In discipline after discipline and industry after industry, digital networks and AI are becoming pervasive, defining a new age for business and for all of us. AI is becoming the new operational foundation of business – the core of a company’s operating model, defining how the company drives the execution of tasks. In a digital operating model, humans may have designed the operational systems, but computers are actually doing the work in real time: painting the digital Rembrandt, setting a price on Amazon, recommending a product on Walmart’s mobile app, qualifying a customer for an Ant Financial loan – all processes that would traditionally have required human intelligence, not only to design but also to execute. When we talk about strong AI, that can simulate human reasoning or weak AI, that can only execute task, weak AI is already enough to transform the nature of firms and how they operate. In this new age of AI, many time-honored assumptions about strategy and leadership no longer apply.
A digital representation is infinitely scalable – it is now possible to easily and perfectly communicate the pattern it represents, replicate it and transmit it at virtually zero marginal cost to a near infinite numbers of recipients, anywhere in the world. Moreover, digitizing the activity makes it easily connectable, also at zero marginal cos, to limitless other, complementary activities, dramatically increasing its scope. Finally, the digital activity can embed processing instructions – AI algorithms that shape behavior and enable a variety of possible paths and responses. Traditionally, the intrinsic scalability, scope amplification and learning potential of technology was limited by the operating architecture of the organizations that it was deployed in. Increasing levels of digitization, analytics and AI/ML can dramatically improve the scalability of a business, making the value curve increase more rapidly as a function of the number of users or their engagement.
As the firms are getting bigger, their operating models are becoming increasingly complex and with complexity come all kinds of problems. Ultimately, complexity becomes the downfall of traditional organizations, increasing operational costs and decreasing service levels.
In retail sector traditionally people in stores are doing suggestions, but for Amazon, that is done by algorithm. At Amazon people are secondary in many of the most critical workflows. From demand forecasting to warehouse management and from supply chain management to capacity planning, software and AI are increasingly running the show. Over and over, Amazona has collided with traditional industrial settings and transforms them with a digitized, automated and increasingly AI-capable alternative. In order to fight with Amazon, Walmart is rearchitecting its operational model on a digital and AI-enabled foundation. It has acquired digital companies like Jet.com and went into partnership with Microsoft to both drive digital transformation and to access cloud capacity, technology and AI capabilities on demand. Trying to improve in-store process, they are also trying to apply the digital capabilities that are now routinely offered in online world for offline too. In China JD.com has leveraged a less aggressive digital operating model to roll out thousands of convenience stores each week.
We can also see digital transformation with street performers, using WeChat (or Alipay) app. Tencent, owner of WeChat and Ant Financial, company behind Alipay are now widely used payment services and even getting into market fund business.
If you are leading a digital organization, you need to appreciate its full potential, along with both opportunities and challenges. If you are leading a traditional organization, you need to understand how to leverage your existing strengths in new ways and transform your operating capabilities to support the new strategies.
Disruption theory defined an existential threat for traditional firms as they confronted waves of technological change. But today new breed of firms, characterized by digital scale, scope and learning is eclipsing traditional managerial methods and constraints. Software, analytics and AI are reshaping the operational backbone of the firm. This transformation is about more than technology; it is about the need to become a different company.
Ant Financial has become the most valuable unicorn in the world with USD 150 billion valuation in 2018. It has more than 700 million users and more than 10 million SMB enterprises. At heart of its success is its ability to leverage data to learn about its users’ need and respond with digital services to address them. The data is assembled into a powerful, integrated platform that uses AI to power such functions as application processing, fraud detection, credit scoring and loan qualification.
The value of the firm is shaped by two concepts:
- Firm’s business model – defined as the way the firm promises to create and capture value.
- Firm’s operating model – defined as the way the firm delivers the value to its customers.
Business model is about strategy, operating model about systems, processes and capabilities. First is about theory, second about practice.
Company must create value for a customer that prompts her to consume the company’s product or service. Company must deploy some method to capture some of the value created. The reason customers choose to use a company’s product or services is called value creation and it has to do with the particular problem the company is solving for customer. This is sometimes also called value proposition or customer promise. On the other hand, company should capture some of this value, preferably less than the value it creates for the customer.
The new breed of digital firms is experimenting and recombining various aspects of value creation and value capture. In a fully digitized business, the options are much broader, because value creation and capture can be separated much more easily and often come from different stakeholders. For the digital firms, underlying all this business model innovation is a very different kind of operating model.
Operational models deliver the value promised to customers. Whereas the business model creates a goal for value creation and capture, the operating model is the plan to get it done. The goal of an operating model is to deliver value at scale, to achieve sufficient scope and to respond to changes by engaging in sufficient learning. So, we can say we have three operating challenges:
- Scale – deliver as much value to as many customers as possible at the lowest price.
- Scope – the range of activities it performs, the variety of products and services it offers its customers. Doing that with centralized R&D organization, warehouse. Investing into brand.
- Learning – the learning function of an operating model is essential to driving continuous improvement, increasing operating performance over time, and developing new products and services.
So, we can structure company model:
- Business model
- Value creation
- Value capture
- Value creation
- Operating model
- Continues improvement
- IP generation
By deploying a fundamentally new kind of operating model, this new type of firm is reaching new levels of scalability, achieving a vastly broader scope and learning and adapting at a much faster rate that does a traditional firm. This is because the digital firm is transforming the critical path in the delivery of value.
Many companies at the dawn at the internet commerce worked hard to solve the trust problem. For Alibaba the solution was to rely on an escrow system, wherein a third party holds payment until a contractual agreement is fulfilled. The solution was to rely on an escrow system, wherein a third party holds payment until contractual agreement is fulfilled. Value creation is related here with offering a substitute for trust. Value capture occurs through the 0.6 percent transaction fee. Alipay needs to increase both the intensive margin of transaction (more transaction per user) and extensive margin (more users). The network effect amplifies the value created by trust in the services. Alibaba spun out Alipay but they still retained rights to collect 37.5 % of its pretax profits.
Ant Financial extended its financial ecosystem with Yu’e Bao, an investment platform that allows Alipay users to earn interest on money in their accounts. They can transfer pocket money from their accounts into market funds and get a 4 percent annual return. Another expansion of Ant Financial are Ant Fortune, Zhima Credit – a social scoring system and Mybank – an internet banking services provider. By 2019 Ant Financial controlled 54 percent of the mobile payments market in China, while Tencent’s WeChat controlled 38 %. Ant Financial also expand in India and South Korea.
Alipay rapidly expanding business model is built on a new kind of operating model. It is based on broad reliance on AI-enabled digital automation. Mybank’s hallmark is 3-1-0 system. 3 minutes to apply, 1 second to approve and 0 human interaction. Alipay uses data and AI to ensure trust. Data are coming from: internal consumer behavior statistics (records of relocation trends, utility bills, money transfers, wealth management, purchasing patterns from Alibaba), transaction data from sellers on Alibaba platforms, publica data such as government databases containing criminal records, citizen identification information and academic profiles, data from Ant Financial’s partners to power Zhima credit scores.
The essence of the digital operating model is avoiding direct human intervention on the critical path of the product or service-delivery process. While employees help define strategies, design user interfaces, develop algorithms, code software and interpret data, the actual processes that drive customer value are fully digitized. With this improvement the marginal cost of serving an additional user on many digital networks is, for all purposes, zero, apart from the small incremental cost of computing capacity, which is easily available from cloud service providers. When fully digitized, a process can easily be plugged in to an external network of partners and providers.
Peloton is a fitness company where you pay 2.200 USD for bike and 39 USD for monthly subscription. Peloton’s initial value creation is to bring the fitness studio to the customer’s home. Value capture model is combination of product sales and subscription. Its organization is structured more like a software company. It is now 700 million revenue company with 4 billion valuation.
Ocado delivers groceries both for its own branded online and mobile services and for variety of third parties. Its operational model is built on data, AI and robotics. It is an AI company disguised as a supply chain company disguised as an online grocer. AI is used not only for routing, but also for predicting when customers are likely to order the products in the first place. Using robots makes their warehouse logistic efficient.
Digital model can be explained with:
- Business model
- Value creation – exceptional increasing consumer value by personalization and engagement
- Value capture and sharing – NEM (number of users * user engagement * monetization) across many related markets (multisided).
- Operation model (value delivery)
- Scale – zero marginal cost network effects
- Scope – aggregation and modularity across markets
- Learning – constant innovation and AI/ML driven improvements
In 2017 Sundar Pichai, Google’s CEO moved its strategy from mobile first to AI first.
Industrializing data gathering, analytics and decision making to reinvent the core of the modern firm in what we call AI factory. The AI factory is scalable decision engine that powers the digital operating model of the twenty-first-century firm. Analytics systematically convert internal and external data into predictions, insights and choices, which in turn guide or even automate a variety of operational choices.
Netflix uses AI to improve user experience. Their model really took of with streaming potential. In 2014 they improved the streaming experience with personalization based on such factors as connection speed and preferred device and determining what movies and shows to cache on edge servers which are deployed closer to viewers.
AI factory components are:
- Data pipeline
- Algorithm development
- Experimental platform
- Software infrastructure
If the data is the fuel that powers the AI factory, the infrastructure makes up pipes the deliver the fuel and the algorithms are the machines that do the work, then experimental platform controls the valves that connects new fuel, pipes and machines to existing operational systems.
Netflix is using data inputs like item ratings, item popularity, plays like duration, time of day and device, queues, metadata about context, presentations, social data, search terms and even some external data. Company is using data to create microclusters or taste communities.
When setting up AI factory it is important to check if data are available and if not, it may be worth that company invest into infrastructure to create it. Executives at many incumbent companies consistently underestimate the challenge and the urgency of the investment they face in cleaning and integrating their data across the enterprise so that they can build an effective AI factory. The tool that makes the data useful is the algorithm. Algorithms have been around for quite some time. The conceptual and mathematical development of classic statistical models such as linear regression, clustering or Markov chains data back more than a hundred years.
Machine learning is about supervised learning, unsupervised learning and reinforcement learning. Regarding unsupervised learning we have three broad types: cluster data into groups, association rule mining and anomaly detection. In reinforcement learning the general idea is to maximize operating performance by minimizing regret.
Google runs more than one hundred thousand experiments each year to test a vast variety of potential data-driven improvements to its services. LinkedIn reportedly runs more than fort thousand experiments. A state-of-the-art experimentation platform will provide the comprehensive set of technologies, tools and methods required to do experimentation at scale. In order to run experimentation platform, we need hypothesis, that we test in randomized control trials (A/B test), with random samples exposed to change (treatment) and second sample doing thigs as usual (control). This approach ensures that any prediction being generated by algorithms actually has a causal effect of outcome.
After the data is aggregated, cleaned, refined and processed, it is made available through consistent interfaces (the APIs). Concurrent with investment in data and software are strategic investments in connectivity and infrastructure to integrate with data platforms. The design of interfaces is critical in ensuring modularity in both code and organization. If there is a standard for sharing data and functionality, we can develop different modules inside company independently. This is also important if we involve external partners. But we should be careful with data governance and security. As scandal with The Cambridge Analytica showed, with data flow through API, companies are very exposed if they don’t do it correctly.
Changing a Company
In 2002 Jeff Bezos send a mail, warning whole company about need to integrate data and functionality on a level of whole company. Digital firms should rest on integrated, highly modular digital foundation. Obidos, Gurupa and Santana are the systems Amazon built to enable its operating capability and to meet its scale, scope and learning objective. Bezos rebuilt Amazon’s retail operation on top of a software platform, which gradually evolved to embedded a state-of-the-art AI factory. Company brought Brian Valentine from Microsoft to build robust software platform. This development lead to development of AWS as the most important part of Amazon business.
Mirroring hypothesis states that organizational ties within a project, firm or group of firms…will correspond to the technical patterns of dependency in the work being performed.
While making similar tasks increasingly efficient over time these patterns can also constraint an organization, building inertia that hampers the response to change. The concept of architectural inertia – the resistance to adaption – in turn informs Clayton Christensen’s disruption theory.
Highly specialized, siloed operating models remain essential in manufacturing and service delivery today. IT system usually mirrored firm’s siloed and specialized architecture. Large firms often use thousands of enterprise applications and IT systems, working with a variety of scattered databases and supporting diverse data models and structures. Traditional operating architecture created serious constraints to firm growth and value. Firms are shaped and limited by their operating models.
Today it is all about building an organization founded on code instead of human labor. As long as digital systems use a well-designed, common interfaces they can connect and combine capabilities, dramatically enriching the range of possibilities. The ideal is to have a common foundation of data inputs, software technology and algorithms all provided by an AI factory. Modern operating models are also characterized by relentless focus on improving performance through learning. While the data is centralized, the company’s experimentation capability is highly decentralized.
Finally, new breed of organization transforms the role of management. Management as supervision, especially of employees performing routine tasks, is finally over. Managers are designers, shaping, improving and controlling the digital systems that sense customer needs and respond by delivering value. They are also innovators and integrators and guardians that work to preserve the quality, reliability, security and responsibility of the digital systems.
Nadella moved Microsoft into cloud. He bet on it, as future of Microsoft. And in 2018 he improved this strategy with making claims about intelligent cloud and intelligent edge. For traditional firms, becoming a software based, AI-driven company is about becoming a different kind of organization – one accustomed to ongoing transformation. It is about fundamentally changing the core of the company by building a data-centric operating architecture supported by an agile organization that enables ongoing change.
In 2014 Nadella decided that it was time for Microsoft to embrace open source. With acquisition of GitHub in 2018, this strategy becomes even more obvious. For Microsoft cloud and AI is plan A, there is no plan B. They are putting 5-6 million of yearly CAPEX into it. Cloud business require massive investments in infrastructure. Microsoft had to deploy an efficient and responsive supply chain, one good enough to compete with Amazon, possibly the best supply chain in the world. Because cloud consumption of a Microsoft product can increase only if the product is actually used, consumer alignment is imperative. Once you are in the consumption business, you are part of your customer’s operations. The responsibility is super real. Cloud provider cannot let their systems to go down.
Microsoft brought Azure into core of their products, redesign it, to make it easier to use and more compatible with traditional Microsoft products. They key of changing Microsoft’s core was the transformation of Microsoft’s own installed base of customers.
The good news in having a cloud business is getting constant feedback from usage that highlights problems and motivated improvements. The bad news is that the engineering organization must react to this in real time, or as close to it as possible.
On top of introducing AI and ML layer into their services and products, Microsoft also needed to transform their organization. Developing CoreServices at the center of Microsoft’s transformation, working to rebuild traditional silos on a common digital foundation. This operating foundation connects the enormous organization to a common software component library, algorithm repository and data catalog, which can be used to rapidly digitize, enable and deploy digital processes across the entire company. DelBene who runs this organization said, that their product is the process.
In AI/ML related field, they define six AI principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability.
The first essential principle in transformation is to develop strategic clarity and commitment. You need to bring unity into company while transforming it. Clear and compelling vision is necessary. Coordination is important. In a lot of transformation, it is interesting to see how CIO and IT organizations can show the biggest resistance against it. So, it is important to develop agile product-focused organization. Agile methods go hand in hand with a transformed, data-centric operating architecture. But transformation must include also culture shift. And building capabilities for new model inside the company and partnering outside. This may signal the emergence of a new generation of business leader.
Digital governance should involve a collaboration across disparate disciplines and functions. And should include ecosystem of partners and customers as well as communities surrounding them.
In a company of one of the authors Keystone Strategy, they have made research about correlation of new model capabilities and business performance. And realize that correlation between them exist, better the model, better the performance.
There appears to be a natural sequence of stages in the journey to become a state-of-the-art AI factory: from siloed data, to pilots, to data hubs to AI factory.
Fidelity launch their project of building AI Center of Excellence in 2011. Their data platform tracked and integrated over 36 million user profiles, interactions and digitized voice calls.
Strategy for a New Age
When Albert-Laszlo Barabasi and colleagues were analyzing WWW, they realize that some nodes in network are more-connected – preferential attachment. Some firms in digital connection (keystones, platform firms, superstar firms or hub firms) would emerge as much more connected and powerful than others. As firms develop increasingly digital foundations, the architecture of the economy is being reconfigured into a huge all-encompassing, AI-powered network consisting of an array of subnetworks – social networks, supply chain networks and mobile app networks. These networks have at least five things in common: they are made up of digital connections, they carry data, they are shaped by algorithms, they ignore traditional industry boundaries and they are becoming more and more important for economy and social system. In order to understand new breed of companies, network analysis is crucial. Network analysis involves understanding the open and distributed connections across firms. Understanding network and learning effects that are accumulated by those firms. Network effects describe the value added by increasing the number of connections within and across networks. Learning effects capture the value added by increasing the amount of data flowing through the same networks.
The intrinsic value created on one network may be realized (and captured) across any of a multitude of other networks the business can now easily connect to. The most important value creation dynamic of a digital operating model is its network effects. The basic definition of a network effect is that the underlying value or utility of a product or service increase as the number of users utilizing the service increase. Networks intrinsic utility increases ad it adds users. There are two main types of network effects: direct and indirect. Companies have also learned that the presence of one type of network effects (direct or indirect) can be leveraged to generate the other type.
Learning effects can either add value to existing network effects or generate value in their own right. The bigger the user base, the greater the scale, the more data is available and the greater the value.
The structure of the network also has an important impact on how a network’s value increase with its size. Network clustering has big effect on sustainability of network-based business model. Global networks are more concentrated around a small number of critical hubs. Barriers to competition are high and sustaining profitability is relatively easy for the dominant player. Uber’s network is highly clustered grouped around individual urban locations. Clustered networks are thus typically highly competitive. Any competitor with local scale can achieve similar efficiencies.
Because of spread of digital networks options for value captured grown dramatically. Digital value capture technologies allow for careful usage metering, sophisticated pricing algorithms that react to product inventory conditions and even outcome-based pricing models.
Multihoming refers to the viability of competitive alternatives specifically to situation wherein users or service providers in a network can form ties with multiple platforms or hub firms (homes) at the same time. When multihoming is common on each side of a platform, it becomes almost impossible for the platform to generate a profit from its business. Firms are trying to avoid multihoming with offering complementary services to partners.
Disintermediation, wherein nodes in a network can easily bypass the firm to connect directly, can also be a significant problem for value capturing. This problem is a frequent one, especially for marketplaces. The most effective way to reduce disintermediation is to reduce transaction fees and make up the revenue on different markets sides.
Although multihoming and disintermediation are the enemies of network-based profitability, network bridging involves making new connections across previously separate economic networks making use of more-favorable competitive dynamics and different willingness to pay. Google bridged the search business with a network of advertisers. This is also main reasons why Amazon and Alibaba and other hubs are moving into many markets.
Strategic network analysis is done in those steps: first is to list the major networks a business is connected to, second it to evaluate the potential of each major network in the business for value creation and capture at scale.
Uber business model depends on their ability of bridging, since their main network is not profitable. Businesses like Uber Walmart, Uber Eats, Uber Health or Uber Pool and Cargo, offers potential for Uber, but all of them have some challenges.
Some of the questions everybody should ask before moving into network business are:
- What is the core service delivered?
- What networks are key to providing that service and that what are their characteristics? Do they have strong learning or network effects? Are they clustered?
- If network and learning effects are weak, how do you strengthen them over time? How do you increase the value delivered?
- If the network effects are strong and there is very little value delivered until critical mass, how do we get there?
- What are the most important secondary networks? Can they enable additional network or learning effects?
- Do we have challenges with network clustering? Multihoming? Disintermediation?
- What are the best value capture opportunities?
- Are there network bridging opportunities? Considering the data, you can accumulate from your core network, is it of value to another network?
A collision occurs when a firm with a digital operating model targets an application (or use case) that has traditionally been served by a more conventional firm. Digital firms may take some time to create value generated in the levels of traditional firms, but once they reach critical mass, value delivered can be truly impressive.
Hotel industry is on a route to collision. Although hotel companies will not cease to exist, their profits are moving to the operating system layer.
Nokia also lost a battle with new business models, when Android and iOS app networks took over. They could either compete with development of their digital operating model or work on their products as complement to the new software-based entrants. They didn’t do anything and so they went out of the market.
We can see collisions in computing business, retail, entertainment, automotive industry.
Beyond influencing individual markets, hub firms are poised to create and control essential connections in key networks. On the other hand, spread of digital operating model is causing a range of new threats.
The Ethics of Digital Scale, scope and learning
Challenges can be grouped into five main categories: digital amplification, bias, security, control and inequality.
Echo chamber effects is one of ways digital amplification is shown. Users are looking only at the context that supports their beliefs.
Two common types of algorithmic bias are: selection bias and labeling bias.
Cybersecurity is becoming main danger in digitally connected world. We are seeing more and more breaches. One of them was a company Equifax. The problem was not so much a breach as behavior of management and company after the breach. Not reporting it, management selling stocks during that time.
Platform control is about hubs controlling their ecosystem and taking responsibility for context and behavior. Think about Cambridge Analytica scandal. It is about drawing a line between offering relevance and violating privacy.
As digital operating models drive network and learning effects, the asymmetric across organizations will tend to grow and markets will become more concentrated.
We need proper regulatory framework and new generation of leaders to acknowledge new responsibilities and work proactively to solve the new challenges. A keystone strategy aligns the objectives of a hub firm with those of its networks. By improving the health of its networks (or business ecosystem), a keystone strategy also benefits the long-term performance of the firm. The central feature of this strategy is its focus on aligning internal and external needs to shape and sustain the health of the networks a firm depends on. The keystone concept is related to the ides of information fiduciary proposed by Jack Balkin and Johnathan Zittrain.
The age of AI is changing the game for all of us. A new meta is like changing the moves allowed on a chessboard or the rules of bridge halfway through a game. As we enter the age of AI, we should pay careful attention to these emerging principles:
- Change is no longer localized, it is systemic – digital technology is an engine of systemwide transformation, digital transformation cuts across every industrial environment at the same time.
- Capabilities are increasingly horizontal and universal – emphasis on primary differentiation on the basis of cost, quality and brand equity is shifting from specialized, vertical expertise to the firm’s position in the network, its accumulation of differentiated data and its deployment of a new generation of analytics.
- Traditional industry boundaries are disappearing, recombination is now the rule.
- From constrained operations to frictionless impact – frictionless systems are prone to instability and have difficulty finding equilibrium. Once in a motion, frictionless systems are hard to stop.
- Concentration and inequality will likely get worse.
The digitization of our economy appears to have moved past an inflection point. And as the digital firm continues to amplify its impact, we are starting to see a marked drop in public trust and cohesion.
In contrast with the current wealth of data, analytics and AI, we still appear to be suffering from a shortage of managerial wisdom. There are four arenas in which this leadership mandate is playing out:
- Transformation – it demands leadership commitment and there is no plan B for digital transformation. Qualification for leaders should start with an understanding of the digital systems they are creating and leading.
- Entrepreneurship – AI is offering one of the biggest entrepreneurial opportunity in the history of civilization. Blockchain is one area where impact is lacking promise.
- Regulation – regulators are racing to catch up with the evolution of technology. Collaboration between public and private is needed.
- Community – communities are an increasingly important complement to regulation in providing checks and balances to digital firms. Think about the power of Wikipedia and open source software.
As digital firms increasingly shape our global economy, their management will be held accountable to a different standard. Managers frequently abandon the common perspective when facing critical decisions. As digital networks and AI increasingly capture our world, we are seeing a fundamental transformation in the nature of firms.