By Myles Suer, #CIOChat Facilitator and CIO.com Contributor
As someone who cares about how business strategy and digital intersect, it is great to see complementary validations of business thinking. This is what I found after reading “Competing in The Age of AI”, released on January 6th. The book’s authors extend and compliment “Designed for Digital” which was reviewed in September. Maybe the authors should take a walk across the Charles River.
In Competing in the Age of AI, authors Marco Iansiti and Karim Lakhani dig into the role of data and AI in driving the digital future. Their perspectives provide additional and supportive thinking from Designed for Digital.
Competing in the Age of AI
Marco and Karim suggest AI has become a universal engine for business execution. For this reason, AI is now core to corporate operating models for those that have a continuing right to win. While AI increasingly defines how companies drive the execution of tasks, it is now transforming how companies compete.
Like Jeanne Ross, Cynthia Beath, and Martin Mocker concept of a digital platform (in Designed for Digital), Marco and Karim suggest a digital operating model. Here, humans may design the operating systems, but computers actually do the work in real time. AI, for this reason, is seen as reshaping the operational foundations of firms and enabling what the authors suggest is digital scale, digital scope, and digital learning. Those that succeed in establishing a digital operating model can erase deep seated limits that constrain growth and the impacted hundreds of years of competitive dynamics. And at the same time, the barriers to entry suggested by fellow Harvard Business School professor, Michael Porter, in the business classic “Competitive Strategy” are increasingly of less relevance.
For digital companies, networks and algorithms become woven into the fabric of their firms. Where they place their oars, they can change how industries function and economies operate. This includes transforming competition. Geoffrey Moore in “Zone to Win” says, “business model disruptions are where the train wrecks happen. They are the Kodak moments. And no established enterprise can reasonably expect to change its business model, ever.” Marco and Karim agree and suggest that ultimately, complexity will become the downfall for traditional, legacy organizations. Simply put, complexity increases operational costs and decreasing service levels in a competition with digital disruptors.
For this reason, the authors suggest that it is time for traditional organizations to move to digital operating models and harness digital scale, digital scope and digital learning. The one contrast worth noting here is that the authors of Designed for Digital suggest that a mastery of digital requires mastery of social, mobile, analytics, cloud, and IoT too. But they and Marco and Karim believe winners have a culture and willingness to experiment and create new digital value propositions. Put simply, the corporate goal increasingly needs to be figuring out how to reimagine existing products and services using AI and other digital technologies.
Rethinking the Nature of the Firm
Marco and Karim importantly suggest AI will cause the corporation and the concept of scale to be rethought. They give an example of Ant Financial. With a staff of a few thousand people, it has grown from 10 million to 700 million users. Their secret is using an integrated platform that uses AI to power application processing, fraud detection, credit scoring, and loan qualification. Everything is automated. To put this in contrast, JP Morgan Chase—a much larger business—has 82 million online customers with 250,000 employees worldwide. Born digital Fintech companies like Ant and SoCal in the U.S. are using digital operating models that leverage digital to transform the financial services industry. In the long run, Marco and Karim believe a collision is coming between disruptors with industry incumbents.
As important, digital is creating a new kind of business model—a new way for firms to create and capture value. To be clear, value creation concerns the reason customers choose to use a company and the particular problems the company solves for them. Marco and Karim believe that as the business model changes so will the operating model which is the way firms deliver value to customers. Traditionally, Jeanne Ross in “Enterprise Architecture as Strategy“, defined an operating model, as the expectations for integration and standardization across business units. With the operating model in hand, enterprise architecture represented the organizing logic for business processes and IT infrastructure. Enterprise Architecture, in effect, reflecting the requirements of the companies operating model. And maturity moved from business silosà standardized tech architectureàoptimized coreàbusiness modularity.
Applying AI to business model thinking and operating models, considers data’s ability to drive a broad variety of functions including personalization, revenue optimization, and recommendations as well as sophisticated analytics to understand the value created by potential new products and services. It, also, involves sophisticated experimentation that allows organizations to learn and understand the opportunities and risks provided by new features and digital products. This, in the end, can transform value creation, capture, and delivery.
Delivering an AI, Digital Operating Model
Establishing a digital operating model is about creating a virtuous cycle between user engagement, data collection, algorithm design, prediction, and improvement. Everything, however, starts with data. Business leaders, CIOs, CDOs, and Enterprise Architects need to consider the accuracy and impact of predicted outcomes and create a system that furthers learning and predictions.
With this, they can create what Marco and Karim label an AI factory. There are four elements to an AI factory. Let’s take a look at each of them:
1) A data pipeline (gathers, inputs, cleans, integrates, processes, and safeguards data in a systemic, sustainable, and scalable way)
2) Algorithm development (generates predictions about future states or action)
3) Experimentation platform (the mechanisms through which hypothesis regarding a new prediction or decision algorithm are tested to ensure changes suggest have their intended effects
4) Software Infrastructure (embedded in the pipeline in a consistent and componentized software compute infrastructure)
Marco and Karim appropriately believe that a data platform starts by establishing a data pipeline. They claim that a state-of-the-art data platform powers the AI factory by data flowing through API publish and subscribe. The purpose of the platform is to make data clean, consistent, and available to applications. Marco and Karim claim what is wanted is a data super market. In the data platform, data is aggregated, cleaned, refined, processed, and made available through consistent interfaces. They suggest that the massive amount of data captured from users, suppliers, partners, and employees is extremely valuable and should not be stored in ad hoc fashion. It needs data governance and security. At the same, they suggest that organizations need to build a secure, centralized system for careful data security and governance, defining appropriate checks and balances on access and security, inventorying assets, and providing all users necessarily protections.
The AI Factory with great data can focus upon algorithm development. So then, what are the different options?
Supervised learning aims to come as close to a human expert or an accepted source of truth. Data inputs here are labeled with a given outcome. Machine learning systems in supervised learning rely on an expert-labeled dataset of the outcome and the potential characteristics or features. To make this work, the first step is to create or acquire a labeled dataset. After training, a validation dataset is used to test the accuracy of the model and determine whether there is an acceptable error rate. It is important to not over cue the algorithms for success here.
What are some examples of supervised learning in practice? There are two. Facebook’s ability to suggest names for friends who may appear in newly uploaded pictures. And the Nest’s ability to automate the temperature in your living room over time.
Unsupervised learning discovers insights in data with few preconceptions or assumptions. There are 3 types of unsupervised learning: 1) clustering data into groups; 2) association rule mining (recommendations for additional products a shopper may want to purchase based on current set in the shopping cart); and 3) Anomaly detection. Unsupervised learning aims to find natural groupings in data.
So, what are some examples of supervised learning in practice. Netflix uses unsupervised learning to discover related groups of customers in analyzing viewing data. Companies of all types use to find the reasons for customer churn. My favorite example is from Amazon which found that people buying diapers also would likely buy beer and wine at the same time. It is funny that when my kids were young, my wife and I would call the early evening the whine and wine hour.
Reinforcement learning requires only a starting point and a performance function. Here, the tradeoff is between exploration and exploitation. Netflix uses this AI approach to figure out which movie selection to present and which artwork to combine with it to maximize the match between user and recommendations.
Rearchitecting the Firm
Marco and Karim discussed a Bezos memo regarding the need for all teams to expose their data and functionality via standardized service interfaces. Digital firms clearly require an integrated, highly modular digital foundation. Their operating model is architected to take an integrated core of software, data, and AI and use it to build a new breed of organizations. Here, the organization reflects the system, and system reflects the organization. Melvin Conway is referenced here who said that that an organization is constrained to design systems that reflect the communication patterns prevalent in the organization.
For many, breaking the architectural inertia of their past established by the links with existing customers prevents them from responding effectively to disruptive change. As organizations become good at doing something in a certain way, they develop routines and systems that reinforce each other and make it difficult to do things differently. Digital transformation after fixing people and processes needs to go after the limits on the scalability and scope of business. This includes the spaghetti code or multiple versions of the same code. By doing this, the authors claim firms will be able to remake themselves so that they are powered by code instead of human labor.
Part of rearchitecting the firm involves eliminating functional silos and establishing a digital operating model. The digital operating model should be designed to unleash the potential of the digital technology. This includes data, AL/ML, APIs, and agile teams. Marco and Karim suggest that the future is delivered through small agile teams equipped with data science, engineering, and product management along with agile processes and digital operating architectures. James Staten, VP Disruptive Innovations, Forrester, agrees and says that “Forrester’s guidance is that leaders should not just form dedicated innovation teams, but they need to empower cross-company (and cross-ecosystem) innovation ideation so they have a broad set of ideas to choose from. It’s crucial that they build out a portfolio of innovations that don’t limit them to just tactical digital transformation moves nor just respond to digital disruptors who have entered their market; but to have in their portfolio, innovations that position them as the market disruptors. Yes, innovation teams should be separate from the strategy team. They should align with them and insure the innovation program that is focused on helping fulfill the company strategy but also be open to moves that align to future customer needs and can expand the company’s market appeal (moves that often go beyond and expand the strategy). This is important because too often company’s strategies are focused on existing services and products only.”
Marco and Karim go on to suggest that data should be centralized, but almost anyone with a hypothesis should be able to launch a live experiment and use the results to implement meaningful changes. At the same time, they say that the digital operating models should promote modularity and reuse of software and algorithms. Here, employees do not deliver products or services, instead they design and oversee a software automated, algorithm driven digital organization that delivers the goods. When fully digitized, a process can readily plug into an external network of partners and provides or even external communities of individuals to provide additional, complementary value. Importantly, as the organization accumulate data by increasing scale, the algorithms get better and the business creates greater value, something that enables more usage and thus the generation of even more data. In an AI powered operating model, managers are transformed into designers, shaping, improving, and controlling digital systems that sense customer needs and respond by delivering value.
Stages of Operating Model Transformation
The authors envision a process of organizational maturity. It clearly starts where MIT-CISR has found that 51% of legacy enterprise are located–silos and siloed data. From here, organizations readily justify the road forward starting with pilots demonstrating the value of analytics-based decisions. With these results, the authors suggest that organizations take a big step forward by creating a data hub. This involves them rearchitect the organization, so it aggregates data from many siloed sources and use this data to identify companywide opportunities. Adoption of a clear, single source of truth is essential to guide decisions on market opportunities, pricing, planning, and organizational excellence.
With this in hand, the final step is an AI factory. This takes major investment in developing a standard operating model for AI. This includes centralized data, powerful algorithms, and reusable software components. MIT-CISR research has found that 74% of global companies are working on building a digital platform moving from silos to a data hub to an AI factory in the language of the authors.
The authors suggest that strategy is critical to success. Competitive advantage is increasingly defined by the ability to shape and control AI networks and harvest the volume and variety of transactions they carry. Competitive advantage moves toward the organizations that most central in connecting businesses, aggregating the data to flows between them, and extracting value through powerful algorithms. Here, companies will win with digitized operating models.
For all of this work, the authors suggest that five principals need to be in place for digital transformation to work.
- There is one strategy
- Architectural Clarity
- Agile, product focused organization
- Capability Foundations
- Clear, multidisciplinary governance
Today is the time to start
Marco and Karim say that AI is past the inflection point. The time is now to respond. In the digital age, the operating rules have changed. Change in no longer localized—it is systematic. It is across all industries as the digital engine of change proliferates. Increasingly, capabilities are increasingly horizontal and universal. At the same time, traditional industry boundaries are disappearing as digital companies move from constrained operations to frictionless operations. Unfortunately, the authors suggest the process of digitization will lead to industry contraction and more inequality.
For those that succeed, transformation will have started at the top and moved motivated innovation teams below. The most difficult work will be in changing the organization and transformation architecture. This includes building the right skills, capabilities, and culture to drive an increasingly digital operating model. As the authors of Designed for Digital suggest the conundrum for legacy businesses is piecing together culture, insights, and competencies to convert a successful predigital company into an agile, innovative digital player. And once the will exists, roughly ¾ of legacy businesses have to deal with their silos or the fact that their businesses are connected with ‘duct tape’ and ‘band aids’. The reality is most legacy businesses will not cross the digital chasm. And finally, although the capabilities of AI create opportunities for new business value, a company will only realize that value if it redesigns their systems, processes, and roles, and reimages its value propositions.
It is rare that two books come along that complement and fill in each other as much as Competing in the Age of AI and Designed for Digital. The reality is that digital change will not only reconfigure industries, but it will determine who will have competitive advantage. The new tools for success will be found in digital scale, digital scope, and digital learning. Companies that weld these technologies will tip the Power of Nations in their way. However, at the same time, as the Klaus Schwab has suggested a few years ago, there will be a need for government action and possibly the concept of a living wage with “47% of the total employment in the US at risk, perhaps over the next decade or two” (The 4th Industry Revolution, World Economic Forum, 2016, page 38).
Find out more in these publications by The Open Group:
Don’t miss The Open Group San Antonio January 27-30, 2020!
Myles. Suer, according to LeadTails, is the 9th leading influencer of CIOs. He is the facilitator for the CIOChat. The chat has executive level participants from around the world in a mix of industries including banking, insurance, education, and government. Mr. Suer, also, has a weekly column at CIO.com (The Adaptive CIO) and has had his articles published in ComputerWorld, Innovation Enterprise, and COBIT Focus. For COBIT, he published “Using COBIT 5 to Deliver Information and Data Governance”. He is reviewing for CIO Magazine, Jeanne Ross’s new book “Designed for Digital”.
Much of Mr. Suer’s experience is as an integration and data practitioner. At Boomi, Mr. Suer, the head of Enterprise Marketing is responsible for aligning Boomi’s offerings with the needs of enterprise customers. Prior to Boomi, Mr. Suer was responsible for product marketing of Informatica’s Intelligent Data Platform. At HP and Peregrine, Mr. Suer led a product management team applying analytics and big data to the company’s IT management products. This includes creating the company’s CIO Scorecard Product. Prior to this, Mr. Suer has had leadership roles in start-ups and large companies. This includes being a software industry analyst. Mr. Suer holds a Master of Science degree from UC Irvine and a 2nd Masters in Business Administration in Strategic Planning from the University of Southern California.