The past decade has seen the emergence of computationally intensive “Machine Learning” (ML) approaches that are applied to very large datasets to identify underlying statistical relationships and derive better predictions. Over the past five years, “Deep Learning” (DL) systems have evolved that allow computers to “self-learn” sophisticated pattern recognition and intelligent response abilities. These are fueling disruptive applications across a range of cognitively complex areas including Natural Language Processing (NLP), radiology and medical diagnosis, digital personal assistants, autonomous driving, finance, and marketing.
Discussions about an evolving “Smart Machine Age,” a “Fourth Industrial Revolution,” and their consequent impact on the future of work are rampant. In their recent book “Competing in the Age of AI,” Harvard Business School professors Marco Iansiti and Karim Lakhani suggest that AI adoption is no longer a novel experiment but instead lies at the “digital core” of a modern successful business.
Acceleration of AI Adoption
From finance to health services, real estate to fast food, AI technologies are being rapidly adopted across industries and around the world to innovate business and operating models. Collectively, AI technologies have been predicted to increase global economic output by $13 trillion by 2029. Empirical evidence of the competitive benefits of AI are emerging. A study by Deloitte of 1,900 executives in seven countries found that 65 percent of respondents reported that their AI adoptions are providing competitive advantage.
The COVID-19 pandemic appears to have accelerated the pace of AI adoption by prompting businesses to seek innovations that reduce costs and circumvent the need for co-located employees and physically present customers. McKinsey has reported that the pandemic has sped up the pace of adoption of digital technologies by seven years, with many firms viewing their investments in AI as positioning them to be competitively stronger than they were before the pandemic.
Overview of the Los Angeles-Long Beach-Anaheim Metropolitan Statistical Area
To gain better insights into the scope, intent, and realized business value of AI adoption, the authors undertook a secondary analysis of published media reports about AI adoption in the greater Los Angeles metropolitan area, also known as the Los Angeles-Long Beach-Anaheim Metropolitan Statistical Area (MSA). With total Gross Domestic Product (GDP) in 2019 of almost $1.09 trillion and a population of over 13.2 million residents, the L.A. MSA is the second largest metropolitan area in the U.S. and the third largest metropolitan area in the world. Its economy is highly diversified, spanning construction, entertainment, financial services, healthcare, information technology, logistics, manufacturing, professional services, public sector, real estate, retail, and wholesale trade. Consequently, the L.A. MSA provides a representative region for studying AI adoption dynamics.
Our research identified a sample of over 50 instances of AI adoption in the region. Figure 1 provides a summary of categories of AI that have been deployed and the reported business intent of these initiatives. Reported instances of AI adoption span a wide range of functional areas and industry sectors.
Figure 1: Categories and intended focus of AI adoption
AECOM, an $18B per year engineering and construction services corporation, is reported to have incorporated machine learning algorithms in its in-house project management platform to achieve “100% accuracy.” BCBG Max Azaria, a worldwide fashion designer, reports implementing a third-party AI service in its marketing efforts to raise its online customer engagement by 20 percent and achieve a 300 percent increase in customer conversation rates.
Other examples of adoption include Amgen in the healthcare industry. Deloitte in professional services, and Northrop-Grumman in manufacturing and national defense. In the realm of ML, Warner Brothers is using decision-assistance software to determine whether to approve films in the greenlighting stage. Cali Group is using DL technology to “transform the restaurant and retail industries.” One of its companies, PopID, uses neural network-powered facial recognition software to enable contactless payment systems, whereby a customer can pay by simply looking into a camera.
Another Cali Group company, Miso Robotics, specializes in the development of robotics hardware for retail kitchen space. Amgen, a biomedical products and services company, is using AI in a variety of areas ranging from NLP applications to assisting health administrators verify compliance with procedures and regulations, to utilizing DL algorithms to help physicians determine whether or not a particular fracture is osteoporotic.
Kaiser Permanente of Southern California has adopted an intelligent Chat-Bot system that communicates with customers through mobile text messaging, assisting them with prescription medication refills. Figure 2 shows the distribution of the instances of AI adoption in the LA area by industry sector. Figure 3 shows the aggregated reported instances of AI adoption over the past 12 years and indicates an accelerating pace of adoption.
Figure 2: Adoption of AI by Industry
Figure 3 Aggregate adoption of AI over time
Business Benefits of AI
Table 1 provides a summary of the reported functional area focus of the AI initiatives and the reported realized business benefits at 26 LA area firms for which we found explicit mention of actual realized business benefits. To enable comparison with national and international patterns, we applied a framework used by Davenport and Ronanki in their analysis of 250 business executives at U.S. and international firms to classify the 26 instances of reported AI business benefits.
Table 1: AI Adoption examples and reported business benefits
|Company Name||Functional Focus Of AI||Reported Realized Business Benefits*|
|AECOM||Analysis||Capture “brings 100-percent accuracy to conventionally cumbersome, time consuming and subjective project management processes.”|
|Allied Universal||Early Threat Detection||20 percent reduction in safety and security incidents.|
|Amgen, Inc.||Analysis, Personalization, Early Warning Detection||Improved patient care, accuracy of diagnoses, intuitive healthcare that can be followed up with when patient is at home.|
|AT&T||Early Threat Detection, Analysis, Automation||Ability to manage 15 million alarms per day for customer incidents, 7 percent reduction in miles traveled for service workers and 5 percent increase in productivity.|
|BCBG Max Azria||Product Recommend||150 percent revenue increase, 300 percent increased conversion rate, 20 percent increased engagement.|
|Cali Group (Caliburger)||Automation||Robotic chef only costs $2000 per month on a subscription-basis vs. $4000 per month of a human, protect against high industry churn rate; increased safety through contactless payment during COVID-19.|
|Carnival Corp – Princess Cruise Lines Ltd.||Product Recommend, Personalization||Reduced cost of prospecting campaigns by 20 percent, tool used to develop wholistic view of each customer and automate personalized interactions.|
|Comcast – NBC Universal||Analysis, Automation||Trials showed 19 percent more brand memorability, 13 percent more likability, and 64 percent more message memorability.|
|Deloitte LLP||Analysis||3000 active users, over 100,000 documents examined.|
|Denim LA, Inc. (DSTLD)||Personalization||104 percent increase in YoY email revenue, 26 percent increase in orders.|
|Grant Thornton International, Ltd.||Analysis||In one case, reduced data prep time per quarter from 24 hrs. to 4 hrs., in another case, a complete review of decades of filings was performed in 2 days, which would have normally taken hundreds of hours over several weeks.|
|Hollar, Inc.||Product Recommend||46 percent increase in conversion rate, 24 percent increase in AIPO, 12 percent increase in avg order value, 8 hrs per week saved.|
|Kaiser Foundation Health Plan, Inc.||Early Warning Detection||Doctors were alerted of patients’ health deterioration before they needed an emergency transfer to ICU, reducing length of ICU stays and mortality rates.|
|Lawrence Berkeley National Lab||Early Threat Detection||Faster wildfire prediction and less false positives.|
|Los Angeles Financial Credit Union||Analysis||Reduced delinquent loans by 96 basis points in 12 months.|
|Moving Analytics, Inc||Customer Engagement||Got – and kept – people walking 60 percent more.|
|Northrop Grumman||Analysis, Automation||Reduced instances of rework, disruption, and down time.|
|Providence St. Joseph’s||Analysis, Automation, Product Recommend||For chatbots: more than 150,000 messages per day, over 20 percent engagement rate, served over 40,000 sessions since launch.|
|Reliance Steel & Aluminum Co.||Early Threat Detection||Ability to examine hundreds of thousands of datapoints, then draw conclusions, create and confirm/reject hypotheses, visualize data for better decision-making.|
|REX Real Estate||Operations||300 percent YoY growth.|
|Snap, Inc.||Customer Engagement||ML allows people to use filters that include themes from advertisers, which increases ad revenue through increased engagement and exposure to advertisers.|
|Tala||Analysis||90 percent loan repayment rate, boosted repayment rate by 14 percent with help from Boundless.|
|Target Corporation||Personalization, Customer Engagement, Product Recommend||15-30 percent revenue growth.|
|The Cheesecake Factory||Sales||Business improvements after 1 year of deployment, but results are not just AI but also online ordering in general, POS integration, and social media campaigns.|
|The Honest Company||Customer Engagement, Sales||19 percent increase in email conversion rate, 7 percent increase in user conversions, 10 percent increase in revenue per user.|
|Vimify||Customer Engagement||21 percent boost of diet and exercise engagement by app users.|
Figure 4 shows the range of business benefits reported by the L.A. area firms compared with those reported by U.S. and international executives. The most common AI business benefit that L.A. MSA companies reported is using AI to optimize both internal operations (40 percent of adopters) and external customer-facing business processes (35 percent of adopters). These levels are quite consistent with those reported by the national and international executives. These focus areas are laudable for their high potential realized value to risk ratio and are consistent with early applications of previous “breakthrough” digital technologies.
We found that 30 percent of AI adopters in L.A. reported the business benefit of enhancing the features, functions, and performance of products. Thirty percent of AI adopters also reported deploying AI to help make better decisions. In contrast with the concerns about potential massive displacement of workers by AI, only 5 percent of L.A. MSA firms reported using AI to reduce headcount. This is consistent with the earlier study of U.S. and international executives.
It is noteworthy that based on our secondary analysis of published media reports, L.A. region firms seem to be lagging U.S. and international firms in realizing AI business benefits across the other five categories shown in Figure 4. The following three cases provide a more detailed look into the scope, business intent, and realized business benefits of AI adoption.
Figure 4: Business benefits of AI Adoption
The Walt Disney Company
The Walt Disney Company (TWDC) is an international entertainment company headquartered in Burbank, CA. The company has been utilizing AI since the early 1990s. In their Disney Store locations, “expert systems” that replicate the abilities of expert decision makers have been deployed to assist store managers. Disney also uses robotic process automation to categorize product SKUs into the optimal categories in its online stores. In its filmmaking arm, TWDC is using “factorized variational autoencoders” (FVAEs) that use DL neural networks to examine audience reactions to determine instant and latent consumer reactions to various scenes and dialogue in movies. Facial recognition and DL techniques are used to assist employees with finding episodes with specific characters and actions. It is speculated that consumers will soon have access to search all of Disney’s content in this way, including ESPN, feature films, and television shows.
O’Melveny & Meyers, LLP
O’Melveny & Meyers is a law firm that was founded in Los Angeles in 1885. The firm now has annual revenues of over US$800m and an employee count of 1,500. It has adopted an AI-enhanced recruiting process by which it is able to assess its top performers and then cross check for those traits in its recruiting efforts. It does this by having all of its associates play AI-based games. The law firm then uses the insights gained through these games in an attempt to “completely remove bias” from the recruiting process, focusing on the results of the performers rather than any other data-point. Further, the company has subscribed to an AI tool to help analyze legal briefs and then recommend corresponding statutes, case law, and other relevant materials to the attorney. The AI-powered search tool is reported to save the average attorney 132-210 hours per year in time spent on legal research through the natural language processing and machine learning methods.
Reliance Steel & Aluminum Co.
Reliance Co. is a metal manufacturing firm that produces over 100,000 different metal products for over 125,000 customers around the world and is headquartered in L.A. Reliance has adopted ML capabilities in order to process sensor data from its machines to determine which equipment should have preventative maintenance and more broadly generate key insights about its manufacturing processes. This has enabled Reliance to draw conclusions, create and confirm/reject hypotheses, and visualize data in order for more informed decision-making.
Caveats to AI Deployment and Adoption
While AI technologies offer compelling potential business benefits, there are a number of significant challenges to AI deployment that business leaders need to carefully consider. These include data and algorithmic bias, limitations to domain- and context-specific applications, and potential unanticipated unintended consequences. Since historical data often reflects past biases and discriminatory practices, AI systems that are designed to learn from historical data are prone to inadvertently learning these biases.
The L.A. Police Department adopted the “Predpol,” predictive policing program that relies on advanced algorithms to determine where crimes are likely to occur in the next 12 hours. Since the program’s adoption, there has been heavy criticism by civil liberties and privacy groups, claiming that PredPol unfairly discriminates against black and Latino ethnicities. The LAPD has since discontinued the use of Predpol due to budget constraints brought on by the coronavirus pandemic, though other concerns may have contributed to the decision.
More broadly, AI applications continue to lack “general intelligence,” also known as common sense, which limits their reliability to the domains and contexts for which they have been trained and tested. This risk is particularly concerning when implementing “autonomous” AI-based systems that are designed to make decisions and act upon those decisions automatically. These risks must be carefully assessed and mitigated through careful choices of objectives, deliberate design, systematic training and testing, and extensive monitoring.
Moving Forward: Lessons and Caveats for Future Adopters of AI
To avoid missing out on potential competitive opportunities within their organizations that are enabled by AI technologies, business executives must be inquisitive, forward-thinking, and action-oriented. They must educate themselves about the main categories of AI that offer potentially important business capabilities and business benefits. They should encourage experimental pilot programs to both explore the potential business benefits of particular AI technologies for their firms and to start developing the technical, organizational and managerial competencies needed to effectively deploy these technologies and realize their potential business benefits.
Fortunately, there is a growing body of knowledge on best practices for adopting AI technologies. A common recommendation, consistent with previous generations of digital technologies, is that adoption should occur in phases, starting with simple tasks and expanding to more complex operations. For the near term, the best path to building more intelligent organizations will be to “blend technology-enabled insights with a sophisticated understanding of human judgment, reasoning, and choice.”
In conclusion, 60 years after the initial research in this area, AI technologies have finally evolved to a level of functionality that offers new business capabilities with significant potential for value creation. AI-enabled business capabilities promise significant benefits across enhanced speed, quality, consistency, scalability, and reduced costs. The potential is especially compelling for manual and cognitive work that is highly repetitive (dull), dirty, dangerous, or expensive (dear).
While much of the attention to date has been on the potential automation of manual work that falls into these categories, the potential opportunities to automate repetitive cognitive (“white-collar”) work are potentially greater. Given the significant share of cognitive and service sectors within the L.A. MS, this could have substantial implications for future employment and the nature of work in the region. Business executives and key decision makers must be forward-thinking in their approaches to evaluating the potential opportunities offered by AI technologies, and develop the talent and competencies (or partnerships) necessary to convert these potential opportunities to actual realized business benefits. Executives will also need to be careful to anticipate and mitigate potential unintended consequences.