Summary-Competing in the Age of AI - Marco Iansiti

Summary-Competing in the Age of AI - Marco Iansiti


  • Enterprise transformations often fail due to a lack of leadership and commitment. Even large investments may not bring change if the organization is fractured and unaligned.

  • Success requires leaders who can inspire ongoing commitment, build bridges across divisions, understand where alignment won’t happen, and make necessary changes. Leadership at all levels is needed, not just at the top.

  • Digital firm leaders need to understand both the technology and human aspects. They need to understand digital systems, ethics, and organizational impact, and inspire the right culture. Integrated, well-rounded perspectives are key.

  • There are huge entrepreneurial opportunities in the age of AI, from transforming existing processes to solving key challenges like cybersecurity, algorithmic bias, fake news, and job creation. Innovation costs have dropped, enabling more people to invent and scale new ideas.

  • However, new ventures must evaluate not just the technology and operations but the business model and competitive implications. Some models like ride-sharing may increase consumer benefits but struggle to be profitable or provide good jobs. Leaders must consider the impact on communities and society.

  • Blockchain is promising but its impact has been limited. To scale, blockchain business models must fit with complex institutions and norms or help transform them. Successful models will likely involve tailoring the technology to specific needs and innovating the business model.

  • In summary, leadership, commitment, integrated thinking, social responsibility, and business model innovation are all needed to successfully transform organizations and scale new digital ventures in the age of AI. Technology alone is not enough.

  • There are many nuanced factors involved in determining how much value increases with network scale.

  • Traditional businesses typically show diminishing returns to scale, as illustrated by Curve A.

  • Even small network or learning effects can increase value, as shown by Curve B. Stronger effects can lead to increasing returns, as in Curves C and D.

  • The goal is to move up the value curve by strengthening the network and learning effects.

  • Network effects mean a product or service's value increases as more people use it. They can be direct (users value other users, e.g. social media) or indirect (different types of users value each other, e.g. Uber).

  • Indirect network effects can be one-sided (only one group values the other, e.g. advertisers valuing social media users) or two-sided (different groups value each other, e.g. Uber riders and drivers).

  • The more connections in a network, the greater the value. Platforms aim to enable connections between users and capture the resulting value.

  • Digital networks often have a clustered structure, where nodes are grouped geographically or by interest. This structure makes the networks vulnerable to competition from focused competitors targeting specific clusters.

  • The strength and structure of network effects can change over time as the network evolves. For example, the network effects of Microsoft Windows decreased as web-based and mobile applications became popular, leading to the rise of competitors like Android, Chrome, and Mac.

  • There are several ways for firms to capture value in digital networks, including usage-based pricing, outcome-based pricing, and charging different sides of two-sided networks based on their willingness to pay. However, not all the value created can be captured due to factors like multihoming and disintermediation.

  • Multihoming, where users or providers can connect to multiple networks or platforms, limits a firm’s ability to capture value. Firms try to reduce multihoming through approaches like loyalty programs, exclusive contracts, and subsidies.

  • Disintermediation, where nodes bypass the network hub to connect directly, also threatens a firm’s value capture. Firms can try to prevent disintermediation by controlling key resources, using legal restrictions, improving the hub’s value proposition, and technological controls.

  • Ultimately, a firm’s ability to capture value depends on developing a competitive advantage that can overcome the threats of multihoming and disintermediation. Strong network effects, especially those based on data and AI, may provide the most sustainable competitive advantages.

  • Fidelity developed an advanced data and AI platform that tracked 36 million customer profiles and interactions.

  • They built agile teams to develop AI applications to improve customer experience, find new revenue opportunities, and gain business insights.

  • Fidelity is transforming its operating model and business model using AI and data. AI is helping Fidelity improve all areas of its business and deliver a better customer experience.

  • In general, as firms adopt digital operating models, they are confronted with many new strategic options to transform their business models.

  • Evaluating these options requires understanding how firms can connect to and control economic networks, leverage network effects, and benefit from data and learning effects.

  • Digital networks ignore traditional industry boundaries, are powered by software and algorithms, carry data, and some nodes become very connected hubs. These hub firms can gain significant power from data and analytics.

  • Competitive advantage comes from shaping and controlling these networks to harvest the data that flows through them. Network analysis, not just industry analysis, is key to strategic thinking.

  • Firms should analyze the networks they connect to, how those networks enable value creation and capture, and the data flows through the networks. Managing networks and data is becoming central to strategy.

  • While industry analysis focuses on isolated industries, network analysis focuses on the open connections across firms and industries.

  • As firms link to networks, they can gain network and learning effects. Network effects come from more connections; learning effects come from more data to power AI. Generally, more is better.

  • The new strategic challenge is understanding a firm’s place within the broad, interconnected economy and all its data flows, not just focusing on a single industry. Firms must consider how to create and capture value through networks and data.

  • Ocado, Ant Financial, and Peloton are digitizing value delivery and enabling business model innovation in their respective industries.

  • They are creating consumer value at an unprecedented scale and scope. They generate revenue by fostering consumer loyalty and engagement, not just transactions.

  • The companies took different approaches based on their industries but digitized critical operating processes to drive scalability, scope, and learning.

  • Ant Financial built capabilities in analytics and AI to automate financial services. Ocado uses AI and algorithms to drive an automated fulfillment system and partnerships with retailers. Peloton uses networks and community along with analytics to boost engagement.

  • In 2017, Google announced it was transitioning to an "AI first" company, making AI the foundation of its business and operating models. AI would power all its products and services.

  • The "AI factory" is the scalable decision engine at the core of digital companies. It industrializes data, analytics, and decision-making.

  • Managerial decisions are embedded in software, digitizing traditionally human processes. This enables superior scale, scope, and learning.

  • The AI factory converts data into predictions, insights, and choices that guide or automate operations.

The key themes are using AI and algorithms to automate and digitize business processes, building consumer loyalty through engagement, and developing an “AI factory” model with data-driven, scalable decision-making at the core of operations.

After a period of tremendous success, Microsoft struggled with competitive and regulatory pressures in the 2000s. As Bill Gates stepped back, CEO Steve Ballmer struggled to keep Microsoft innovative. The company faced issues in launching new products, failed in some new markets, and lost relevance with developers.

When Satya Nadella became CEO in 2014, he renewed Microsoft’s purpose and strategy. The new mission was to “empower every person and every organization on the planet to achieve more.” The strategy focused on becoming a productivity platform enabled by the cloud, AI, and open-source software.

Implementing the new strategy required overhauling Microsoft’s operating model and architecture. Microsoft had to build a massive cloud infrastructure and supply chain to compete with Amazon. It took years of work, but Microsoft gained operational capabilities and now has a responsive supply chain and systems to manage it.

The cloud model also provides benefits like constant customer feedback and data to improve products. Microsoft has analytics to see how customers use products and make improvements. The cloud means Microsoft is closely integrated with customer operations, so reliability and uptime are critical.

Nadella brought Microsoft Azure into the core of the company early on. Azure was redesigned to work with Microsoft’s existing products and run both Windows and Linux. Microsoft gave customers incentives to move to Azure. Transforming the existing customer base was key to changing Microsoft. Leaders like Bill Laing and Scott Guthrie restructured Azure to be more user-friendly, compatible, and powerful. They broke down silos to transform how Azure was built and delivered.

In summary, Nadella renewed Microsoft's purpose, moved to a cloud and open source strategy, overhauled operations, made Azure central and compatible, gained operational excellence, used data and analytics, focused on customers, and transformed product development. Microsoft went through a difficult transition but emerged as an innovative leader in the cloud. Here's a summary:

  • Equifax, a major credit reporting agency, suffered a massive data breach in 2017 that exposed the personal information of 147 million people. The breach was the result of poor cybersecurity practices and slow response.

  • Data breaches and cyberattacks have become common, impacting many major companies. There is a duty to protect customer data, but doing so is increasingly difficult given the growing reliance on data and digital systems.

  • In addition to traditional cyberattacks, there are new threats like digital hijacking, where online systems and platforms are misused for malicious purposes. For example, the Christchurch mosque shooter live-streamed his attack on Facebook, and Russian groups used social media to spread misinformation and influence elections.

  • Responding to these threats requires cooperation across individuals, companies, and governments. Not all harmful digital incidents are illegal, creating "gray areas" in how platforms control and monitor their ecosystems.

  • Facebook and other platforms aim to ensure their tools are used for "good" but determining how to properly control platforms without impacting free speech is challenging. Facebook's data was misused by Cambridge Analytica to target political ads, showing how platform data can be exploited if not properly controlled.

  • In summary, cyber threats are increasingly sophisticated and impactful, ranging from data breaches to digital hijacking to misuse of platform data. Responding requires shared responsibility, investment in security, rapid response to incidents, and balanced control and monitoring of digital platforms. Failing at any of these areas can have major consequences.

  • Digital platforms like YouTube, Netflix, and streaming services are transforming the entertainment industry.

  • YouTube benefits from huge network effects and essentially dominates its market. Netflix relies more on content from studios and faces more competition.

  • Digital firms now compete with traditional media companies and have huge content budgets. They are providing over-the-top internet-based video services at scale.

  • Traditional media companies are reacting by merging and transitioning to digital. Some like Comcast and Disney have made progress.

  • Cars are becoming more connected and digital, threatening traditional automakers. The opportunity to connect with consumers during commutes is worth billions.

  • Exploiting this requires a digital, data-centric model to provide on-demand services and targeted ads. Car-sharing services are showing the way, but autonomous vehicles are the biggest opportunity.

  • Alphabet (Google) is poised to shape the automotive experience and capture value. Android, Google Maps, and ads are ready. Waymo could earn hundreds of billions.

  • This will transform the auto industry as transportation becomes more about services than ownership. Differentiation will decrease and hardware will commoditize.

  • Auto manufacturers have two options: challenge digital companies or work with them as suppliers. Both are challenging. They are trying to participate in software and services but face issues of scale.

  • Some automakers have invested in car-sharing and autonomous driving. They may need to collaborate at scale to compete, like with the HERE mapping consortium.

  • We see new digital operating models transforming industries. Software, data, and AI enable new business models across industries, changing competition.

  • Some markets are becoming more concentrated and "winner-take-all." Industries are connecting through a digital fabric.

  • A generation of "hub firms" like Google, Amazon, Facebook, Tencent, and Alibaba are occupying central positions, providing value but also capturing much of it.

  • These firms are poised to control key networks and connections in the economy.

  • Advanced analytics and AI provide an opportunity for Walmart to enhance the in-store experience by capturing customer data and patterns, personalizing the shopping experience, and improving store operations. However, implementing these technologies raises privacy concerns and questions about how technology and humans will interact.

  • Amazon Go stores are an example of how technology can transform the retail experience by eliminating cashiers and checkout lines. The scalability of these digital stores poses a threat to traditional retailers.

  • In China, WeChat Pay and other digital payment systems have become ubiquitous, enabling new applications and business models. Tencent, the maker of WeChat, has built a massive digital platform and is using data and AI to expand into new markets, threatening various industries.

  • The emergence of digital technologies like AI is challenging long-held assumptions and transforming the economy. Although the changes can be disconcerting, we have entered a new age where organizations must leverage digital networks and AI to create and capture value.

  • Leaders must understand how to manage, grow, and govern their organizations in this new environment of digital transformation. This book aims to provide that understanding.

  • The Netflix recommendation problem is complicated because they have to figure out which movie to recommend and which visuals (artwork) to pair it with for each user.

  • Like navigating down a mountain, Netflix spends time exploring options and exploiting solutions from their models. They explore by randomizing visuals shown to users. They exploit by using improved models to show personalized recommendations.

  • Netflix improves dynamically by cycling between exploration and exploitation to maximize user engagement. They aim to find the best artwork for each user based on their tastes.

  • The Netflix challenge is a type of reinforcement learning model called the multi-armed bandit problem. It's like a gambler deciding between slot machines with unknown payouts. The goal is to maximize reward while balancing exploration and exploitation.

  • Multi-armed bandit problems are important for optimizing processes. They are used for recommendations, pricing, clinical trials, ad selection, and more. They can guide the behavior of agents in the real world or simulations.

  • Reinforcement learning, used by AlphaGo, involves an agent interacting with an environment to maximize a reward. AlphaGo Zero used reinforcement learning with just the rules of Go to beat AlphaGo. DeepMind applied this to drug discovery and protein folding.

  • To reliably use AI predictions, companies need experimentation platforms to test hypotheses. They run randomized control trials, like A/B tests, to see if predictions actually cause outcomes. Companies like Google and LinkedIn run thousands of experiments per year.

  • Netflix built an extensive experimentation platform integrated with their algorithms. They A/B test all major product changes and use it to improve streaming, content delivery, images, UI, email, playback, registration, and more. They aim for scientific rigor in decision making.

  • Data platforms, algorithms, and experimentation platforms need software, connectivity, and infrastructure. Data flows through a pipeline, cleaned and processed, then made available through APIs for applications to use. This lets teams build and deploy new AI apps rapidly.

  • Enterprises traditionally operate in silos, but data platforms should be modular with consistent interfaces to enable aggregation, insight, analytics, and AI.

  • Digital transformation is accelerating and bypassing traditional organizational constraints. This allows for rapid growth in users, engagement and revenue. However, it also creates new leadership and governance challenges that existing institutions struggle with.

  • Digital networks lead to greater concentration of transactions, data, power and wealth in a few network hubs. This exacerbates inequality across workers and firms. The scale and speed of this trend could lead to greater frustration and anger.

  • Digitization has moved past an inflection point and is reducing public trust and cohesion. While digital innovation provides benefits, relatively unconstrained digital operating models also amplify economic disparities, spread misinformation and open us up to cyber threats.

  • These vulnerabilities require a broad, considered response. Leaders need to consider employees, customers, partners and communities - not just shareholders. Responses could include investments in depressed areas, universal basic income, etc. Leadership decisions shape our collective community, so leaders may be judged by society as a whole, not just investors.

  • We need wisdom and proper consideration to address the social dislocations from the transformation of value creation and delivery. As in the Industrial Revolution, addressing grievances and finding ways for workers to return to work will require fairness and wisdom.

  • The age of AI mandates wiser leadership of increasingly digital firms. Digital tools and AI are redefining firms and managerial tasks. But despite successes, we need to do better at managing new digital assets and capabilities.

  • Transformation starts at the top but is difficult. No organization should stand still. Cloud technologies and expertise are available to all, but transforming organizations and building skills are hard. Leaders must motivate change despite difficulties.

  • Dabbling in transformation and pilots is not enough. Real transformation threatens the status quo, so leaders must be fully committed. Incremental progress is not transformative and sustainable change requires perseverance.

  • Digital companies rely on thousands of experiments using A/B testing and randomized controlled trials to optimize their services, increase customer satisfaction, and boost revenue. While data is centralized, experimentation is decentralized, allowing anyone to test hypotheses.

  • Digital operating models promote reusable and modular software, using frameworks like React and Apache Storm. Competitive advantage comes from data rather than proprietary tech. Development is often open-source.

  • In digital operating models, software and algorithms deliver the product or service, not employees. This removes constraints on scale, scope, and learning. Marginal cost to serve more users is negligible. Complexity is solved through software and outsourcing. These models are infinitely scalable if computing power and data increase.

  • Digital technologies are modular, enabling many business connections. Once digitized, a process can plug into external networks and communities. Scope and value multiply. Learning and network effects lead to increasing returns to scale.

  • Management roles change to designing, improving, and controlling digital systems; envisioning digital system evolution; integrating systems; and ensuring quality, reliability, security, and responsibility.

  • Though the potential of data-centric operating models is huge, many traditional firms hesitate to transform due to a desire to protect existing capabilities, routines, and organizational boundaries. Technology is the easy part; organizational change is hard.

  • Becoming an AI company requires conviction and patience. Microsoft transformed into a cloud company under Satya Nadella, then into an "intelligent cloud and intelligent edge" company. Transformation involved changing both the business and operating models.

  • Principles for transformation include: reimagining your business model; adopting a customer-obsessed growth mindset; building a data-centric operating architecture; using algorithms and AI; and organizing for ongoing change.

  • Research shows the value of transformation in boosting revenue and market cap. The speed and extent of change determine how much value is created. Fidelity Investments transformed into a technology-driven financial services company, boosting assets under management.

  • The likelihood of multihoming and disintermediation for the Parkinson's app is low because the value comes from integrating related networks. As the app gathers more data and engages physicians, multihoming is even less likely.

  • To determine value capture opportunities, analyze the networks involved. The Parkinson's app creates significant value for patients, physicians, researchers and insurers at scale due to learning and network effects. To reach critical mass, the app should be free for patients and physicians. Other options include:

  • Provide the app for free and benefit from increased branding and exposure to the pharma business. Even a small increase in pharma revenue could fund the app.

  • Targeted ads, physician referrals, insurance subsidies and anonymized data monetization. The app could be a good business and add value.

  • The app could bridge networks for more value creation or capture, e.g. insurers push the adoption of similar apps; physicians enable bridging to other disease networks.

  • Strategic collisions happen when a firm with a digital operating model targets an application traditionally served by a conventional firm. Because they have different scale, scope and learning dynamics, collisions transform industries and reshape competitive advantage.

  • It takes time for digital operating models to generate economic value comparable to traditional models. Executives often don't initially believe digital models will catch up. But after reaching critical mass, digital models can overwhelm traditional firms.

  • Examples: Airbnb vs Marriott and Hilton; Airbnb's lean, data-centric model enables rapid growth while traditional hospitality is constrained. Airbnb moves labor to the edge, uses data and AI for insights and growth. Airbnb and Booking have quickly gained share, increased services and boosted concentration. Marriott responded with a Starwood merger to compete, harnessing data and redesigning its model.

  • The dynamics: Digital models add a software layer, acting as an "operating system." By moving bottlenecks outside, they remove constraints on scale, scope and learning. Network and learning effects amplify value. More demand spurs more supply; more supply spurs more demand. Data trains algorithms to optimize decisions and engagement. The more data, the more optimization and user engagement. AI, learning and network effects enable rapid growth.

  • In summary, digital operating models can transform competitive dynamics and advantage in an industry through strategic collisions with traditional firms. By leveraging scale, scope, learning, network and AI in new ways, digital models can quickly gain the upper hand.

  • AI and digital technologies are transforming the economy at an unprecedented speed and scale. This is generating opportunities but also challenges like job displacement and social dislocation.

  • Capabilities are becoming more universal and horizontal. Competitive advantage is shifting from specialized, vertical expertise to universal capabilities like data, analytics and algorithm development. This is eroding traditional differentiation strategies.

  • Traditional industry boundaries are disappearing as digital interfaces allow business models to cut across industries. Industries are merging with each other as capabilities become more universal and data/analytics transferable across contexts. Digital networks are not constrained like human organizations.

  • Digital operating models are removing traditional operating constraints, allowing new companies to scale at unprecedented rates. This frictionless impact also makes these systems hard to stop once in motion and prone to instability. Frictionless systems amplify information, opinions, biases and aggression at vast scale.

  • In summary, the age of AI is transforming how businesses operate and compete. Capabilities, business models, and entire industries are being reshaped. While this enables new opportunities, it also introduces new challenges that leaders must grapple with.

  • Software-based entrants like smartphones posed a threat to traditional hardware makers like Samsung and Nokia.

  • Samsung survived by focusing on hardware and components, though less profitable than software players. Nokia failed by refusing to change and betting on the wrong software (Windows OS).

  • The pattern of software/data-centric entrants challenging traditional firms is repeating in many industries:

  • In computing, mainframes/minicomputers were disrupted by PCs and now cloud computing is disrupting traditional software. Microsoft overtook mainframe makers, now Amazon/Microsoft lead in cloud.

  • In retail, e-commerce platforms like Amazon digitized transactions and now have data-centric operating models that traditional retailers struggle against.

  • In entertainment, music/video streaming services like Spotify, Netflix have disrupted traditional music labels, pay TV. Napster started music streaming but failed; successful models found sustainable value capture through ads, subscriptions.

  • Simply putting a business online is not enough to topple industry giants. The difference is having a software- and data-centric operating architecture, which unlocks network effects, personalization, and new sources of value.

  • Once firms have this kind of architecture, transformation to new tech generations is easier. Experience and a less siloed/fragmented operating model help.

In summary, across industries we see a pattern of traditional companies facing competition from highly scalable, data-driven, software-centric business models. Success requires transformation to digitized, platform-based operating models that can drive value in new ways.

  • Peloton allows members to binge on workout classes the way Netflix subscribers binge on shows. Peloton offers live and on-demand spin, yoga, strength training, and treadmill classes.

  • Peloton has built a strong community. It has over 170,000 members on its Facebook page and many subcommunities. The live classes provide a shared experience where members can track each other’s performance, virtually high five, and follow progress. Instructors engage with members during live classes.

  • Peloton’s business model combines product sales and subscriptions. The bike requires a subscription, and Peloton has over 1 million subscribers with a 95% renewal rate. Members can also subscribe to just the digital content for $20/month.

  • Peloton’s operating model relies on scale. A single spin class can have 500 to 20,000 members. After live classes end, they become on-demand options. Peloton has expanded to other fitness areas like yoga, strength training, and treadmill workouts.

  • Peloton is product-focused but aims to create the “iPhone of fitness equipment.” It spent years perfecting its bike design. With $100M in funding, it improved manufacturing, delivery, software, content, and its supply chain.

  • Peloton is structured like a software company. It has 70+ software engineers and a sophisticated analytics platform. It gathers extensive member data to constantly improve the experience through class selection, new products/services, etc. The data increases loyalty, reduces churn, and provides opportunities for scope expansion into areas like nutrition, healthcare, and insurance.

  • Peloton has grown quickly, reaching $700M in revenue and a $4B valuation on $1B invested. Its software, data, and networks have enabled fast scaling.

  • Ocado, an online grocer, uses AI and robotics to reliably and efficiently deliver 50,000 perishable items to 1M customers in the UK. Ocado struggled initially but now exceeds expectations.

  • Ocado has a centralized data platform with detailed info on products, customers, partners, supply chain, and delivery. The data is in the cloud and used by teams to optimize all operations like delivery routing, robotics, fraud detection, and spoilage prediction. Ocado achieves 98.5% on-time delivery.

  • AI algorithms optimize Ocado’s operations like routing thousands of trucks, predicting customer orders and supplier needs, and managing warehouse bots that pick and move items. The algorithms maximize efficiency and freshness.

  • Ocado’s warehouses use 35 miles of conveyors and 10,000 bots to move 100,000s of boxes daily. Algorithms route boxes and model the entire system. Operations are flexible and scalable.

  • Ocado’s technology and operations teams use machine learning and iteration to constantly improve. Vision, trial-and-error, and high-volume iteration drive progress.

  • Satya Nadella reorganized Microsoft's engineering groups.

  • Scott Guthrie led Azure software, Todd Homdahl and later Rani Borkar led hardware, and Mike Neil led advanced hardware.

  • Guthrie pushed agile methods and restructured teams around customer needs. Teams had to respond faster to operations.

  • Microsoft entered an AI transformation, with Cloud and AI under Guthrie, and Experiences and Devices under Rajesh Jha.

  • Microsoft had invested in AI since the early 2000s under Harry Shum. The announcement accelerated AI development and products.

  • Microsoft's AI strategy centers on its developer ecosystem and Azure. Azure offers AI services like search, vision, language, and speech.

  • Kurt DelBene led the transformation of Microsoft's data, IT, and operations. DelBene reorganized IT into Core Services Engineering and Operations.

  • Core Services Engineering went proactive, ran like a product team, used agile methods, and gained budget responsibility. DelBene brought in leaders and engineers from product groups.

  • Core Services identified Microsoft's data, built data catalogs and lakes, used AI for unexpected behaviors, and provided a platform for the company.

  • Brad Smith became president to address issues like privacy, security, accessibility, sustainability, and digital inclusion.

  • Smith and Harry Shum's team governed Microsoft's AI use. After Tay, they made policies for "responsible bots" and six AI principles.

  • The policies are shaping the organization as CELA joins development, sales, and more. Microsoft is learning from experience to balance innovation and risk.

  • The authors have spent over a decade studying digital transformation, networks, and AI across hundreds of companies.

  • This book distills what they have learned about how AI and software are reshaping companies and the economy.

  • Chapter 2 introduces the concept of the "digital unicorn" - fast-growing tech startups worth over $1B. It examines 3 examples: Ant Financial, Ocado, and Peloton. These companies use digital networks and AI to achieve scale, scope, and learning.

  • Chapter 3 focuses on the "AI factory" - the scalable systems companies build to leverage data and AI for automation, analysis, and insights. It uses Netflix as an example. The factory has 3 parts: AI algorithms, data pipelines, and infrastructure.

  • Chapter 4 contrasts traditional company architectures with the integrated, data-centric architectures of modern tech companies like Amazon. New architectures remove constraints on scale, growth, and learning.

  • Chapter 5 examines how companies transform into AI-driven organizations. It looks at Microsoft's transformation and studies 350 other companies. Advanced companies had superior growth and financials. It describes common AI implementations and Fidelity's transformation.

  • Chapter 6 discusses the strategic implications of digital networks and AI. It introduces "strategic network analysis" to analyze opportunities. It examines Uber's strategic position.

  • Chapter 7 looks at competitive dynamics between traditional and tech-driven companies in areas like smartphones, home sharing, and automotive. It discusses broader implications.

  • Chapter 8 examines ethical issues like digital amplification, algorithmic bias, data privacy, platform control, and equity. It considers responsibilities of leaders and regulators.

  • Chapter 9 describes the broader implications for companies and governments. It lays out new rules for the AI era shaping key arenas and the future.

  • Chapter 10 focuses on the leadership challenge. It identifies opportunities for managers and entrepreneurs. It considers actions for leaders of traditional and tech companies as well as regulators and communities. It summarizes how to lead in an increasingly digital world and shape the future.

  • The authors believe AI-powered transformation can benefit any organization with commitment and investment. While startups have advantages, established companies can also adapt and succeed. The book aims to provide insight to prepare for the collisions ahead.

• Digital companies with strong network and learning effects can generate rapidly increasing value through self-reinforcing loops. As the network grows and more data is generated, the services and value improve, attracting more users and data. This continuous cycle fuels increasing returns to scale.

• Compared to traditional companies where value levels off with scale, digital companies have the potential for exponential value growth. This makes it very difficult for traditional companies to compete. Although they won't disappear, their profits will shift to the digital "operating system" companies.

• The smartphone industry offers an example of this dynamic. Nokia dominated the industry for years with a traditional, product-focused operating model. Although innovative, each Nokia phone was optimized separately, with different software and interfaces. This prevented Nokia from building a developer ecosystem and benefiting from network effects.

• In contrast, the iPhone and Android provided a consistent digital platform, enabling the formation of a huge app developer network. As the networks grew, value accelerated rapidly due to increasing returns to scale. Within five years, Apple and Android displaced Nokia and other incumbents, capturing nearly all industry profits.

• Nokia had two options: transform into a digital operating model, or focus on becoming a component supplier to the dominant smartphone OS companies. Nokia chose the latter path, but it was too late. The market had tipped in favor of iOS and Android.

• The lesson is that once critical mass is reached, digital network effects can quickly overwhelm traditional operating models. Incumbents must transform early or risk rapid decline. The next decade will see many similar battles as digital disruptors collide with traditional industries.

That's the high-level summary and key lessons from the analysis. Let me know if you would like me to clarify or expand on any part of the summary. Here is a summary of the relevant information from the passage:

  • Netflix gathers an enormous amount of data from its members including:

  • Interactions with recommendations like scrolls, mouse-overs, clicks and time spent

  • Members’ searches within the Netflix service

  • Social interactions with friends on the service

  • Demographics, location, language, temporal data, etc.

  • Netflix combines this internal data with external data like critics reviews and box office performance

  • This data is cleaned, integrated and used to provide personalized recommendations and customize each member’s experience

  • Netflix has created about 2000 “taste communities” that group together members with similar interests

  • The depth and breadth of Netflix’s data is unmatched in the industry

  • “Datafication” refers to systematically extracting data from normal business activities and transactions to provide new value. Netflix has “datafied” TV entertainment.

  • Companies need to invest in cleaning and integrating their data to build an effective “AI factory”

  • After gathering data, companies must clean, normalize and integrate the data. This is challenging work that requires investment to address inaccuracies and inconsistencies.

  • Algorithms are the rules and models that make the data useful by generating predictions, insights and solutions to problems.

  • Netflix would use algorithms to predict customer churn by analyzing usage, satisfaction, demographics and relationships with other users.

  • Predictive algorithms have become much more widely used and accurate due to increases in data and AI

  • AI algorithms can provide predictions, recommendations, image recognition, translation and more. Some systems use multiple algorithms simultaneously.

  • Although algorithms are now widely used, many of the foundational statistical models have been around for over 100 years. Neural networks have been around since the 1960s but are now being used at scale.

  • Most AI systems use statistical machine learning models like supervised learning, unsupervised learning or reinforcement learning.

  • Supervised learning algorithms are trained on expert-labeled data to predict an outcome as accurately as possible, like classifying images as cats or dogs. Here is a summary of the impact of AI on society:

Positive impacts: •AI can automate many routine tasks and jobs, freeing up humans to do more creative and fulfilling work. This could positively impact productivity and economic growth.

•AI can enhance and augment many existing jobs. For example, AI can help doctors diagnose diseases, assist teachers in the classroom, and augment the work of engineers and scientists.

•AI can help solve many complex problems that currently challenge society like developing renewable energy, reducing pollution, improving transportation systems, and accelerating medical research.

Negative impacts: •AI may significantly disrupt labor markets and displace many human jobs. Many routine jobs like cashiers, telemarketers, and assembly line workers are at high risk of automation. This could lead to job losses and workforce disruption.

•Bias and unfairness in AI systems can negatively impact many people. For example, biased AI used in hiring, lending, and healthcare can disadvantage and harm vulnerable groups.

•Lack of transparency and explainability in many AI systems can be problematic. It may be difficult to understand why an AI system behaves in a particular way or makes a specific decision. This can reduce trust in AI and limit its applications.

•Autonomous weapons powered by AI could be very dangerous if misused. There are also concerns about an AI arms race leading to increasingly automated weapons.

•Loss of human control and oversight is a risk as AI systems become more autonomous and powerful. There are fears about advanced AI systems causing unforeseen disruptions and potentially even harming humanity.

In summary, AI will likely positively transform our lives and society in many ways, but we must be proactive and manage the risks from the negative consequences and unintended impacts of AI. With proper safeguards and oversight in place, AI can achieve its promise to benefit humanity. But we must be vigilant and thoughtful about how we develop and apply AI technology.

  • Traditional organizational structures based on functional silos have limits to scale, scope, and learning. They hit diminishing returns.

  • Before innovating its operating model, Amazon looked like a traditional firm with disconnected organizational and technological silos. It needed an architectural change to increase scalability and scope.

  • Jeff Bezos aimed to redesign Amazon's technology and organization. He built a software and data platform to run significant parts of operations. The new platform took time but propelled Amazon to leadership.

  • The new platform enabled two-pizza teams to work independently while sharing code and data. This preserved agility while enabling scale and scope economies. The platform also enabled machine learning and AI deployment across Amazon.

  • AWS, Amazon's cloud division, democratizes access to services and AI tools. Customers can apply them to their own problems, fueling broad AI progress. Amazon offers software and pre-packaged tools for customers to go from data to insights.

  • Amazon's transition exemplifies a trend of AI-driven firms with software, data, analytics, and agile teams enabling huge scale, scope, and learning. These operating models threaten traditional firms.

  • Digital operating models have virtually boundless functionality via unlimited connections and data aggregation. They need a shared foundation—a platform encompassing data, technology, and AI— that teams can rapidly deploy for various use cases.

  • An ideal digital operating model has: a data, software, and algorithmic foundation from an AI factory; interfaces enabling applications for operating tasks; agile, cross-functional teams developing applications; and a focus on continuous learning through real-time data and experimentation.

The key to building a robust AI factory is ensuring modularity in both code and organization. Clear interfaces allow for decentralized innovation at the module level; as long as there is a standard for sharing data and functionality, each module can improve independently. APIs compartmentalize the problem and enable agile teams to focus on specific tasks without compromising the whole.

For companies like Alibaba that expose data to external partners, consistent and secure data platforms are crucial. Alibaba’s Taobao, for example, lists over a billion items from third parties. Clear, secure APIs enable required functionality and allow internal developers and external sellers to access over 100 data platform modules. Well-designed APIs free up Taobao’s engineers and unleash external creativity.

A state-of-the-art AI factory improves data governance and security. Massive amounts of sensitive data from users, suppliers, partners and employees must be stored securely and governed properly with checks on access and usage. APIs control the flow of data in and out of systems. They force companies to define what assets are available internally and externally. Data flowing through APIs can make or break digital companies. Cambridge Analytica accessed too much Facebook data through a hole in an API.

Infrastructure for an AI factory is often on the scalable cloud, built with standard components and open source software, and connected to core processes that shape value delivery.

The Laboratory of Innovation Science at Harvard built an AI system to map lung tumor shapes from CT scans in 10 weeks. They leveraged their AI factory and crowdsourced algorithm design contests on Topcoder. The top 5 algorithms, using CNNs and random forests, performed as well as human experts. This shows organizations don’t need vast data, resources or talent to build an AI factory.

However, in large, complex companies with siloed, outdated systems, embedding AI in operating models is challenging but critical. Operating architecture must be designed strategically at senior levels.

Jeff Bezos’ 2002 mandate required all Amazon teams to expose data and functionality through service interfaces and communicate only through those interfaces. All interfaces had to be externalizable. This overcame growth problems from weak connections across acquired businesses and technology. Consistency in data and operating models was required. The mandate illustrates how twenty-first century firms must be architected differently, with different business and operating foundations, to thrive.

  • Digital systems can amplify the impact of human biases, both implicit and explicit. Even if only a small percentage of individuals demonstrate biased behavior, at scale, the impact can be significant.

  • There are two major types of algorithmic bias:

  1. Selection bias: The input data does not accurately represent the population or context being analyzed. This can lead to flawed predictions and decisions. Examples include Amazon's HR system penalizing female candidates and AI beauty contests favoring white contestants.

  2. Labeling bias: Bias introduced when data is labeled or tagged, often by crowdworkers. Studies show labeling bias in image datasets, medical diagnoses, and other areas. Algorithms trained on biased data can amplify the bias.

  • Some algorithmic bias is inevitable given limitations in data and human knowledge. However, by choosing appropriate models and datasets, bias can be reduced. Understanding and addressing bias is crucial for ethical AI.

  • Cyber threats are increasing in scale, frequency, and impact. Traditional breaches, like Equifax, expose sensitive data by exploiting vulnerabilities. A new type of attack uses AI and massive datasets to generate synthetic data or manipulate people.

  • The Equifax breach in 2017 exposed the data of nearly 150 million Americans. The attackers exploited a vulnerability in open-source software that Equifax used. Although Equifax had been warned about the vulnerability, it failed to patch its systems. The breach was a result of Equifax amassing and monetizing huge amounts of personal data, creating a "nightmare scenario."

  • Overall, the increasing power and scale of digital systems present complex new challenges around bias, cyber threats, privacy, and other issues. With great power comes great responsibility, and we must apply ethical principles to the development and application of these advanced technologies. Failing to do so could have devastating consequences.

Traditional firms have siloed operating architectures that consist of specialized, largely autonomous functions and units. These operating models evolved to manage operational complexity by breaking up organizations into separate units focused on individual tasks. These units were given independence to maximize flexibility and minimize communication challenges.

Some of the earliest examples of distributed commercial operating architectures date back to 15th-century Italy, where the wool and textile trades had specialized units for production, distribution, banking, and insurance. The Dutch East India Company, founded in 1602, achieved scale by integrating shipping operations but had a multi-unit structure to manage its complex operations. Its siloed operating model worked well given communication and management challenges at the time.

Mass production enabled by the Industrial Revolution drove increased specialization and standardization. Unlike traditional craft methods where workers did all tasks, mass production meant each worker focused on a single component or stage. This specialization allowed scale and learning advantages. Firms organized into functional silos focused on finance, marketing, operations, etc. Information technology systems were built along these functional boundaries, limiting impact and scale.

Amazon CEO Jeff Bezos tried to break this architectural inertia with his 2002 memo calling for a transition to a software-, data-, and AI-driven operating model. Rather than siloed functions, this digital operating model is integrated yet modular, with software at its core powering new organizational forms and economies of scale, scope, and learning. To understand this transition, we need to explore the historical roots of operating models and why traditional architectures are so entrenched.

The key takeaway is that firms have traditionally had siloed, functionally oriented operating architectures, but new digital models with integrated software, data, and AI at their core can enable new forms of organization and advantage. Bezos understood early on the need to transform Amazon's operating architecture to achieve a sustainable competitive advantage in the digital age.

  • A new meta refers to a new reality that transcends existing constraints and rules. It is like changing the rules of a game midway.

  • We are entering a new meta driven by AI that is changing how firms operate and create value. This is enabling growth and opportunity but also leaving us struggling to cope with the implications.

  • A similar fundamental shift happened with the Industrial Revolution. It drove a move to specialized work and engineered production processes. This disrupted traditional methods, concentrated wealth, and caused unrest.

  • The current shift to a new meta is:

  1. Systemic: Change is happening across all industries globally, driven by relentless improvements in digital technology and AI. This differs from the localized waves of change in the Industrial Revolution.

  2. Impacting all occupations: AI and software will transform the nature of virtually every job. Many tasks will be automated, creating opportunities but also dislocation.

  3. Happening at an accelerated pace: The pace of change shows no signs of slowing as more resources are devoted to advancing AI and computing. We are only at the beginning of massive, systemwide change.

  4. Creating both opportunities and challenges: AI will enhance and replace many human tasks, enabling new ventures but also job displacement. Estimates suggest up to half of jobs could be impacted.

  • In summary, we are entering a new meta that will drive a fundamental shift in how value is created and captured. Managing the scale and pace of change will be key to harnessing the opportunities and mitigating the challenges. A glance at history provides clues for how to navigate such a shift.

  • Network effects generate value in a business’s products and services. Specifically, for search engines, users value speed, accuracy and comprehensiveness which improve with more use. Advertisers value more users because more data improves ad targeting.

  • Companies can leverage one type of network effect to generate the other. For example, Facebook started with direct network effects of people connecting with friends. But then enabled indirect network effects by allowing content creators, games and websites to access users. Gaming consoles also started with indirect network effects of players and game makers, and added direct network effects by enabling communication between players.

  • Although larger networks often mean more value, the relationship is complex. Businesses with weak network effects are easier to start but any advantage is less sustainable. Businesses with strong network effects have fewer competitors, more market concentration and a bigger competitive advantage.

  • Learning effects can add value to network effects or provide value on their own. For example, Google's algorithms improve the more people search, creating a learning effect. Learning effects depend on scale - more data means algorithms can solve more complex problems. The impact of data on competitive advantage depends on the algorithm, type of data and scale of data needed. Simple algorithms and widely available data provide little advantage. Complex algorithms and unique, large-scale data can create barriers to competition.

  • Network structure also impacts how value increases with size. Global networks concentrate around hubs, have high barriers to entry and are easy to profit from. Clustered networks are fragmented, with value depending mostly on local scale. Competitors can more easily reach critical mass in a local cluster. Uber is an example of a clustered network, with competition in each city, while Airbnb is more of a global network. Here is a summary of the passage in 350 words:

The authors conducted research on 350 large companies to assess their data, analytics and AI capabilities. They found significant differences across companies. Top companies have developed sophisticated data and AI capabilities which have led to superior business performance.

The research focused on manufacturing and service companies with 6000 employees and $3.4 billion in revenue. The AI maturity index measured data analysis, advanced analytics and AI capabilities. Companies at the bottom used basic tools and had siloed, scattered data. Top companies had integrated data platforms and used AI for automation and insights.

Top companies benefited in several ways. They had a single source of truth about the business and used analytics for personalized customer experiences, reduced churn, predicted equipment failure and real-time process automation. They understood markets, acquired customers and optimized advertising. They had a 360-degree customer view for tailored offers and support.

Top companies also used data and analytics in engineering, manufacturing and operations. They consolidated product lifecycle and supply chain data, analyzed it frequently and automatically. They understood operational efficiency, product quality, anticipated downtime and improved compliance. Some used IoT for connected products to optimize manufacturing, service and customer value.

Top companies had sophisticated data platforms for agile teams to rapidly deploy applications, improve business performance and the customer experience. They used data for forecasting, recommendations and functions like optimizing business strategy and employee development plans.

Financially, top companies had 55% gross margin, 16% earnings before tax and 11% net income, vs. 37%, 11% and 7% for bottom companies.

There are four stages of digital transformation:

  1. Siloed data: Little sharing or barriers.

  2. Pilots: Demonstrate value without big changes. Done by vendors and consultants.

  3. Data hubs: Rearchitect to aggregate data and identify opportunities. Requires investment and change.Faces resistance.

  4. AI factory: Standard AI operating model. Centralized data, algorithms, components. Policies and governance for issues like privacy and bias. Builds AI skills across organization.

Fidelity Investments is an example. In 2011, they launched an AI Center of Excellence for initiatives governed by leadership. They identified many use cases and needed capabilities. They hired top data scientists attracted to the culture, data and use cases. They developed data-focused product managers to identify and deploy applications.

In 2012, Fidelity focused on an integrated data strategy. They centralized data for a 360-degree customer view to tailor offers, anticipate needs and improve service. This data foundation enabled AI applications across the firm to transform customer experiences. Fidelity became an AI-enabled firm, with data as a strategic asset to optimize key processes.

  • Uber connects multiple networks including riders, drivers, food providers, and healthcare providers. However, Uber’s core rider-driver network faces challenges in value creation and capture.

  • Uber’s rider-driver network lacks direct network effects and geographic clustering weakens network effects. The presence of more riders and drivers does not directly benefit others. Local network effects mean competitors can gain scale in specific locations.

  • Uber benefits from learning effects that help optimize its service but multihoming on rider and driver networks hinders value capture. Many riders and drivers use multiple ridesharing apps.

  • Disintermediation threatens Uber’s transaction-based model. Once riders and drivers connect, they have little incentive to continue booking rides through the Uber hub. Uber deters disintermediation through policies like withholding contact information until payment but these strategies are not always effective.

  • Network bridging presents opportunities for Uber. Uber Eats, Uber Health, and other partnerships can generate new value by connecting Uber to more networks. Data and scale from Uber’s core business enable it to bridge into new markets.

  • Overall, while Uber faces substantial challenges in value capture from its rider-driver network, network bridging into complementary markets may help sustain and grow its business model. Connecting across networks is key to improving value creation and capture.

Alipay is a mobile payment system launched by Alibaba in 2004. It allows customers to make payments by scanning QR codes with their mobile phones, without requiring any additional hardware. Alipay charges merchants a 0.6% fee per transaction. Alipay users can use the app to pay for a variety of goods and services as long as the vendor also uses Alipay.

Alipay became very popular in China due to a lack of competition from state-owned banks and the large market opportunity. Ant Financial, the company that operates Alipay, rapidly expanded the services offered in the Alipay app. These include Yu’e Bao, an investment platform; Zhima Credit, a social credit scoring system; MYbank, an internet bank; insurance; and many other services. By 2019, Alipay had over 700 million users and dominated China’s mobile payments market.

Alipay relies on a modern digital operating model. It makes extensive use of AI and automation, for example to instantly approve small loans with no human review. It also relies on a sophisticated data platform that collects information on all of its users’ transactions and behaviors. Alipay uses this data and AI to ensure trust and security in its transactions, checking information on both buyers and sellers in real time.

Alipay gets data from four main sources: its own records of user behavior, transaction data from Alibaba sellers, public government data, and data from its business partners. It uses this data to calculate credit scores and customize services for each user. The more data Alipay collects over time, the better its AI systems get at using that data to make decisions.

In summary, Alipay has built a very successful business by creating a sophisticated mobile payments infrastructure in a market underserved by traditional banks. It uses a modern digital operating model powered by AI, automation, and data to provide users with a broad range of customized financial services. Alipay demonstrates how data and AI can transform a traditional industry like banking.

Traditional operating models are characterized by siloed organizational architecture where work is highly specialized and compartmentalized into different units. This model was pioneered by Ford's assembly line, which focused on standardization, specialization, and scale. It spread to automotive and other manufacturing industries, enabling huge increases in productivity and scale.

Over time, this model also spread to service industries like retail, hospitality, and banking. Modern manufacturing facilities like Foxconn’s iPhone assembly lines still rely on a siloed architecture with narrowly specialized, standardized work.

Although IT systems have been adopted to improve traditional operating processes, they have largely mirrored and reinforced the siloed architecture. Data and applications remain embedded within organizational units. Integrating them across silos is difficult, time-consuming, and unreliable.

While the traditional model enabled success and innovation, it also faces limits. As organizations grow in scale, scope, and complexity, diseconomies emerge. Bureaucracy, inefficiency, and inertia build up. Optimal size is reached, and productivity and innovation suffer. IT has not substantially alleviated these constraints. Integrating disparate legacy systems is painful.

In summary, traditional operating models shape and limit organizations. Their siloed architecture makes value delivery subject to diminishing returns.

  • A fully digitized business has many options to create and capture value because value creation and capture can be separated easily and often come from different stakeholders. Google's services are free but they make money from advertisers.

  • A digital firm has a very different operating model. The operating model delivers the value promised to customers. It is the plan to achieve the goals of the business model.

  • The goals of an operating model are:

  1. Scale: Deliver value to as many customers as possible at the lowest cost. Achieve efficiency through volume or complexity.

  2. Scope: The range of activities a firm performs. Achieve efficiency across multiple product lines or business units.

  3. Learning: Continuously improve, increase performance, and develop new products/services. Innovation and learning help firms stay viable and competitive.

  • A firm performs best when its business model and operating model are aligned. The more a firm can drive scale, scope and learning, the greater its value.

  • Digital technology in operating models can replace human labor and remove constraints, allowing for new levels of scalability, scope and learning. Three examples:

  1. Ant Financial (of Alibaba): Built for scale. Uses an escrow system (a third party holds payment until agreement is fulfilled) to facilitate trust in e-commerce transactions. Grew rapidly by increasing number of buyers, sellers and transactions. A network effect creates increasing returns to scale.

  2. Amazon: Has an operating model optimized for scope and learning. Offers a wide range of products and services. Has a decentralized structure that encourages experimentation and innovation. Uses data and algorithms to gain insights and improve offerings.

  3. Tesla: Has a digitally-enabled operating model focused on learning and innovation. Continuously improves products based on data from customers' use of vehicles. Over-the-air software updates add new functionality. Aims to have a fully autonomous vehicle.

The nature and purpose of firms is well understood. Firms exist to enable the coordination of complex tasks and lower transaction costs compared to individuals working through markets alone. A firm's value is shaped by its business model, which defines how it promises to create and capture value, and its operating model, which defines how it actually delivers value.

A company's business model has two parts:

  1. Value creation: Solving a customer problem or need. This could be based on quality, cost, experience, etc. The factors that determine value creation can change, e.g. technology and ease of use have become more important for cars. The specific problem a company solves and its positioning determine its approach to value creation. For ride-sharing companies, value creation relies on network effects between drivers and riders.

  2. Value capture: Generating revenue and profit from the value created. For traditional companies like auto makers, this is the difference between the sales price and cost of manufacturing. For digital companies like ride-sharing, it's based on consumption and pay per use. Although the revenue share going to drivers is high, the companies still need to generate margin.

Digital companies are innovating new business models by experimenting with different ways to create and capture value. Traditional companies typically create and capture value from the same source in a straightforward manner. Digital companies often create value for one set of stakeholders and capture value from another, or enable value creation and capture across a platform.

The key capabilities that enable new digital business models are:

  1. Digital scale and scope: Leveraging technology to serve huge numbers of stakeholders with a wide range of services. E.g. Ant Financial serves 700M+ customers with payments, lending, insurance, investing, etc.

  2. Continuous learning and improvement: Using data and AI to gain real-time insights into stakeholders and improve services. E.g. Ant Financial uses data to determine fraud risk, loan qualifications, new features, etc.

  3. Ecosystem engagement: Working with external partners in a collaborative way. E.g. ride-sharing companies rely on networks of independent drivers.

  4. Light, agile operating models: A small number of employees using technology to coordinate complex operations. E.g. Ant Financial employs only 10K people, while Bank of America employs 209K.

The new breed of digital firms like Ant Financial, Ocado, Peloton, and Google are leading the transformation of the economy by deploying these new capabilities. Existing companies need to understand and adapt to this trend to compete.

  • Technology and digital transformation are rapidly changing industries and organizations. Leaders today need to deal with constant change and disruption. Innovation and entrepreneurship are key ways to navigate this change.

  • Regulations are struggling to keep up with the pace of technological change. Key areas of focus include privacy, antitrust, traffic safety, and bias. Europe has led on privacy with GDPR, but its impact on startups is a concern. Regulation of hub companies is complex, and collaboration with industry is needed. Challenges like inequality and bias are hard to fix, so flexible, collaborative regulatory approaches are most promising.

  • Communities are an important complement to regulation. Open source software communities have had a huge impact, from Linux to TensorFlow. Community models like Wikipedia show how open, transparent processes can reduce inaccuracy and bias over time. Communities could help solve issues like algorithmic bias and fake news. Companies collaborate in communities too, like the Apache and Mozilla foundations. Community leadership and participation should be encouraged.

  • In summary, technology change requires innovation and community participation to navigate. Regulation is necessary but hard to get right and keep up to date. Community models offer promising ways to solve complex, dynamic challenges, and policy should promote community leadership. Overall, a mix of policy, corporate, nonprofit, and grassroots community efforts will be needed to govern the digital economy in a way that benefits all.

Traditional organizations struggle with operational complexity as they scale up, incurring higher costs and decreasing service levels. In contrast, Amazon embraces digital scale and improves with size and complexity. Its digital systems, powered by AI and machine learning, scale easily and continue improving.

Amazon's product suggestions are generated by algorithms that analyze huge amounts of data to determine what customers are likely to want. This system improves over time and with more data. It does not face human complexity costs like communication overhead. It can also connect recommendations across products.

Amazon has collided with and transformed traditional industries by digitizing operations. Its service improves with volume, whereas traditional businesses bog down in complexity. As Amazon grows, traditional competitors lose out.

Amazon Echo and its voice assistant Alexa show how Amazon is expanding into new areas. Alexa started with simple commands but has gained thousands of skills as it gathers more data. It continues transforming how people do tasks. And it connects users to many services and products, with potential to address more needs over time.

Amazon's model is scaling well, disrupting major industries and threatening established competitors. However, its broad impact is also drawing more scrutiny. Its future growth depends on balancing benefits to consumers against economic disruption. Competitors like Walmart are responding to remain competitive.

Walmart, though a longtime tech leader in supply chain, is rearchitecting on a digital and AI foundation to confront Amazon. It is acquiring digital firms, partnering with Microsoft, and growing online revenue. But it must also evolve physical stores using data, analytics and AI to match digital convenience and personalization. Ironically, improving in-store shopping means applying the capabilities that online shopping now offers. Physical retail's frustrations show how far it must go to meet the expectations that e-commerce has set.

  • Qihong Liu and Vanessa iness from their study of 150,000 product offerings found evidence that platforms and hub firms wield excessive market power and shape competition.

  • However, there are factors like multihoming, network clustering, and competition that curb the dominant behavior of these firms. For example, Uber and Lyft's ability to raise prices is limited by widespread multihoming of riders and drivers and network clustering of local ride-hailing services.

  • To reduce multihoming and clustering, Uber and Lyft have implemented features to tie drivers and riders to their platforms through app features, pricing, incentives, and acquisitions.

  • While platform firms shape the economy and harvest significant profits, simplistic solutions like breaking them up make little sense. Effective regulations and community involvement are needed instead.

  • Leaders of platform firms have a responsibility to address the challenges posed by their business models. Firms like Google, Microsoft, and Facebook are investing in research to address issues like algorithmic bias and misinformation.

  • Platform firms that occupy central network positions, like Facebook and Equifax, have become "keystone species" in the digital economy. Like keystone species in an ecosystem, their activities impact the entire network. They thus have a responsibility to sustain the health of their networks.

  • The concept of "information fiduciaries" suggests that firms like Google and Facebook that control extensive consumer data have a responsibility not to harm the communities they collect information from. They could adopt fair information practices and privacy guarantees in return for less regulation.

  • Leaders of major platform firms are increasingly aware of their impact and responsibility to sustain not just their shareholders but the broader communities and economy. They should take actions to enable the long-term health of the networks they depend on.

  • Cambridge Analytica was a data analytics firm that offered services to influence voters using psychographic profiles created from Facebook data.

  • In 2015, Cambridge Analytica worked for the Brexit campaign and Ted Cruz's presidential campaign. After Cruz dropped out, they worked for Donald Trump's campaign.

  • In 2018, it was revealed that in 2014, Cambridge Analytica had obtained data on 50 million Facebook users from a researcher named Aleksandr Kogan. Kogan had created a personality quiz app that accessed users' data and the data of their friends.

  • Facebook's policies allowed app developers to access friend data at the time. But selling or transferring that data for advertising or monetization was prohibited. When Facebook found out in 2015, they suspended Cambridge Analytica's access and demanded they delete the data. Cambridge Analytica claimed they did delete it but had not actually done so.

  • The story illustrates the control challenges of digital platforms. Digital platforms gain power through openness and connectedness, but this also exposes them to unforeseen and unintended uses, which can be hard to curb while still enabling innovation. Defining and enforcing "good" uses of a platform is difficult. The more open the platform, the bigger the risks. But too much control can also hamper usefulness and adoption.

  • The challenges around control, bias, security, and others point to new ethical questions for companies and societies in the digital age. Meanwhile, network effects are leading to increased market concentration, raising concerns about equity and fairness. Powerful platforms like Apple's iOS and App Store or Amazon's marketplace can effectively control access to large groups of consumers or suppliers, forcing others to accept their terms to reach those communities. • Drivers use multiple ride-sharing apps to find the cheapest fares. This makes it easy for riders and drivers to switch between services, threatening Uber’s business. • However, Uber has implemented measures to keep riders and drivers loyal, including penalties for drivers who break the rules. As a result, riders and drivers switching away from Uber is not currently a major problem. • Still, competition from other ride-sharing services poses a threat to Uber’s profitability. Unless Uber can achieve major cost savings from experience, its core ride-sharing business may remain unprofitable.
    • Uber’s future profitability depends on connecting its ride-sharing networks to other networks to generate more value and revenue. Examples include grocery delivery, food delivery (Uber Eats), and healthcare (Uber Health). Some of these opportunities seem more promising than others. • Questions entrepreneurs should ask about their network-based businesses include: › What is your core service and main network? Analyze its characteristics. › How can you strengthen network and learning effects over time? How can you provide more value? › If you need scale to deliver value, how will you achieve critical mass? › What secondary networks could you connect to? Could they enhance network or learning effects? › Do you face challenges like network clustering, customers/suppliers using multiple networks (multihoming), or networks cutting you out (disintermediation)? › How will you overcome these challenges?

In summary, while Uber faces significant threats from competition and the ability of riders and drivers to easily switch services, connecting to additional networks may help ensure its long-term success if executed well. Asking strategic questions about networks and value creation can help in developing a successful strategy.

  • Digital platforms like Google, Facebook and Amazon have gained control of access to billions of consumers. This gives them a competitive advantage and ability to extract value.

  • The speed of digital transformation is increasing exponentially. Digital technologies like software platforms, data, analytics and AI are rapidly impacting various industries. Industries like media, banking, automotive, travel, etc. are being disrupted.

  • The spread of digital operating models is posing new challenges:

  1. Digital amplification: The algorithms that drive engagement and ad revenue on platforms can amplify misinformation and harmful content. Anti-vaccination propaganda and other kinds of false information can spread widely.

  2. Bias: The algorithms and data used by digital platforms can reflect and amplify societal biases. Studies have found discrimination on platforms like Airbnb and in financial services.

  3. Security: The large datasets used by digital firms are vulnerable to cyberattacks and can threaten consumer privacy.

  4. Control: There are concerns about the concentration of power and control in a few large tech companies. Regulators are scrutinizing their market dominance.

  5. Inequality: Although digital platforms provide economic opportunity, they are also associated with increasing inequality in the economy. The rewards tend to accrue to highly skilled, highly educated workers.

  • These challenges highlight new ethical considerations for managers and leaders. They need to understand how their digital capabilities and data can be misused, and take responsibility to address the issues.

  • No one can plead ignorance anymore. We must all understand these issues and be prepared to act to ensure the health of businesses, society and political systems.

  • Supervised machine learning uses labeled data to train algorithms to predict outcomes. The data is split into training and validation sets. The training set determines the model parameters, and the validation set tests the model's accuracy.

  • Unsupervised learning finds patterns in unlabeled data. It can be used for clustering data into groups, association rule mining to find relationships between items, and anomaly detection.

  • Reinforcement learning requires a starting point and performance function. It involves exploring the problem space to find the best solution. It requires balancing exploration and exploitation.

  • Applications of machine learning include:

-- Recommendation systems -- Image recognition -- Predicting customer churn -- Detecting fraud -- Maintaining systems and machinery -- Analyzing social media data -- Creating customer segments -- Finding topics in text

  • Companies can use their existing data from systems, technologies, and databases for machine learning. For example, insurance companies have data on claims that could be used to detect fraud. Healthcare organizations have medical data that could be used to help doctors diagnose patients.

  • We have entered an age of artificial intelligence that is transforming competition and the economy. Imperfect, "weak" AI is enough to fundamentally change how firms operate and compete.

  • Digital technology and AI are enabling new kinds of operating models that are scalable, have a wide scope, and enable continuous learning. These new operating models are overtaking traditional ones.

  • The digitization of activities like photography has enabled tech companies with digital operating models, like Facebook and Google, to thrive. These companies analyze massive amounts of data to improve their services. They have transformed photography and related industries.

  • Traditional firms have struggled to adapt to these new competitors with digital operating models. Kodak was ultimately defeated not by a competitor but by the emergence of smartphone and social media companies.

  • Digital operating models introduce new challenges, like privacy, security, and bias concerns. Leaders of these companies face different challenges than leaders of traditional companies.

  • Amazon is an example of a company with a digital operating model that is transforming traditional industries. By harnessing digital tech, analytics, and AI, Amazon has scaled massively, expanded its scope, and continuously improved. This has allowed it to outcompete traditional retailers.

  • For traditional companies, size often becomes an impediment due to operational complexity. Digital operating models do not have the same limits, so companies like Amazon have been able to scale in ways that traditional retailers cannot.

  • In summary, we are in a new age of artificial intelligence that is enabling a new breed of company with digital operating models. These companies are reshaping competition and transforming the economy.

  • Alibaba's Zhima Credit scores consumers based on their connections and actions on Alibaba's platform. The algorithms calculating the scores are constantly improving to make better decisions. Consumers with good scores get perks while those with low scores face additional fees.

  • Alibaba's AI-driven fraud prevention system monitors hundreds of user actions to detect suspicious behavior. Low-risk actions proceed automatically while risky actions trigger manual review.

  • Alibaba runs hundreds of experiments daily to learn and develop new products and features. Alibaba's growth came from combining data sources and rapid innovation by agile teams.

  • Alibaba's data and algorithms enable new financial services. Alibaba uses scenarios and prototypes to develop new solutions, refining them to attract consumers and mainstream the technology. Alibaba automates customer service using data mining and semantic analysis.

  • Digital operating models avoid direct human intervention in critical processes. While humans help with strategy, design, algorithms, and more, the actual value-delivery process is automated. Key processes rely on integrated customer and operational data. This scales easily, has minimal marginal costs, enables connections with external networks, and speeds learning and innovation.

  • In digital operating models, employees oversee an automated, algorithmic "organization" that delivers value. This fundamentally changes management, growth, and constraints. Two examples are Peloton and OpenTable.

  • Peloton sells connected exercise bikes and a subscription to live and on-demand spin classes. This digitally transforms the fitness industry by overcoming limits of time, space, and capacity. Peloton brings the studio experience home, providing many workout options. Its model differs from traditional gyms and equipment sellers.

  • OpenTable provides online restaurant reservations, enabling diners to easily book tables and restaurants to fill seats. OpenTable builds profiles of diners and provides recommendations based on preferences, habits, and social connections. Diners earn points for booking that provide perks. OpenTable helps restaurants optimize operations and market to diners.

  • AI factories are at the core of operating models for many companies. They guide critical processes and decisions while humans handle peripheral tasks.

  • AI factories create a virtuous cycle of user engagement, data collection, algorithm design, prediction, and improvement. They use data to refine and train algorithms that make predictions and drive automated responses or inform human decisions.

  • The AI factory solves the problem of analyzing huge amounts of data by applying mass production methods to data processing and analytics.

  • Netflix is an example of a company with an AI-centric operating model. Its AI factory gathers data and trains algorithms that influence its content recommendations, user experience, and content deals.

  • The AI factory has four main components:

  1. The data pipeline gathers, inputs, cleans, integrates, processes, and safeguards data.

  2. Algorithm development generates predictions that drive critical operating activities.

  3. The experimentation platform tests changes to prediction and decision algorithms.

  4. The software infrastructure embeds the pipeline and connects it to internal and external users.

  • For Netflix, inputs to the data pipeline include billions of item ratings, millions of new ratings daily, millions of stream plays daily, millions of items added to queues daily, and metadata on all content.

  • Netflix's algorithms use the data to recommend content, create a personalized user experience, determine what content to cache locally, and decide what original content to develop. Experiments help determine the effects of changes to the algorithms and user experience.

The key takeaway is that AI factories are becoming central to how leading companies operate. By systematically gathering huge amounts of data and developing algorithms and predictions, they are able to gain insights and automate decisions that would be impossible for humans alone. Companies like Netflix show how AI factories can transform industries by enabling personalized, customized experiences and services. Here is a summary of the key ideas:

  • Artificial intelligence is transforming the way firms operate and compete. AI is becoming the core of a company’s operational model, the “runtime” that shapes how a company executes tasks.

  • AI is changing the very nature of companies by enabling digital scale, scope, and learning. Digital and AI-driven processes are more scalable, connectable, and able to learn and improve. This is erasing limits that have constrained firm growth for hundreds of years.

  • Powerful AI is not required to drive major changes. “Weak” or narrow AI that performs specific tasks traditionally done by humans is sufficient to transform businesses. Systems that can replicate human judgment and reasoning, or “strong” AI, are not needed.

  • Examples like The Next Rembrandt show how AI can be applied to create works of art and transform artistic fields. AI is influencing all disciplines and industries.

  • Competing in the age of AI requires rethinking firms and how they operate. Software, algorithms, and AI are changing the nature of companies. The core of the firm is becoming a “scalable decision factory” powered by software, data, and AI.

  • Firms need a new operating architecture to take advantage of digital networks and AI. They must transform themselves into “AI companies” that leverage data, networks, and AI. This requires new strategies and leadership.

  • AI brings strategic opportunities and threats. There will be collisions as digital firms driven by AI compete with traditional organizations. There are also ethical considerations with the new scale, scope, and learning enabled by AI.

  • The age of AI is changing the rules of the game for business, society, and all of us. Leaders must adapt to the implications of firms defined by software and AI. Regulatory institutions and communities also need to understand the shift to an era when algorithms and networks are in control.


Did you find this article valuable?

Support Literary Insights by becoming a sponsor. Any amount is appreciated!