Value Creation within AI-enabled Data Platforms

Introduction

Digitalization changed the view on how firms create value. Today, one will hear frequently that ‘data is the new oil’, a phrase coined by Clive Humby (Palmer, 2006) and later used inter alia by European Consumer Commissioner Meglena Kuneva (2009) or German Chancellor Angela Merkel (SZ, 2015). The analogy puts it straight that data became an additional factor of production in the twenty-first century. Nevertheless, it is unclear how value creation will be achieved, if one compares the production of petrol from oil, with a ‘production’ based on data.

In his book Competitive Advantage, Porter (1985) introduced the ‘value chain’ as a concept to understand the internal activities in a firm (see Della Corte & Del Gaudio, 2014, for a literature review). Additional to the primary activities along the value chain, Porter described secondary activities, which include the usage of information technology (IT) for support processes such as human resource management or financial accounting. In the 1990s, alternative concepts about activities of a firm and competition were proposed, such as ‘value constellation’ between external partners in the economy (Normann & Ramírez, 1993), ‘hypercompetition’ (D’Aveni, 1994) or ‘disruption’ (Bower & Christensen, 1995). In the last decade, new perceptions in marketing triggered the development of complementing approaches for conceptualizing value from an external perspective ¹

This external perspective delineates the creation of value as a process of economic exchange between a firm, customers ²

Stabell and Fjeldstad (1998) extended Porter’s single model to a triad of ‘the value chain, the value shop and the value network’. They introduced three distinct models to understand the value creation within firms, supported by different types of technologies such as conveyor belt, emergency operating theatre or, respectively, railway networks as described by Thompson (1967). This model was the 1990s answer to Ronald Coase (1937) with his discussion about ‘The Nature of the Firm’ [quote]:

[It] is necessary to understand why a firm exist at all, since this gives us a clue as to the direction in which to look in order to uncover what determines what a firm does.

With the rise of digital business platforms, this question has to be discussed again. Therefore the research question of this article is how digitalization (‘data as new oil’) can be included into a framework of internal value creation in a firm and how the application of leading-edge digitalization—especially artificial intelligence (AI)—could be extrapolated to a future value creation (which will be called ‘AI-enabled data platform’ later in this article).

Research Approach

The starting point for this article is the model of Stabell and Fjeldstad for value creation in the internal processes of a firm, which inherits Porter’s perspective of IT (and data) as a secondary function only. However, the development of the 2000s pointed out the importance of IT, and soon after Stabell and Fjeldstad (1998), Cusumano and Gawer (2002) and Evans (2003) introduced the model of ‘platforms’, taking up already existing work (see Baldwin & Woodard, 2009, for the history of the platform concept). Their examples consisted of product platforms such as Intel’s chips (Gawer & Cusumano, 2013) and Microsoft’s operating systems, both connecting producers of software packages and software users. Similar cases are video/CD/DVD recorders and game consoles. With the rise of Google, Facebook, Amazon, Alibaba, Tencent, Uber, AirBnB and so on, the tangible base vanished, as all those new business platforms do not ‘produce’ in a traditional sense but provide a digital hub in two-sided markets with (external) data as core resource.

OPEN IN VIEWER Figure 1.

Research approach as discusses in the text (1) with following derivative approaches concerning ‘external’ value creation such as hyper competition/disruption, cooperation in nets, global value chain and so on; (2) value chains with ‘long-linked’ technologies such as assembly lines in manufacturing; value shops with ‘intensive’ technology such as an emergency room in a hospital; and value networks with ‘mediating’ technologies such as railway networks (description according to Thompson, 1967); (3) value chains with with production planning, value shops with data- and document-centric systems, value networks with Capacity & Yield Management; (4) for example, causal inference or process (pattern) mining.

Source: The Author.

The concept of ‘multi-sided platforms’ (see Hagiu & Wright, 2011), of ‘platform-mediated markets’ (see Eisenmann, Parker, & Van Alstyne, 2011) or, respectively, of ‘two-sided markets’ (see Rochet & Tirole, 2003, 2006; Weyl, 2010) raised a number of questions concerning pricing, competition, and market failure. Therefore, research concentrated on the market power of platform due to network effects (see Katz & Shapiro, 1994; Liebowitz & Margolis, 1994), but the question of value creation by business platforms was not in the focus.

The research approach in this article starts with a review of the chronological development of value creation in the 2000s, will derive a synthesis of value creation (within a firm) and ‘digital’ information technologies towards a comprehensive framework for value creation and will extrapolate the proposed framework to the future (refer to Figure 1).

Hence, ‘AI-enabled data platforms’ will be introduced as a new type of value creation, which will be based on the usage of ‘external’ data in contrast to the traditional perspective of value creation in a firm supported by internal data.

The framework provides a synthesis of the three ‘industrial’ and two ‘digital’ models and can be aligned along an axis starting with the ‘zero-dimensional’ value shop via one-dimensional chains and two-dimensional networks to two-sided platforms and beyond to multi-sided ones. Along this framework, key features of those models of value creation are challenges, core competencies, basis for pricing and resources to be controlled (see Figure 2). The discussion will focus on the development of IT from support function to core production factor and on the usage of ‘internal’ and ‘external’ data.

OPEN IN VIEWER

Figure 3.

Source: The Author.

A Literature Review of the Model of ‘Shops, Chains and Networks’

The value chain was successful to describe value creation in discrete (linear) manufacturing industry, whereas already features of (highly interconnected) process industry require modifications to the original model. Even other types of industries such as banking can be analysed with a value chain model, if one defines the ‘production’ as a transfer of risk, maturity or scale from an input (e.g., a large long-term loan) to an output (e.g. many small short-term savings). Nevertheless, the value chain falls short to describe hotels, airlines or other service industries.

Stabell and Fjeldstad (1998) proposed an extension to alternative types of value generation: value shops, value chains and value networks (see Figure 2). In a ’value shop’ (in the sense of a workshop), different specialists work together to provide answers to wishes or problems and expertise and reputation are the key resources for a value shop. In the traditional ’value chain’, efficiency and quality of production and logistics are key along the transformation from procurement via production to sales. In a ‘value network’, an infrastructure of connections has to be operated and (planned) capacity and (actual) flow have to be aligned. It is important to note that the ‘networks’ described by Stabell and Fjeldstad represent internal value creation within one firm—by providing a network such a telecom, a railway or an airline network. They have to be distinguished explicitly from external networking of many companies within global supply chain and/or ecosystems (as discussed in the literature, e.g., by Keane, 2014; Normann & Ramirez, 1993; Vargo et al., 2017).

All three models are based on traditional production factors such as buildings, machines or infrastructure like a telco network. Those three models of value creation also have different basis for pricing. A product from a value chain will be sold at a market value according to demand and supply (and with all well-known pricing approaches from simple rebates to prices adjusted to the price sensitivity of the consumers). As a value shop—like a hotel or a hospital—answers to wishes of guests or problems of patients and the subjective ‘value’ for guests or patients is key for pricing.

These models differ in the way they applied IT to control the key resources. Twenty years after the work of Stabell and Fjeldstad, the role of IT has developed from a pure support function (e.g., in the sense of enterprise resource planning [ERP]) to an essential and integrated part of the ‘production’ itself. In the value shop, the ‘value’ for the client is based on the connection of different types of data, records, documents and so on: coming back to the Paris five-star hotel, a client expects that they know his or her preferences and wishes, loyalty program and address data. In a modern hospital, all different data inputs from images of the magnetic resonance scanner to diet requirements have to be stored in a form to offer access to all with a need to know, but with high data security of the personal and medical data. In the value chain, supply chain management and production planning systems are key to achieve economies of scale in modern complex production processes (from master data management via just-in-time/just-in-sequence to build to order). In the value network, IT systems solve the capacity optimization problems on different time scales (from long-term route planning via seat capacity management to staff scheduling). Therefore, IT is added to the description of the original model as shown in Figure 2.

Marketplaces and Platforms as ‘Digital’ Types of Value Creation

With ongoing digitalization, ‘business platforms’ have been discussed for two decades with different examples (see overview in Kim, 2015), such as:

hardware such as videogame systems, for example, Sony PlayStation, Microsoft Xbox, Nintendo (e.g. Eisenmann, Parker, & Van Alstyne, 2011; Schmalensee & Evans, 2007)

Those platforms have in common that there is an oligopoly of few providers, which connect participants from two sides of a market (e.g., a certain videogame system connecting videogame producers and videogame players). A second common feature is the second-order network effect: different from peer networks, for example, telco networks with the typical relation of:

Value V ∼ n^*(n-1), where n is the number of network nodes,

platforms follow a two-sided relation with a same-side part and a cross-side-part:

V ∼ n^*(n-1) ^* m^*(m-1) / (n+m)² with n and m nodes on both sides of the network.

In 2017, the top 10 companies by market capitaliazation in the world were Apple (>$900bn), Amazon, Alphabet/Google, Microsoft (all >$700bn), Facebook, Alibaba, Berkshire Hathaway, Tencent (around $500bn), before JPMorgan Chase and ExxonMobil (both <$400bn; all numbers from Touryalai, Stoller, & Murphy, 2018). At the Jackson Hole 2018 symposium ‘Changing Market Structures and Implications for Monetary Policy’, Van Reenena (2018) called those platform companies ‘Superstar Firms’ (borrowing from Rosen, 1981). Currently the global oligopoly consist of Google, Apple, Facebook, Amazon (‘GAFA’) and Baidu, Alibaba, Tencent/WeChat, (‘BAT’) plus more regional or local business platforms.

Development of a Framework for Value Creation Including ‘Platformification’

This article proposes that business platforms can be classified in a two-dimensional matrix (Figure 3). The two examples, videogame systems and Uber-type services are both characterized by direct sales of products (videogame hardware) or services (taxi rides) to consumers plus some on-boarding of the second side of the market (game software or drivers with own cars). The second dimension is the difference between platforms with anonymous customers and those platforms with registered customers (i.e., personal data from addresses and bank or credit card credentials to the whole transaction history, e.g., with all time and geolocation tracking of the rides).

OPEN IN VIEWER

Figure 2.

Source: The Author.

This classification helps to define business platforms in sensu stricto as a combination of a marketplace with the identification of registered customers. A company such as Facebook on-boards users of the social media services on the one side and companies advertizing for their products by auctions of individualized advertisement space. As Minor, Hossain, and Morgan (2011) figured out in an experimental analysis, a competition of such business platforms shows a general tendency to develop into monopolistic structures (‘winner takes it all’; Noe & Parker, 2005; Thiel, 2014) or, respectively, into oligopolies, if the platforms address different client groups (e.g., diverse regions, communities, special interest groups, etc.).

The proposed framework for value creation can be extended to business platforms (refer to Figure 2), which rest upon information asymmetry and a ‘reversed’ value stream from the consumer via the platform to companies. The consumers, for example, as users of social media or search engines ‘produce’ digital traces of data as a new ‘raw material’. The business platform aggregates all digital traces into a digital representation of consumer behaviour and sell tailored advertisement spaces to companies (typically in real-time auctions on a best match basis as the customers’ transactions occur). The challenge of this model of value creation is the control of both sides of the market, that is, to ‘lock-in’ consumers in first order and advertisers in second order. The core technology for this model of value generation are engines for matching and recommendation, as well as for predictive analysis, which extract correlations about user behaviour from the aggregated collection of all data and make statistical forecasts for the next action (click, selection, search, desire, etc.). Those engines are able to provide prescriptive analysis, that is, to navigate the client to the ‘next best actions’. Apple’s Siri, Amazon’s Alexa, Microsoft’s Cortana, or Google’s digital assistant are going to strengthen this situation. This technology—extended to the extreme front-end of the human–machine communication—has the potential to develop as a gatekeeper for all ‘digital traces’ and all following steps in the (reversed) value creation of digital business platforms.

Extrapolation towards AI-Enabled Data Platforms with a ‘Map of the World’

Taking into account the current development in the field of AI, one can extrapolate the current business platforms to future ‘AI-enabled Data Platforms’. This article proposes the new concept of ‘AI-enabled Data Platforms’ as hubs with the aggregation of tremendous amounts of transactional data from different sources and extraction of ‘internal’ information about value creation (of other players) based on available ‘external’ data.

Business platforms typically apply AI for matching, recommendation and/or prediction engines, which are all based on statistical (or probabilistic) methods to analyse correlations in customer behaviour to derive patterns for a given customer situation (such as proposal for amended orders, upgrades or additional usage of the platform). It would be beyond the scope of this article to elaborate on the different approaches of AI, and the reader is referred to the (vast and growing) literature for an overview on AI, such as Ertel (2017). The field of AI inter alia includes knowledge representation, cognitive modelling, agent-based modelling, pattern recognition/perception, natural language processing, robotic incl. e.g. autonomous vehicles, generative query networks, genetic algorithms, machine reasoning, and different tools for machine learning (ML). ML—only a subset of AI—covers a large spectrum from traditional methods such as decision trees/random forests via Artificial Neutral Networks (ANN) to leading-edge ‘deep learning’ based on advanced ANN (see, e.g., Milkau & Bott, 2018). Although AI is much more than just ML (see, e.g., Shalev-Shwartz & Ben-David, 2014), the use of AI by current business platform can be simplified to pattern recognition derived from statistical correlations in the gathered data and estimation of average behaviour.

This pure probabilistic approach was already enhanced by Google with concurrent ‘experiments’ by changing, for example, the entry screens (as an input) and measuring customers’ reaction (as an output; Varian, 2010) and Google’s Knowledge Graph (Madrigal, 2012), as well as by Facebook’s Social Graph to map the relationships among users. The Social Graph has been referred to as ‘the global map of everybody’ (Ugander, Karrer, Backstrom, & Marlow, 2011). Both graph-based approaches substantiated the vision of Tim Berners-Lee about a ‘semantic web’ as an evolution from the original World Wide Web (largely documents for humans to read; see Berners-Lee, 1989) to a web of ‘knowledge representation’ (Shadbolt, Hall, & Berners-Lee, 2006)—although not by semantics but by correlation in tremendous amounts of aggregated data. Nevertheless, those platforms provide actual value for customers, as a trivial example of searching for an installation manual may illustrate. The more products are sold with ‘customer self-serve’ only, the more people use search platforms and social media to ask for support, helpful hints and tricks or even explanation videos. The success of the ‘value chain’ model to support internal optimization of (back-end) processes had the collateral effect that also ‘customer journeys’ were defined ex-ante as standardized and scalable processes. This gap of ‘actual value for the customer’ set the stage for the rise of platformification. Advanced AI concepts go beyond the statistical approaches. Either ‘causal inference’ can be applied to derive structural models of the underlying processes (typically by the use of directed acyclic graphs and intervention, see especially Pearl, 2009), or ‘process mining’ can be used based on pattern recognition (within directed transaction data, see, e.g., Liesaputra, Yongchareon, & Chaisiri, 2015) to reconstruct ‘hidden’ internal processes from ‘externally’ available data sets. Both approaches are starting points to develop approximation for a ‘map of the world’ (of those segments of the economic world, for which data runs through those platforms; see Figure 2). Although ‘knowledge’ is a hard term to define precisely, ‘AI-enabled data platforms’ come rather near to the implementation of a ‘knowledge platform’. This can be illustrated, for example, by a future self-driving car, which ‘knows’ when your flight is delayed, pick you up at the airport and takes the best route home along a new pizzeria with a discount offer for your favourite pizza. Such a ‘knowledge platform’ has the potential to create multiple value by solving a complex multi-party coordination problem: for the platform due to advertisement fees from the pizzeria, for the tired and hungry customer with a decent meal, for the pizzeria due to more revenue and even for the environment by better overall traffic coordination (simplified and without any discussion about the danger of ‘nudging’).

Of course, those ‘knowledge platforms’ cannot substitute physical production or delivery processes. However, they can utilize multi-dimensional external data (from customers’ behaviour via geolocation data to flow information about logistic networks) to intermediate the traditional ‘silo’ structures of value creation and to provide benefits for all stakeholders. This is a paradigm shift from internal data (within a firm) as key for efficient value creation to leveraging external data by platforms with multiple-stakeholder value.

Usage of Data for Value Creation: Internal Versus External

In the ‘traditional’ models of value creation—shops, chains and networks—optimization of production of goods and services and, consequently, competitive advantage was achieved by the exploration of internal data: internal expert knowledge, continuous monitoring of process parameters, or long-term transaction histories of clients. In this perspective, a firm was a separated entity with internal value creation and weak interfaces to the outside market (typically along supply chain or service networks). In the twenty-first century, business platforms started to leverage the data from transaction flow to create not even competitive advantage but to dominate global value creation if measured by market capitalization (Thomas, 2017). This triggers a split between models of value creation:

firms, which optimize their internal value creation traditionally by economies of scale and offer products or services to customers on the market

For example, many if not most of the activities of the Alphabet holding neither do fit to the core activities nor are an organic extension for horizontal and/or vertical integration. However, those activities can complement and strengthen the supremacy based on the tremendous collection of digital traces in nearly every sector of life:

Waymo as specialist for self-diving cars (including the ability to prescribe a ‘best’ route)

For an ‘AI-enabled data platform’ in the future, an overall data supremacy will be a key for value creation by the integration of different activities (and different sources of data) under one corporate structure.

Discussion

The interplay between IT and models of value creation changed over the last two decades:

From Porter (1985) to Stabell and Fjeldstad (1998) the models of value creation were based on traditional technology used for production from a ‘mechanical’ perspective, and IT was seen as a simple support function.

Digitalization did not lead to a general disintermediation (due to a reduction of transaction costs for searching etc. lowering the boundaries of a firm) but to different types of business platforms. The dominating model of value creation at the beginning of the twenty-first century are business platforms, which achieved a supremacy of information asymmetry, aggregate as much as possible data, and are able to monetarize the digital traces. Business platforms seek for aggregation of external data, that is, they offer services (usually ‘for free’) to collect customer digital traces as ‘raw material’ for an information refinery with a reversed value creation (i.e., consumer ‘pay’ with data).

This ‘reversed’ model of value creation may be one reason that business platforms could soar to the top of global rankings concerning market capitalization and R&D budget. As the supremacy is based on network effects in a two-sided market, an established business platform has a rather strong competitive position: The winner takes it all. The current development of business platforms is characterized by an extension of these new models of value creation to other economic sectors such as entertainment, groceries or financial services (especially Tencent/WeChat and Alibaba/Ant Financial) and boosting of the development of market capitalization (with Tencent and Alibaba doubling their market cap in 2017). Lower-tier business platforms achieved strong positions in one certain industry (such as travel/hotel portals as second-tier platforms), in one market (such as domestic comparison portals like “Check24” in Germany as third-tier platforms), or in a local niche (such as local food delivery platforms connecting consumers to—logistically local—delivery restaurants as fourth-tier platforms).

An extension of the current business platforms will be the creation of ‘AI-enabled data platforms’ with ‘a map of the world’, as they aggregate data from multiple stakeholders including consumers, producers, logistics and financial transactions. This data pool covers multiple layers of transaction flow (from the complete behaviour patterns of consumers via the entire physical production and trade chains to financing and financial risk management). The combination of those data aggregates will be the basis for a ‘map of the world’, which will be derived by the extraction of ‘knowledge’ about internal value creation by artificial intelligence and, especially, by causal inference and process mining based on external data. With this new competitive advantage, such a ‘knowledge platform’ can create a ‘man in the middle’ rent, but also create multi-stakeholder value by solving complex multi-dimensional optimization problems.

From the point of view of this article, the following questions are open issues:

Is the development towards ‘AI-enabled data platforms’ a consequent development driven by digitalization or, respectively, could there be a flashback, for example, due to ‘over-nudging’?

The scientific description about the interplay between IT, business models and value creation —as already realized in the real economy—does still require further research.

Conclusion

Emmanuel Macron, the president of France, said in an interview with Wired in March 2018 that ‘artificial intelligence will disrupt all the different business models and it’s the next disruption to come’ (Thompson, 2018).

Digitalization and especially the emergence of ‘business platforms’ raises new questions about the nature of a ‘digitalized’ firm and how value is created within a firm in the digital age. One century after the end of the dominance of Standard Oil Company, business platforms exploit ‘data as the new oil’ posing a new kind of challenges for all incumbent companies with a traditional production of goods or provision of services.

In this article, an extension of the approach of Stabell and Fjeldstad with the triad of ‘the value chain, the value shop and the value network’ is proposed as a framework to add (current) business platforms and the (future) AI-enabled data platform as ‘digital’ models of value creation.

The capability of leading-edge AI—for example, causal inference and process mining—provides the basis for ‘AI-enabled data platform’ to extract information about ‘internal’ value creation in firms by analysis of ‘externally’ available data aggregated by those platforms. This leads to rather new forms of value creation in the twenty-first century, which are not covered by traditional theories. Likewise, this challenges incumbent firms, which have to decide how to deal with the intermediation of business platforms (between them and the clients) and with the coming ability of AI-enabled data platforms to intermediate internal value creation based on the data running through those platforms. This article calls for more research to clarify the impact of such data-based business models compared to production-based models for value creation in a firm.

Acknowledgements

The author wants to thank Wolfgang König (Goethe University Frankfurt), Ritva Tikkanen, (Justus Liebig University Gießen) and two anonymous referees for valuable comments and suggestions which improved this article substantially.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: The views expressed in this article are those of the author and not necessarily those of the organizations mentioned.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.