfaction, are often considered organismic responses [24]. The final element of the framework, response, refers to the behavioral outcomes resulting from the organismic state. These responses appear in the form of observable behaviors such as customer decision-making or behavioral intentions [25].

Although the SOR framework was originally developed in the field of environmental psychology, it has been widely adopted in various research areas, including consumer behavior, service management, and technology adoption studies, to explain customer decision-making processes [26,27,28]. More recently, the SOR paradigm has been applied to emerging digital service contexts. For example, Yang and Lee [24] applied the SOR perspective to examine the factors influencing the adoption of remote services in smart city environments by comparing two smart cities in South Korea. In addition, Khoa and Huynh [29] used the SOR framework to investigate the impact of digital transformation on consumer loyalty to fitness service platforms. Vafaei-Zadeh et al. [30] designed a research model combining the SOR framework and task–technology fit (TTF) theory to verify the determinants that significantly affect the adoption of AI-driven customer service. Based on these prior studies, the SOR framework is considered appropriate for examining user experiences in super-app ecosystems because users interact with multiple integrated services within a unified digital platform environment, and their perceptions of platform attributes may influence their internal evaluations and subsequent behavioral intentions.

3. Research Model and Hypotheses

This study develops a comprehensive research model based on the SOR framework to explain user responses within AI-enabled super-app ecosystems (see Figure 1). In this context, performance efficacy, service efficiency, and perceived security are conceptualized as key stimulus factors representing the functional affordances of super-app platforms. Drawing on the information systems literature, performance efficacy and service efficiency are closely related to system quality and performance expectancy, which have been identified as critical determinants of user evaluation and behavioral outcomes [31,32]. These constructs reflect the extent to which super apps provide reliable, seamless, and efficient service integration across multiple functions. Perceived security is also included as a key stimulus factor, given its established role in shaping user trust and reducing perceived risk in digital environments, particularly in platforms that integrate multiple services and AI-driven interactions [33,34]. User satisfaction is introduced as an organism construct that reflects users’ internal evaluation of the platform experience, and continuance intention represents the behavioral response. In addition, this study incorporates AI system capability as a key platform-level construct that reflects the system’s ability to integrate, adapt, and personalize services across heterogeneous domains. From an affordance perspective, AI system capability functions as a mechanism that shapes how users experience and utilize system attributes, thereby influencing the extent to which functional affordances are effectively realized and translated into satisfaction and continuance intention.

3.1. Performance Efficacy

Performance efficacy refers to the extent to which users perceive that using a super-app enhances their ability to accomplish tasks effectively and achieve desired outcomes [32]. For this reason, performance efficacy has been widely examined in the IS literature as a key antecedent influencing users’ intentions to adopt new services. Shi et al. [35] demonstrated that the efficacy of recommendation systems in travel planning acts as a systematic factor that positively influences user cognitive trust. Luo et al. [36] found that performance efficacy, which represents users’ expectations regarding the performance of robot advisors in the financial industry, had the strongest influence on perceived value among several antecedent variables. Song et al. [37] verified that the performance efficacy of AI assistants is positively associated with users’ passion for technology. These findings suggest that users’ evaluations of a technology’s performance capability play an important role in shaping their perceptions and attitudes toward new digital services. Evaluating performance efficacy requires users to assess the capability and effectiveness of a system to support their tasks systematically and comprehensively. Super apps can also be considered a relatively new phenomenon in the mobile application ecosystem and have rapidly extended their scope to include the financial and non-financial sectors. Therefore, users’ perception of the effectiveness of super app functionality is a critical factor that shapes their overall evaluation. In this study, performance efficacy is conceptualized as the first stimulus factor and is operationally defined as the degree to which users believe that the functionality of a super app can enhance their task performance. Based on this reasoning, this study proposes that performance efficacy positively influences user satisfaction with super apps.

H1. 

Performance efficacy is positively related to satisfaction of super-app users.

3.2. Service Efficiency

Service efficiency refers to the extent to which the super-app enables users to complete tasks with minimal time and effort by providing streamlined and responsive service processes [38]. In other words, the perceived level of service efficiency can be understood as the extent to which users can access and utilize desired services or functions quickly and efficiently. In digital service environments, reduction in time and effort has become an important criterion for users’ evaluation of service value. Previous studies have also suggested a positive relationship between service efficiency and user satisfaction. Furtado et al. [39] noted that faster booking and payment processes are associated with higher levels of customer satisfaction. Similarly, El-Said and Al Hajri [40] empirically demonstrated that the speed of robot-based services positively affects users’ experiential satisfaction. These findings support the idea that efficient service delivery can significantly enhance users’ evaluations of their service experiences. Based on these studies, this study defines service efficiency as the extent to which a super-app ecosystem enables users to access and coordinate multiple services efficiently within a unified platform. Given the integrated nature of super apps, efficient access to various services is expected to influence user satisfaction with a platform positively. Accordingly, considering service efficiency as a second stimulus factor, we propose the following hypothesis:

H2. 

Service efficiency is positively related to satisfaction of super-app users.

3.3. Perceived Security

As super apps integrate numerous functions within a single platform, a greater number of user touchpoints may expose them to higher security risks than traditional standalone applications. Moreover, the potential scale of damage could be significantly larger because user data are centralized within the platform. Therefore, perceived security, which refers to the level of trust users have in the security of a specific service [41], can be considered a core antecedent that affects user satisfaction. The importance of perceived security has been consistently highlighted across diverse research contexts. Previous studies have shown that perceived security plays a critical role in shaping key user perceptions and behaviors. For example, Azhari et al. [42] found that perceived security is positively associated with perceived usefulness in mobile banking. Similarly, Kuo et al. [43] reported that perceived security enhances users’ continuance intention to use ERP systems, whereas Puspitasari et al. [44] demonstrate a positive effect on user satisfaction with mutual fund investment applications. Extending these findings to emerging technologies, Zhang [45] verified a significant relationship between perceived security and trust in face recognition systems, and Kwon and Lee [46] empirically validated the importance of security in the context of smart car software usage. Collectively, these findings suggest that perceived security serves as a fundamental determinant of user evaluation and behavioral intentions across various digital service environments. Based on these prior studies, the present study operationally defines perceived security as the level of trust that super-app users have in the security of a super app. Accordingly, we propose the following hypothesis:

H3. 

Perceived security is positively related to satisfaction of super-app users.

3.4. AI System Capability

AI system capability refers to a platform-level capability that enables the integration, adaptation, and personalization of heterogeneous services through AI-driven mechanisms, while facilitating users’ ability to realize functional affordances within the system. From an affordance perspective, AI plays a dual role in super-app environments. On the one hand, it introduces new action possibilities through functionalities such as personalized recommendations, intelligent assistance, and adaptive interactions. On the other hand, it serves as an enabling mechanism that supports the effective realization of functional affordances by facilitating seamless coordination and interaction across services.

In super-app environments, where multiple services are tightly integrated, AI system capability enhances user experience by enabling more efficient and effective task execution. By improving task efficiency, reducing cognitive effort, and supporting goal achievement, AI system capability contributes to more favorable evaluations of the platform. Prior research has shown that AI-driven functionalities significantly improve user experience and interaction effectiveness in digital environments [47,48,49]. Accordingly, higher levels of AI system capability are expected to directly enhance user satisfaction in super-app environments.

H4-1. 

AI system capability is positively related to the satisfaction of super-app users.

Performance efficacy, service efficiency, and perceived security represent key stimulus factors reflecting the functional affordances of super-app platforms, capturing users’ ability to achieve goals, optimize processes, and manage risks. However, the extent to which these affordances translate into user satisfaction may vary depending on the level of AI system capability. From an affordance perspective, AI system capability functions as an enabling mechanism that shapes how users realize platform functionalities. By providing adaptive and context-aware support, it allows users to more effectively utilize system attributes and navigate complex service environments.

As a result, the positive effects of performance efficacy and service efficiency on satisfaction are likely to be strengthened when the AI system capability is high, as users can more fully realize the benefits of system performance and efficiency [50,51]. Similarly, AI system capability may enhance the effect of perceived security by enabling users to better understand and manage potential risks in complex digital environments. Prior research suggests that perceptions of risk and uncertainty play a critical role in shaping user evaluations of digital platforms [52,53]. By reducing uncertainty and supporting informed decision-making, AI system capability can strengthen the positive impact of perceived security on user satisfaction.

H4-2. 

AI system capability moderates the relationship between performance efficacy and the satisfaction of super-app users.

H4-3. 

AI system capability moderates the relationship between service efficiency and the satisfaction of super-app users.

H4-4. 

AI system capability moderates the relationship between perceived security and the satisfaction of super-app users.

3.5. Satisfaction

Prior research in information systems (IS), marketing, and psychology has consistently demonstrated that user satisfaction positively influences continuance intention. Recent studies support this relationship. For example, Yang and Lee [24] reported a significantly positive relationship between user satisfaction and continuance intention in the context of remote services in smart cities. Similar findings have been reported in studies on government applications [53], social media platforms [54], mobile banking services [55], and intelligent personal assistants [56]. The same relationship can be reasonably expected in the context of super apps. As super apps integrate multiple services within a single platform, users are likely to continue using the platform when they are satisfied with the overall service experience. In other words, satisfaction with the integrated service environment may encourage users to maintain their relationship with super apps rather than switch to alternative standalone applications. Therefore, this study operationally defines satisfaction as users’ overall affective evaluation of their experience using a super app, including feelings of contentment and positive sentiments toward its features and services, and assumes that user satisfaction with super apps positively influences users’ continuance intentions.

H5. 

Satisfaction is positively related to continuance intention of super app users.

4. Methods

4.1. Instrument Development

The survey items in this study were derived from existing studies on SOR frameworks, super apps, and IS. Following recent research in AI and IS, system-level capabilities are often reflected through users’ perceived functional outcomes, as users evaluate the effectiveness of AI capabilities based on the benefits they experience [57]. Rather than directly measuring perceptions of specific AI features, the constructs of performance efficacy, service efficiency, and perceived security were designed to capture users’ evaluations of core functional affordances within AI-enabled super app environments, consistent with the conceptualization of AI as an embedded system-level capability rather than an isolated feature-level attribute.

Table 1. Survey items.

Construct	Item No.	Items	References
Performance efficacy	PE1	Using the super app improves my efficiency in completing daily tasks.	[58]
	PE2	Using the super app enhances the convenience of managing various services.
	PE3	The super app helps me handle activities more effectively.
Service efficiency	SE1	The super app includes many of the services, brands, or shops I prefer without requiring additional app downloads.	[11,59]
	SE2	The super app integrates diverse businesses and resources, and offers shared activities (e.g., promotions, events) within the platform ecosystem.
	SE3	The super app allows me to use consistent interfaces and mechanisms without needing to learn a new app.
Perceived security	PS1	The super app can ensure the security of my personal information and data.	[53]
	PS2	It is safe to conduct various transactions or service processes through the super app.
	PS3	The super app operates reliably and is unlikely to become unavailable due to system failures.
AI System capability	ASC1	The AI in the super app provides comprehensive information and support.	[53,57,60]
	ASC2	The AI in the super app provides information that is tailored to my personal needs.
	ASC3	The AI in the super app responds promptly to my requests.
	ASC4	The AI in the super app quickly resolves problems I encounter.
Satisfaction	SA1	I am content with my experience using the super app.	[58]
	SA2	I am pleased with my experience using the super app.
	SA3	I am delighted with my experience using the super app.
	SA4	Overall, I am satisfied with the super app.
Continuance intention	CI1	I intend to continue using the super app.	[58]
	CI2	I would keep using the super app if access is available.
	CI3	I will continue using the super app in the future.
	CI4	I plan to use the super app regularly in the coming days.

presents the 24 measurement items for the seven constructs, which were evaluated using a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

4.2. Data Collection

To examine the determinants of users’ continuance intention toward super apps, this study conducted an online survey in January 2026 through a professional research agency. Data were collected separately for two user groups in South Korea: 306 responses from financial super-app users and 308 responses from non-financial super-app users, resulting in 614 valid observations. The survey comprised three sections. The first section introduced the purpose of the study and asked a screening question to ensure respondent eligibility by confirming whether the participants were current super-app users. The second section contained the measurement items for the focal constructs in this study. The final section gathered the respondents’ demographic information.

To address potential common method bias, the unmeasured latent marker construct (ULMC) approach was first applied. The results indicate that the inclusion of a latent marker variable leads to less than a 10% increase in the R² values of all endogenous constructs, suggesting that common method bias is not a serious concern [61]. In addition, a full collinearity test based on variance inflation factor (VIF) values was conducted, given that the data were collected using a single survey instrument. The results show that multicollinearity is not an issue, as all VIF values range from 1.000 to 2.904, which is below the recommended threshold of 5 [62].

Table 2. Sample demographics.

	Respondents (n = 614)
Characteristics	Number	Percentage
Gender
Male	250	40.7%
Female	364	59.3%
Age
20–29	57	9.3%
30–39	153	24.9%
40–49	185	30.1%
50–59	156	25.4%
60+	63	10.3%
Education
~High school	118	19.2%
University	413	67.3%
Graduate school	73	11.9%
Etc.	10	1.6%
Monthly Income ($)
Less than 2000	135	22.0%
2000~3500	228	37.1%
3500~5000	140	22.8%
5000+	111	18.1%
Occupation
Self-employer	75	12.2%
Start-up/SME worker	271	44.1%
Major company worker	32	5.2%
Public service worker	92	15.0%
NGO worker	25	4.1%
Freelancer/Etc.	119	19.4%
Group
Financial super app users	306	49.8%
Non-Financial super app users	308	50.2%

presents the demographic characteristics of the sample.

4.3. Measurement Model Validation

To assess the research model and examine the proposed hypotheses, we applied partial least squares structural equation modeling (PLS-SEM) using SmartPLS 4. This analytical method was chosen based on its ability to maximize the explained variance of endogenous constructs and produce stable path estimates, particularly when the data deviate from the normal distribution assumptions, a characteristic observed in the present dataset [63,64]. The measurement model demonstrated satisfactory convergence and internal consistency. Specifically, all factor loadings exceeded 0.6, average variance extracted (AVE) values were above 0.5, composite reliability (CR) values exceeded 0.6, and Cronbach’s alpha values were greater than 0.7, thereby meeting commonly accepted thresholds [65] (See

Table 3. Construct validity and reliability.

Construct	Items	Factor Loading	Std. Error	t-Value	AVE (>0.5)	Composite Reliability (>0.7)	Cronbach’s Alpha (>0.7)
Performance efficacy	PE1	0.927	0.007	127.755	0.852	0.945	0.913
	PE2	0.923	0.008	113.642
	PE3	0.920	0.007	131.658
Service efficiency	SE1	0.881	0.018	48.241	0.785	0.916	0.863
	SE2	0.889	0.011	78.640
	SE3	0.889	0.011	77.823
Perceived security	PS1	0.926	0.008	114.253	0.857	0.947	0.917
	PS2	0.941	0.005	176.006
	PS3	0.911	0.009	101.218
AI System capability	ASC1	0.817	0.017	48.611	0.706	0.906	0.861
	ASC2	0.791	0.021	38.382
	ASC3	0.852	0.014	62.735
	ASC4	0.897	0.009	98.419
Satisfaction	SA1	0.933	0.006	146.752	0.857	0.960	0.944
	SA2	0.937	0.006	152.032
	SA3	0.898	0.010	92.314
	SA4	0.935	0.007	138.936
Continuance intention	CI1	0.933	0.007	134.059	0.840	0.954	0.936
	CI2	0.910	0.010	87.253
	CI3	0.909	0.010	93.575
	CI4	0.913	0.008	111.707

). Discriminant validity was further evaluated using the Fornell–Larcker criterion, cross-loading examination, and heterotrait–monotrait (HTMT) ratio (See

Table 4. Fornell–Larcker criterion.

	PE	SE	PS	ASC	SA	CI
PE	0.923
SE	0.737	0.886
PS	0.471	0.495	0.926
ASC	0.681	0.625	0.426	0.840
SA	0.813	0.680	0.604	0.660	0.926
CI	0.808	0.670	0.548	0.623	0.887	0.916

,

Table 5. Construct cross-loadings.

	PE	SE	PS	ASC	SA	CI
PE1	0.927	0.680	0.419	0.634	0.751	0.740
PE2	0.923	0.682	0.417	0.596	0.733	0.753
PE3	0.920	0.679	0.469	0.655	0.766	0.744
SE1	0.617	0.881	0.417	0.524	0.583	0.571
SE2	0.670	0.889	0.393	0.551	0.603	0.596
SE3	0.670	0.889	0.504	0.583	0.620	0.611
PS1	0.360	0.407	0.926	0.350	0.514	0.441
PS2	0.453	0.464	0.941	0.407	0.572	0.536
PS3	0.487	0.498	0.911	0.421	0.586	0.537
ASC1	0.599	0.529	0.423	0.817	0.577	0.562
ASC2	0.526	0.472	0.227	0.791	0.490	0.465
ASC3	0.562	0.537	0.335	0.852	0.533	0.501
ASC4	0.598	0.556	0.425	0.897	0.609	0.556
SA1	0.777	0.634	0.537	0.618	0.933	0.822
SA2	0.783	0.629	0.518	0.625	0.937	0.830
SA3	0.693	0.606	0.609	0.588	0.898	0.798
SA4	0.755	0.648	0.576	0.615	0.935	0.833
CI1	0.768	0.642	0.483	0.575	0.831	0.933
CI2	0.753	0.609	0.531	0.602	0.822	0.910
CI4	0.734	0.596	0.410	0.528	0.764	0.909
CI4	0.706	0.607	0.578	0.575	0.829	0.913

and

Table 6. HTMT Ratio.

	PE	SE	PS	ASC	SA
SE	0.829
PS	0.511	0.553
ASC	0.766	0.723	0.470
SA	0.874	0.753	0.648	0.730
CI	0.874	0.744	0.586	0.691	0.942

). Although several HTMT values (0.874–0.942) marginally exceeded the recommended threshold of 0.85, the bias-corrected confidence intervals did not contain a value of 1, indicating that the discriminant validity among the constructs remained acceptable [66].

5. Results

Table 7. Summary of hypotheses testing results (Decomposed Model).

Hypothesis	Financial Super App Users				Non-Financial Super App Users
	Path Coefficient	S.E	t-Value	Supported	Path Coefficient	S.E	t-Value	Supported
H1: PE → SA (+)	0.577	0.054	10.672	Yes	0.520	0.056	9.265	Yes
H2: SE → SA (+)	0.075	0.053	1.410	No	0.038	0.055	0.693	No
H3: PS → SA (+)	0.206	0.037	5.512	Yes	0.308	0.042	7.247	Yes
H4-1: ASC → SA (+)	0.141	0.052	2.720	Yes	0.139	0.053	2.632	Yes
H4-2: ASC × PE → SA (+)	0.021	0.042	0.492	No	−0.005	0.047	0.097	No
H4-3 ASC × SE → SA (+)	0.003	0.038	0.077	No	−0.066	0.051	1.289	No
H4-4: ASC × PS → SA (+)	−0.075	0.041	1.818	Yes	0.113	0.040	2.809	Yes
H5: SA → CI (+)	0.902	0.012	73.570	Yes	0.872	0.017	50.748	Yes

presents the results of the PLS-SEM analysis based on a bootstrapping procedure to test the proposed hypotheses. The analysis was conducted separately for financial and non-financial super-app users to examine potential group differences. For financial super-app users, performance efficacy showed a significant positive effect on satisfaction (H1: β = 0.577, t = 10.672), supporting H1. In contrast, service efficiency did not significantly influence satisfaction (H2: β = 0.075, t = 1.410), leading to the rejection of H2. Perceived security was found to have a significant positive impact on satisfaction (H3: β = 0.206, t = 5.512), supporting H3. In addition, AI system capability (ASC) showed a significant positive effect on satisfaction (H4-1: β = 0.141, t = 2.720), supporting H4-1. Regarding the moderating effects, ASC did not significantly moderate the relationship between performance efficacy and satisfaction (H4-2: β = 0.021, t = 0.492), or between service efficiency and satisfaction (H4-3: β = 0.003, t = 0.077), leading to the rejection of H4-2 and H4-3. However, ASC significantly moderated the relationship between perceived security and satisfaction (H4-4: β = −0.075, t = 1.818), supporting H4-4. Finally, satisfaction showed a strong positive effect on continuance intention (H5: β = 0.902, t = 73.570), supporting H5. For non-financial super-app users, performance efficacy also showed a significant positive effect on satisfaction (H1: β = 0.520, t = 9.265), supporting H1. Similar to the financial group, service efficiency did not significantly influence satisfaction (H2: β = 0.038, t = 0.693), resulting in the rejection of H2. Perceived security significantly enhanced satisfaction (H3: β = 0.308, t = 7.247), supporting H3. ASC also demonstrated a significant positive effect on satisfaction (H4-1: β = 0.139, t = 2.632), supporting H4-1. With regard to moderating effects, ASC did not significantly moderate the relationship between performance efficacy and satisfaction (H4-2: β = −0.005, t = 0.097) or the relationship between service efficiency and satisfaction (H4-3: β = −0.066, t = 1.289), leading to the rejection of H4-2 and H4-3. However, ASC significantly moderated the relationship between perceived security and satisfaction (H4-4: β = 0.113, t = 2.809), supporting H4-4. Satisfaction strongly predicted continuance intention (H5: β = 0.872, t = 50.748), supporting H5.

To further examine the internal mechanisms of the model, a mediation analysis was conducted (

Table 8. Mediation test.

	Mediated Path	Original Sample	t-Value	Significance (p < 0.05)	95% CIs (Bias-Corrected)
Financial Super-App Users	PE → SA → CI	0.521	10.468	0.000	[0.435, 0.598]
	SE → SA → CI	0.067	1.410	0.079	[−0.014, 0.143]
	PS → SA → CI	0.186	5.557	0.000	[0.133, 0.243]
	ASC → SA → CI	0.127	2.713	0.003	[0.062, 0.215]
Non-Financial Super-App Users	PE → SA → CI	0.453	8.756	0.000	[0.368, 0.538]
	SE → SA → CI	0.033	0.694	0.244	[−0.046, 0.111]
	PS → SA → CI	0.268	7.081	0.000	[0.206, 0.330]
	ASC → SA → CI	0.121	2.661	0.004	[0.050, 0.200]

). The results indicate that satisfaction significantly mediates the relationship between performance efficacy and continuance intention for both financial (β = 0.521, t = 10.468, p < 0.001) and non-financial users (β = 0.453, t = 8.756, p < 0.001). Similarly, perceived security shows a significant indirect effect on continuance intention through satisfaction in both groups (financial: β = 0.186, t = 5.557, p < 0.001; non-financial: β = 0.268, t = 7.081, p < 0.001). In addition, ASC exhibits a significant indirect effect on continuance intention via satisfaction (financial: β = 0.127, t = 2.713, p < 0.01; non-financial: β = 0.121, t = 2.661, p < 0.01). In contrast, the indirect effect of service efficiency on continuance intention is not significant in either group (financial: β = 0.067, t = 1.410, p = 0.079; non-financial: β = 0.033, t = 0.694, p = 0.244). All significant indirect effects were supported by 95% bias-corrected confidence intervals that excluded zero, further confirming the mediating role of satisfaction in the proposed model.

Additionally, a multi-group analysis (PLS-MGA) was conducted to examine the potential differences between financial super-app users (n = 306) and non-financial super-app users (n = 308). The three-step measurement invariance of the composite model procedure was applied to ensure the validity of the group comparison prior to performing the MGA. The results confirmed configural invariance (Step 1), compositional invariance (Step 2), and equality of the composite mean values and variances (Step 3), thereby establishing full measurement invariance. These results indicate that meaningful comparisons of the structural path coefficients between the two groups can be conducted. The MGA results revealed that most structural relationships did not differ significantly between financial and non-financial super-app users (see

Table 9. Multi-group analysis.

	Differences	Significance (p < 0.05)	Supported
H1: PE → SA (+)	−0.057	0.231	No
H2: SE → SA (+)	−0.037	0.316	No
H3: PS → SA (+)	0.102	0.038	Yes
H4-1: ASC → SA (+)	−0.001	0.498	No
H4-2: ASC × PE → SA (+)	−0.025	0.329	No
H4-3 ASC × SE → SA (+)	−0.069	0.137	No
H4-4: ASC × PS → SA (+)	0.188	0.001	Yes
H5: SA → CI (+)	−0.031	0.070	No

Note: Difference values are calculated as βNon-financial − βFinancial.

). Specifically, no significant group differences were observed in the direct effects of performance efficacy, service efficiency, or ASC on satisfaction, as well as in the effect of satisfaction on continuance intention. Similarly, the moderating effects of ASC on the relationships between performance efficacy and satisfaction and between service efficiency and satisfaction did not significantly differ between the two groups.

However, several relationships showed statistically significant differences between the two user groups. In particular, the effect of perceived security on satisfaction differed significantly between financial and non-financial super-app users (Δβ = 0.102, p < 0.05). In addition, the moderating effect of ASC on the relationship between perceived security and satisfaction also differed significantly between the two groups (Δβ = 0.188, p < 0.01). These findings suggest that whereas the core structural relationships remain stable across financial and non-financial super apps, the role of perceived security and its interaction with AI system capability varies depending on the context of super-app usage.

6. Discussion

6.1. Key Findings

The results of the analysis revealed several important findings. First, the structural relationships among the core SOR variables show largely consistent patterns across both groups. PE and PS significantly affected SA, whereas SE did not show a significant impact. SA, in turn, strongly influenced CI. These findings indicate that users of financial and non-financial super apps evaluate their platform experiences primarily based on functional effectiveness and perceived security, rather than service efficiency. In particular, the non-significant relationship between SE and SA across both groups suggests that efficiency alone may not be sufficient to shape user satisfaction on complex multi-service platforms, where functional effectiveness and perceived security play important roles. One possible explanation is that super apps integrate multiple services within a single platform ecosystem, which may lead users to evaluate the platform based on overall functionality and security rather than on basic operational attributes such as speed or timeliness. As users increasingly rely on super apps to perform a variety of tasks within a single interface, the relative importance of service efficiency may diminish compared to performance- and security-related factors. In mature super-app ecosystems, seamless and efficient service integration may already be perceived as a baseline or hygiene expectation rather than a distinctive source of satisfaction. As a result, users may place greater emphasis on whether the platform effectively supports task accomplishment, personalized service coordination, and secure interactions across heterogeneous service domains. This finding is consistent with prior research suggesting that efficiency-related attributes do not always serve as decisive determinants of technology adoption or service evaluation in complex service environments. For example, Sheel and Nath [67] found no significant effect of effort efficiency on customer intention to adopt blockchain technology to enhance supply chain security. Similarly, Shajrawi and Aburub [68] found no significant association between responsiveness and service differentiation in the context of ERP adoption in the hotel industry. More recently, Chau et al. [69] also reported that usability and responsiveness were not the strongest predictors of sustained satisfaction in AI-powered customer service environments, whereas users placed greater emphasis on the tangible value delivered by AI systems.

Second, the moderating effects of ASC revealed meaningful differences across the two user groups. For financial super-app users, ASC does not significantly moderate the relationships between performance efficacy and satisfaction or between service efficiency and satisfaction. However, ASC shows a significant moderating effect on the relationship between perceived security and satisfaction. This finding suggests that AI-enabled system capabilities—such as intelligent assistance, automation, and data-driven service support—may influence how users interpret security-related attributes of the platform. This is consistent with prior studies suggesting that AI-enabled functionalities can enhance users’ evaluation of platform performance and trust by generating functional value perceptions [60]. When such capabilities are effectively embedded in the system, they may reinforce users’ evaluation of platform reliability and trustworthiness, thereby shaping the impact of perceived security on satisfaction.

For non-financial super-app users, a similar pattern is observed. ASC does not significantly moderate the relationships between performance efficacy and satisfaction or between service efficiency and satisfaction, but it significantly moderates the relationship between perceived security and satisfaction. This may indicate that, in non-financial super-app environments, AI system capability functions as an important signal that enhances users’ confidence in the platform. As users recognize the functional benefits enabled by AI capabilities, such as personalized services and intelligent recommendations, these capabilities may strengthen the role of perceived security in shaping satisfaction.

Finally, the MGA results provide additional insights into cross-group differences. The analysis reveals that most structural relationships, including the direct effects of PE, SE, ASC, and SA, do not differ significantly between users of financial and non-financial super apps. However, two relationships show statistically significant differences across the two groups: the effect of PS on SA and the moderating effect of ASC on the relationship between PS and SA. These results indicate that, although the fundamental drivers of continuance intention remain stable across super-app contexts, the role of perceived security—and its interaction with AI system capability—differs depending on the type of super-app service.

Overall, the findings highlight that core platform attributes consistently shape satisfaction and continuance intention, whereas AI system capability influences how users interpret these attributes across different super-app contexts.

6.2. Theoretical and Practical Implications

Building on the empirical findings presented here, this study provides several theoretical implications for the literature on super apps and AI-enabled digital platforms. First, it extends the SOR framework to the emerging context of AI-enabled digital service ecosystems represented by super apps. Whereas prior studies applying the SOR framework have primarily focused on traditional service attributes or technology acceptance factors, relatively limited attention has been paid to how AI-enabled functionalities influence user evaluations of digital platforms. Building on this gap, this study advances the SOR framework by conceptualizing AI system capability as a system-level construct that shapes user responses in complex platform environments.

Second, this study contributes to the literature by reconceptualizing the role of AI in digital platforms from an affordance perspective. Rather than treating AI as a set of discrete features or perceived value constructs, the findings highlight that AI operates as a system-level capability that both introduces new functionalities and enables users to effectively realize functional affordances embedded in the platform. This dual role provides a more nuanced understanding of how AI shapes user evaluations, demonstrating that its impact extends beyond direct performance enhancement to include facilitating users’ ability to utilize and benefit from platform attributes. By positioning AI system capability as an affordance-enabling mechanism, this study offers a theoretical advancement in understanding AI-enabled service systems.

Finally, this study contributes to the super-app literature by demonstrating that the effects of AI system capability are contingent on service context. By distinguishing between financial and non-financial super apps, the findings show that the extent to which AI system capability enhances user satisfaction and strengthens the effects of functional affordances varies across different service environments. Financial super apps are typically associated with goal-oriented and risk-sensitive tasks, whereas non-financial super apps focus more on convenience and experiential services. The results suggest that AI system capability plays a particularly important role in helping users navigate complexity and manage uncertainty in these contexts. This finding highlights the importance of considering service context when examining the role of AI in shaping user evaluations and behavioral outcomes within complex digital platform ecosystems.

The practical implications of this study are as follows: First, super-app service providers should develop various strategies to enhance PE and PS and continuously improve user satisfaction. Firms should ensure that users can complete multiple tasks within a single integrated ecosystem to maximize the advantages of super-app platforms. This requires not only improving the performance of individual platform functions but also implementing seamless integration among different services within the platform. In addition, because super apps provide a wide range of functions within a single environment, continuous reinforcement and regular updates of security modules are necessary to ensure that users can safely utilize various platform services.

Second, the findings highlight the importance of AI system capability as a key driver of user satisfaction and platform effectiveness. Super-app providers should focus on strengthening AI system capabilities that support personalization, intelligent assistance, and adaptive interactions, as these functionalities enable users to perform tasks more efficiently and navigate complex service environments. By enhancing AI system capability, firms can help users better utilize platform functionalities and achieve their goals more effectively, thereby improving overall user experience.

Finally, as AI system capability strengthens the effects of platform attributes on user satisfaction, firms should recognize its role in helping users better understand, utilize, and benefit from platform functionalities. In particular, AI can support users in managing complexity and uncertainty in multi-service environments, thereby enhancing the effectiveness of both performance-related and security-related features. To maximize these benefits, firms should ensure transparency in AI-supported processes and clearly communicate how AI functionalities assist users in achieving their goals and maintaining secure interactions.

7. Conclusions

This study investigated the factors influencing users’ continuance intention to use super apps in an AI-enabled complex digital ecosystem environment. To achieve this objective, a comprehensive research model grounded in the SOR framework was developed to capture system-level user evaluations. The model conceptualizes PE, SE, PS, and ASC as stimulus factors; SA as the organism; and CI as the behavioral response. In addition, this study extends the traditional SOR framework by incorporating AI-enabled affordances, specifically ASC, as a moderating variable and by comparing financial and non-financial super-app users through MGA. The findings demonstrate strong explanatory capability for predicting users’ continuance intention toward super apps. In particular, the results highlight the important roles of platform attributes such as performance effectiveness and perceived security in shaping user satisfaction and continuance intention. Furthermore, the study reveals that AI-enabled affordances influence how users interpret platform attributes and that these effects vary depending on the service context of super apps. By distinguishing between financial and non-financial super-app environments, this study provides a more nuanced understanding of how AI-enabled functionalities interact with platform characteristics to influence user evaluations and behavioral intentions. Overall, the findings offer meaningful insights for both academic researchers and practitioners seeking to understand user behavior within complex AI-enabled digital service ecosystems.

Despite these contributions, this study has several limitations. First, although this study conducted a comparative analysis of financial and non-financial super apps, the empirical data were collected solely from users in South Korea. As super apps are more mature and widely adopted in East Asian markets, particularly in countries such as South Korea and China, user perceptions of integrated services and AI-enabled functionalities may differ from those in Western contexts, where platform ecosystems tend to be more fragmented and less centralized. These structural and contextual differences may influence how users evaluate platform attributes and interact with AI system capabilities, thereby limiting the generalizability of the findings to other regions. Future research should extend this study by collecting data from multiple countries, particularly from Western markets, to examine whether the relationships identified in this study hold across different technological and institutional environments. Such comparative analysis would provide a more comprehensive understanding of user behavior in global super-app ecosystems. Second, although this study conceptualized AI as a system-level capability embedded within the super app ecosystem, the measurement constructs of performance efficacy, service efficiency, and perceived security primarily focused on users’ evaluations of general functional affordances rather than AI-specific attributes. Accordingly, this study did not directly measure users’ perceptions of specific AI characteristics such as perceived intelligence, explainability, personalization, adaptiveness, or autonomous coordination capabilities. Future research could incorporate more explicit AI-related dimensions to provide a more fine-grained understanding of how users evaluate AI-enabled platform experiences and how AI-specific attributes shape user satisfaction and continuance intention within super-app ecosystems.