The successful development of applications powered by machine learning AI requires both the right set of skills and expertise, as well as the right tools and technologies. Relatively few organizations in the world have all the necessary internal expertise and capabilities to build, deploy, and operate sophisticated AI applications that will drive meaningful value. The vast majority of organizations will need to engage with partners to provide the required expertise and technology stack to build, test, deploy, and manage the applications.
Machine Learning: Development and Deployment Workflow
There is significant financial upside and business benefit in understanding how to avoid the potential pitfalls of an AI development initiative so that you can quickly capture positive ROI. To get a sense of the challenges in developing and deploying AI applications at scale—and why the right expertise, partners, and development platform are critical—let’s look at what’s involved in a machine learning development process. In this section, I outline the sequential workflow in developing and deploying a machine learning AI application. This process is well understood by machine learning experts.
1. Data Assembly and Preparation
The first step is to identify the required and relevant data sets, and then assemble the data in a unified image that is useful for machine learning. Because the data come from multiple disparate sources and software systems, there are often issues with data quality such as data duplication, gaps in data, unavailable data, and data out of sequence. The development platform must therefore provide tools to address those issues, including capabilities to automate the process of ingesting, integrating, normalizing, and federating data into a unified image suitable for machine learning.
2. Feature Engineering
The next step is feature engineering. This involves going through the data and crafting individual signals that the data scientist and domain experts think will be relevant to the problem being solved. In the case of AI-based predictive maintenance, signals could include the count of specific fault alarms over the trailing 7 days, 14 days, and 21 days; the sum of the specific alarms over the same trailing periods; and the maximum value of certain sensor signals over those trailing periods.
FIGURE 6.2
3. Labeling the Outcomes
This step involves labeling the outcomes the model tries to predict (e.g., “engine failure”). Often the specific outcomes are not clearly defined in the data since the original source data sets and business processes were not originally defined with AI in mind. For example, in AI-based predictive maintenance applications, source data sets rarely identify actual failure labels. Instead, practitioners have to infer failure points based on combinations of factors such as fault codes and technician work orders.
4. Setting Up the Training Data
Now comes the process of setting up the data set for training the algorithm. There are a number of nuances to this process that may require outside expertise. For classification tasks, data scientists need to ensure that labels are appropriately balanced with positive and negative examples to provide the classifier algorithm enough balanced data. Data scientists also need to ensure the classifier is not biased by artificial patterns in the data. For example, in a recent fraud detection deployment for a utility, a classifier trained on historical cases on a large country-wide data set incorrectly identified a number of suspected fraud cases on a remote island. Further examination revealed that because the island is so remote and hard to access, investigators traveled there only if they were certain of fraud. All historical cases investigated on the island were therefore true positive labels. Consequently, the classifier always correlated the island location with incidence of fraud, so the algorithm had to be adjusted.
5. Choosing and Training the Algorithm
The next step is to choose the actual algorithm and then train it with the training data set. Numerous algorithm libraries are available to data scientists today, created by companies, universities, research organizations, government agencies, and individual contributors. Many are available as open source software from repositories like GitHub and Apache Software Foundation. AI practitioners typically run specialized searches across these libraries to identify the right algorithm and build the best-trained model. Experienced data scientists know how to narrow their searches to focus on the right classes of algorithms to test for a specific use case.
6. Deploying the Algorithm into Production
The machine learning algorithm then must be deployed to operate in a production environment: It needs to receive new data, generate outputs, and have some action or decision be made based on those outputs. This may mean embedding the algorithm within an enterprise application used by humans to make decisions—for example, a predictive maintenance application that identifies and prioritizes equipment requiring maintenance to provide guidance for maintenance crews. This is where the real value is created—by reducing equipment downtime and servicing costs through more accurate failure prediction that enables proactive maintenance before the equipment actually fails. In order for the machine learning algorithm to operate in production, the underlying compute infrastructure needs to be set up and managed. This includes elastic scale-out and big data management abilities (e.g., ingestion, integration, etc.) necessary for large data sets.
7. Closed-Loop Continuous Improvement
Once in production, the performance of the AI algorithm needs to be tracked and managed. Algorithms typically require frequent retraining by data science teams as market conditions change, business objectives and processes evolve, and new data sources are identified. Organizations need to maintain technical agility so they can rapidly develop, retrain, and deploy new models as circumstances change.
The science of AI has evolved and matured over the last several decades. We are now at a point where not only are the underlying technologies available, but also organizations now have access to domain experts, data scientists, and professional services providers that can help them harness the power of AI for competitive advantage.
Business Benefits of AI
AI technologies deliver real business benefits today. In particular, technology companies like Google, LinkedIn, Netflix, and Amazon use AI at large scale. McKinsey Global Institute (MGI) estimates that technology companies spent $20 billion to $30 billion on AI in 2016.24 Some of the most established applications for AI delivering concrete business benefits are in online search, advertising placement, and product or service recommendations.
In addition to technology companies, sophisticated industries advanced in digitization, such as financial services and telecom, are starting to use AI technologies in meaningful ways. For example, banks use AI to detect and intercept credit card fraud; to reduce customer churn by predicting when customers are likely to switch; and to streamline new customer acquisition.
The health care industry is just starting to unlock value from AI. Significant opportunities exist for health care companies to use machine learning to improve patient outcomes, predict chronic diseases, prevent addiction to opioids and other drugs, and improve disease coding accuracy.
Industrial and manufacturing companies have also started to unlock value from AI applications as well, including using AI for predictive maintenance and advanced optimization across entire supply chains.
Energy companies have transformed operations using AI. Utility companies use advanced AI applications to identify and reduce fraud, forecast electricity consumption, and maintain their generation, transmission, and distribution assets.
There are several emerging applications of AI in defense. Already, the U.S. military uses AI-based predictive maintenance to improve military readiness and streamline operations. Other use cases include logistics optimization, inventory optimization, and recruiting and personnel management (e.g., matching new recruits to jobs).
I will cover some of these industry use cases in more detail in chapters 8 and 9.
AI in Action: Addressing a $1 Billion Customer Retention
Opportunity
To illustrate how AI can address complex concerns shared by virtually every business, let’s look at an example of using AI to improve customer retention in the financial services industry. Enterprises focus considerable resources on keeping customers satisfied, successful, and engaged. Particularly in B2B industries, determining the health of a customer account can be challenging. Often, this is a matter of dedicated account managers making calls and manually tracking customer behavior. And in many cases, a company may not learn about a customer’s intention to switch to a different provider until it’s too late.
In the corporate banking market, banks compete for business based on a number of factors, including product offerings, interest rates, and transaction fees. Banks generate revenue from fees charged on customers’ transactions and from interest earned by lending out the funds in customers’ accounts. A bank’s corporate account managers, therefore, carefully track their customers’ transaction activity and cash balances, since these are key revenue drivers. This is largely a manual process, managed using spreadsheets and based on reports generated from the bank’s CRM and other systems. But since many internal and external factors—investment activity, mergers and divestitures, competitive dynamics, etc.—can impact a customer’s transaction volume and balances in any given period, there is a lot of noise in those indicators.
Corporate account managers therefore struggle to identify, as early as possible, signs that a customer may be permanently reducing or terminating its business with a bank for preventable reasons. If account managers can determine an at-risk customer early enough, they may be able to take action. For instance, a customer might reduce its business because it is financially overextended on some loans. In this case, the account manager might offer to restructure the loans or offer loan counseling services. Or maybe a competitor has offered the customer a better interest rate, in which case the account manager can extend a competitive rate.
Faced with this complexity, a leading financial services company is using an AI suite to develop an AI-powered application to assist corporate account managers in effectively identifying and proactively engaging with potentially at-risk corporate customers. The bank employs hundreds of corporate account managers that serve tens of thousands of corporate customers, representing aggregate cash balances of several hundred billion dollars. Any improvement in customer retention in this high-margin line of business represents significant economic value to the bank.
The AI application ingests and unifies data from numerous internal and external sources, including multiple years of historical data at various levels of frequency: customer transactions and account balances; changes in rates paid on cash balances; credit risk; GDP growth; short-term interest rates; money supply; and account-specific corporate action data from SEC filings and other sources. By applying multiple AI algorithms to these data in real time, the application can identify profiles of at-risk customers, predict those who are likely to reduce their balances for preventable reasons, and send prioritized alerts to account managers, enabling them to take proactive action.
Using the AI application results in far more accurate predictions and timely identification of at-risk customers than the traditional approach used by account managers. The bank estimates the incremental annual economic value of applying this AI application is approximately $1 billion—pure bottom line profit.
The Economic and Social Impacts of AI
AI will have profound consequences for society and business. According to PwC’s 2017 study projecting an increase of $15.7 trillion in global GDP by 2030 due to AI, half of that total gain will derive from labor productivity improvements and the other half from increased consumer demand. PwC estimates the potential value creation across specific industries could reach $1.8 trillion in professional services, $1.2 trillion in financial services, $2.2 trillion in wholesale and retail, and $3.8 trillion in manufacturing.
According to the same PwC study, the impact will not be evenly distributed across the globe. While North America currently leads, with Europe and developed Asian economies following, China is expected to eventually surpass others. In fact, China has made it a national priority and goal to be the world leader in AI by 2030. For organizations everywhere, and particularly those competing with Chinese counterparts, the urgency to digitally transform and invest specifically in AI capabilities grows.
AI-driven growth is a welcome antidote to a decades-long slowdown in productivity gains in developed economies. But the potential downside of advances in AI will be sharp and painful for those who do not adapt. For some organizations, their very existence is at risk.
Renowned Harvard Business School professor Michael Porter—whose book on competitive strategy is a classic in the field—speculates a “new world of smart, connected products” underpinned by AI and big data represents a sea change in the fundamental dynamics of competition.25 Porter suggests this is not simply a matter of competitive advantage; it is existential. Recall that 52 percent of Fortune 500 companies have been acquired, merged, or have declared bankruptcy since 2000. The threat of organizational extinction is very real.26
For some time, academics, scientists, and market researchers have raised the alarm about AI’s impact. The dialogue had been fairly muted, largely constrained to the technology and scientific communities. Occasional press coverage of high-profile AI pessimists—like Vanity Fair’s 2017 article on Elon Musk’s “crusade to stop the A.I. apocalypse” —appeared, stoking the fire of AI backlash.27 By the early part of 2018, concerns about the consequences of AI hit the international stage as a main topic at Davos and in subsequent media coverage. Estimates and predictions suddenly took on additional meaning as mass media and pundits stitched together viewpoints on the topic. Dire scenarios such as a robot apocalypse, the elimination of all jobs, and the destruction of civilization hit the media.28
Most people with even a casual understanding of the state of AI today interpret this hand-wringing as wildly overblown.29 History shows the change of magnitude represented by AI is initially met with fear and skepticism, eventually giving way to mass adoption to become business as usual. But that does not diminish the cultural and social downsides. For some types of jobs, the impact will be stark. And the potential for human bias to distort AI analysis is real.
Many jobs will be lost. Retraining existing workforces to adapt will be a challenging public issue. The sooner governments and industry respond with thoughtful policies, the more stable the impact will be. At the same time, most economic AI value cases still require human front-line operators to execute. Certainly, middle managers and white-collar workers will typically work alongside AI in the near term.
As with other technology advances, AI will soon create more jobs than it destroys. Just as the internet eliminated some jobs through automation, it gave rise to a profusion of new jobs—web designers, database administrators, social media managers, digital marketers, etc. In 2020, AI is expected to create 2.3 million jobs while eliminating 1.8 million.30 Some of these new jobs will be in computer science, data science, and data engineering. Ancillary jobs at consulting firms will continue to grow, both at traditional firms like McKinsey and BCG and at decision science firms like Mu Sigma.
But the reality will likely be a much broader impact on society, and we cannot yet anticipate all the different types of jobs that will result from widespread implementation of AI. Speaking in 2017 about an e-governance initiative, India’s prime minister Narendra Modi said: “Artificial intelligence will drive the human race. Experts say that there is a huge possibility of job creation through AI. Technology has the power to transform our economic potential.”31
Positive viewpoints abound. With the global population growing to 9.7 billion people by 2050, AI-enabled agriculture could help farmers meet the demand for 50 percent more crops on ever-shrinking land resources.32 AI-powered precision medicine will identify and treat cancer much more effectively. Better cybersecurity will detect and prevent threats. AI-en
abled robots will assist the elderly, providing more independence and a better quality of life. AI will dramatically improve the ability to scan billions of online posts and web pages for suspicious content to protect children and others from trafficking and abuse. Climate change, crime, terrorism, disease, famine—AI promises to help alleviate these and other global ills.
I am familiar with a number of AI applications designed to deliver significant social value beyond their commercial impact: An AI application to predict the likelihood of opioid dependency so doctors can make better decisions in prescribing drugs and save millions of people from addiction. An AI application to predict public safety threats so agencies can better protect human lives. An AI application to detect money laundering so financial institutions can better combat the $2 trillion-a-year criminal industry.
In my meetings with organizations around the world, every enlightened business and government leader with whom I’ve spoken is actively working to understand how to harness AI for social, economic, and environmental good. We have hardly scratched the surface of what’s possible in improving human life and the health of the planet with AI.
The Battle for AI Talent
Because AI is now essential for every organization that wants to benefit from the analysis of large data sets, the competition for trained data scientists is robust. Consequently, there is a significant global shortage of AI talent today. Existing talent is extremely concentrated at a few technology companies like Google, Facebook, Amazon, and Microsoft. By some estimates, Facebook and Google alone hire 80 percent of the machine learning PhDs coming into the job market.33
Digital Transformation Page 13