**1. Introduction**

Digitalization describes the introduction and usage of digital technologies in a social, individual, and organizational context [1]. In 2018, a survey of 3958 information technology (IT) leaders revealed that 61% see higher revenue growth than their competition when using digital technologies [2]. To achieve competitive advantage through digitalization, organizations have to massively transform their organizational structures, strategies, methods, business models, and enterprise architectures [1]. In the past, the implementation of new IT was mainly the task of the organization and the IT department; nowadays, user-friendly IT allows business units to shape digitalization [3]. If the business unit implements a new IT system on its own without a central IT department being involved during the development or the subsequent control, the phenomenon is called shadow IT [4]. The effect of shadow IT on the enterprise architecture of an organization has two perspectives: On the one hand, shadow IT introduces innovation into an organization and allows a higher flexibility for business units [5,6]. On the other hand, inefficiencies occur [7]: The heterogeneity rises, because business units choose technology on their own. Complexity increases, because shadow IT is connected or exists parallel to

formal systems. Integration, by linking or unifying shadow IT and the redundant enterprise system, could eliminate redundancies and solve associated problems [8]. A survey of 490 CIOs revealed that 64% regard IT integration as a priority, while simultaneously desiring a high degree of innovation and flexibility in the IT architecture of their organization [9]. Thereby, organizations find themselves in an area of conflict, as IT integration might eliminate the benefits that shadow IT offers [10]. Because organizations must be aware of these drawbacks before performing an IT integration, this paper aims at presenting a framework for shadow IT integration drawbacks.

Most studies on IT integration focus on the benefits and provide classifications for its factors [11]. Research on IT integration drawbacks focuses merely on monetary factors such as indirect and direct costs [12,13], or states that some integration technologies are more expensive than others [14,15]. Research on non-monetary factors considers IT integration barriers in environments such as hospitals [16] or governments [17] or for special methods like enterprise application integration (EAI) [15,18]. Thereby, none of these studies focus on the phenomenon of shadow IT [11]. Additionally, IT integration research in general lacks a link with the existing theory base [19]. Therefore, this paper contributes to the scientific discussion as it presents a theory-based view of shadow IT integration drawbacks using the systemic theories of path dependency and switching costs. Besides, practitioners can use the resulting framework to assess the drawbacks when coming to a decision on shadow IT integration.

The paper is structured as follows: At first, we introduce the problem that shadow IT causes in the enterprise architecture. Then, we illustrate our research approach consisting of our systemic theory base and the literature review that we conducted. Afterwards, we present the results and discuss them. Finally, we provide a conclusion and note possible future research directions.

#### **2. Shadow IT in the Enterprise Architecture**

Shadow IT describes IT systems that business units implement individually in their business processes, whereby they are not involved in an organizational IT managemen<sup>t</sup> [20]. On the one hand, it has technological aspects because shadow IT occurs in various forms, such as local applications, spreadsheets, end devices, cloud services, or combined solutions [5], and needs technical support to function in an organization [6]. Yet, shadow IT also has social components because business employees are highly involved during its implementation and usage [21,22]. Therefore, we regard shadow IT as a socio-technical phenomenon [4].

In the beginning, shadow IT is often experimental and small, as it responds to an emergen<sup>t</sup> need in the business unit [5,22]. Once established in the organization, shadow IT can grow large because business units share the benefits that the system provides for them [23]. Due to inertia on an individual as well as an organizational level, business users continue using shadow IT [22]. Thereby, it often gets intertwined with the enterprise architecture of an organization [24]. Shadow IT reinforces by emerging and reemerging in a cycle of time and cost pressures [4]. Thereby, it shapes the enterprise architecture and can become an important part of it [25]. Shadow IT exists alongside formal enterprise systems and either complements, expands, or supplements them [26]. Studies show that a redundancy of data or functionality exists in a majority of the cases [26], and that as a result shadow IT causes various inefficiencies in the enterprise architecture [7]: First, IT departments often do not know about shadow IT, which leads to an non-transparent enterprise architecture, the inability to manage it, and related risks. Second, responsibilities are often unclear, which reduces business-IT-alignment. Third, shadow IT increases the complexity of the enterprise architecture in various ways. The low standardization and integration and high heterogeneity prevent automation [10] and thereby hinder digitalization [27].

Organizations can solve these inefficiencies by converting shadow IT into business-managed IT. Thereby, they identify and include shadow IT in IT managemen<sup>t</sup> [5,28]. However, redundancies of shadow IT and enterprise systems will remain, and organizations must take architectural measures to solve them. IT integration is an established concept to cope with these types of problems [16]. A common database or data interface can solve data redundancies, and a unification of shadow IT and the redundant enterprise system can remove functional redundancies [8,29].

*Systems* **2018**, *6*, 42

Like in other IT integration decisions [30], organizations must valuate drawbacks and benefits of an integration of shadow IT and the redundant enterprise system. While many studies analyze IT integration benefits, no study discusses the drawbacks [11]. Some studies provide a classification based on direct costs, such as the implementation of the IT integration, and indirect costs such as training of employees [12,13]. Others compare different costs for different integration technologies, such as interfaces and unification of systems [14,29] or methods such as EAI versus point-to-point integration [17,31]. A discussion of non-monetary costs exists in the analysis of IT integration barriers. Here, research presents barriers that can occur during integrating systems, such as resistance to change [18] or technical incompatibility [32]. Overall, research lacks a synthesis of the discussions on monetary and non-monetary costs and a theory-based specification of these drawbacks to the phenomenon of shadow IT. Shadow IT research has started to target integration by pointing out redundancies [26] or providing evidence based on a small number of cases [27]. None of the research on shadow IT integration has presented a theory-base yet, and IT integration research lacks that in general [19]. Additionally, existing research on IT integration focuses on the benefits of an IT integration [11]. By providing a theory-based synthesis of the drawbacks, we can close this research gap. As a result, organizations can weigh the drawbacks against the benefits to come to an integration decision of shadow IT and enterprise systems. To be able to achieve this goal, we pose the following research question: Which drawbacks do organizations face when deciding about the integration of shadow IT and an enterprise system?

## **3. Research Approach**

To answer our research question, this section first presents our theory background. Then, we show our research method that led to the conceptual model that we present in the next section.

#### *3.1. A Systemic Viewpoint on Shadow IT Integration: Path-Dependency and Switching Costs*

Path dependency is a concept introduced from evolutionary economics and explains occurring inefficiencies in a complex system from a systemic viewpoint [33]. Coming from the discussion on economical processes, the theory explains that historic choices of technology combined with several types of self-reinforcing effects, such as economies of scale, emotional reactions, or political processes that lead to increasing returns and the establishment of a dominant design [34]. However, after the dominant design has been established, markets may reside in an inefficiency, where a seemingly less appropriate solution has the greatest market share although other solutions might be technologically more appropriate [34]. In this inefficient state, users can no longer freely switch to another technology, but find themselves in a lock-in with the current dominant technology [33].

In this lock-in situation, the costs to switch to another technology are very high due to the high intertwinement of the technology with the organization [35]. These so-called switching costs are originally defined as "onetime costs that customers associate with the process of switching from one provider to another" [36], but are also applicable for switching technologies in an organization [35]. Switching costs are not only monetary costs but also include emotional or cognitive costs, such as the search for a new technology; learning; transaction costs; and costs due to loyalty, habit, and emotion [37]. Burnham categorizes eight switching costs into financial, procedural, and emotional types of switching costs using factor analyses on a survey on perceptions of 144 customers to change their service provider [36].

Information systems research increasingly recognizes that path-dependency is a relevant concept in the field of enterprise architecture [35,38] and suggests that it is an important theory base for the research field of IT integration [19]. We assume that path-dependency is suitable to shed light on the specific problem of shadow IT due to the following reasons: the last section explained that various effects lead to the intertwinement of shadow IT with the organization [4,23], and the high involvement of business users during its implementation and usage [21]. This points to self-reinforcing effects that affect the evolution of the enterprise architecture and switching costs that arise. Often, shadow IT even becomes an important part of the enterprise architecture [24]. In these cases, the historic choice of shadow IT implementation can lead to various inefficiencies in terms of transparency, redundancy, and governance in the enterprise architecture, where organizations find themselves with resulting problems such as missing automation, regulatory requirements, data integrity, and unclear responsibilities [26]. These facts sugges<sup>t</sup> that the self-reinforcing effects can even lead to a lock-in with the existing shadow IT.

In cases where the lock-in and the occurring inefficiencies stem from the redundancy of the shadow IT with the enterprise system, organizations can integrate the two systems [8]. Thereby, business units must change from using the old shadow IT to using the new, integrated system. Thereby, they must adapt their work routine. We therefore regard the transition of the old system to the new system as a switch and apply the concept of switching costs to assess the shadow IT integration drawbacks.

## *3.2. Research Method*

Because IT integration is a mature topic with an existing body of research [19], we used a literature review to develop our conceptual model [39]. A lot of the conducted research in the field is done as case study research or practitioner surveys [19], which assure that our study is also practically relevant. Additionally, our research method needed to reflect the fact that shadow IT is a socio-technical phenomenon [4]. Therefore, we chose the Path Biography Method (PBM) to conduct the analysis in our literature review [40]. The approach of the PBM goes beyond the mainly quantitative-empirical methods of IS research focusing on technical aspects [41] as well as the managemen<sup>t</sup> research that mainly concentrates on organizational aspects [42]. Rather, the PBM integrates both aspects of our research problem [40]. Additionally, it is useful for research areas that span disciplinary boundaries [40] and, although it is a fairly recent method, research suggests that it is relevant in the area of IT integration [43].

Figure 1 shows our research method. Our literature review consisted of a structured literature search and the analysis, where we used the PBM to conduct the coding.

**Figure 1.** Research Method.

The first step in our research procedure was the collection of possible shadow IT integration drawbacks from shadow IT integration literature. Research on this specific topic is scarce and has covered its drawbacks only briefly [11]. Therefore, we expanded our search to literature on IT integration in general. Thereby, we consulted an existing review on IT application integration [11] and complemented these findings by a search on literature starting from 2017 on using the keywords *integration costs*, *integration barriers,* or *integration drawbacks* in title or abstract in the databases of IEEE, AISeL, and Sciencedirect. Additionally, because possible drawbacks from integrating shadow IT mainly stem from its loss and the following loss of its benefits [27], we scanned a former review on shadow IT on literature of shadow IT benefits [28]. We complemented this review by conducting a literature search in IEEE, AISeL, and Sciencedirect using the keywords *shadow/feral systems,* and

*shadow/grey/hidden/rogue IT* in combination with *information technology/services/systems/security,* in title or abstract from 2017 on. After we finished the literature collection, we scanned it for specific drawbacks, excluded those that made only vague or relative statements or did not mention any specific factors, and in the end, removed duplicates.

In a second step, we used an approach of coding [44] to assign the found criteria to pre-set codes following three sub-steps: First, we used the concept of Burnham that divides switching costs into procedural, financial, and relational costs [36]. Principle 1 of the PBM requires one to focus on the self-reinforcing effects that cause the inefficiency [40], which is, in our case, the redundancy of shadow IT and the enterprise system. To adapt the model to this specific situation and make it valuable for shadow IT research, we added a fourth cost category that accounts for the loss of shadow IT benefits. Second, to comply with principle 2 of the PBM, we additionally differentiated between the technical, organizational, and level-spanning drawbacks. Third, we mapped them to the local level, the global level, or the interaction of the levels following the third principle of the PB. Thereby, our knowledge from shadow IT literature and the fact that most of the literature was practical case study research helped us to evaluate the criteria on a fit to the specific phenomenon of shadow IT integration. One author conducted the analysis and discussed the results with the other authors. In several iterations, we thereby refined and evaluated the criteria and their mapping. As a result, a three-dimensional framework of shadow IT integration drawbacks emerged that we will present in the next chapter.
