Defining artificial intelligence

There remains a lack of clarity about what AI is—and isn’t— even among those who are eager to harness its potential. Our work shows that companies which are leaders in digitization are also those most able to deploy AI (see box below, “Our study on AI as the next digital frontier”). But where do you draw the line between “traditional” digitization, data analytics, automation, and AI? For many of our conversation partners, the link to machine learning is critical, with AI simulating cognitive behavior and able to teach itself. One example: an algorithm that identified and analyzed the core characteristics of families of malware, which it then used to teach itself to identify emerging new strains.

The data conundrum

Data are the lifeblood of AI. To extract true value from the data requires huge quantities of it, but where will that data come from, and who should own it? Participants in our conversations with experience of data sharing describe a world of considerable suspicion between those who have the data and those able to use it for AI purposes. In some cases, where it is shared, only the most basic raw numbers are disclosed, without context or clarity about what they represent. Since AI experts are looking for patterns and outliers, that may be all they need.

Our study on AI as the next digital frontier

At the Viva Technology Forum, the McKinsey Global Institute presented its new discussion paper, Artificial intelligence: The next digital frontier?

Key insights:

• Tech giants such as Google and Baidu dominate investment in AI, accounting for about three-quarters of the total. We estimate the big tech firms spent $20 billion to $30 billion on AI in 2016. Investment by venture capital, private equity, grants, and seed investments is also growing and totaled $6 billion to $9 billion.
• AI adoption outside of the tech sector is at an early and often experimental stage. Only 20% of C-level executives aware of AI say they use any AI-related technology at scale or in a core part of their business, according to our survey of 3,000 C-level executives across 10 countries and 14 sectors. Only 12% of 160 AI potential applications we reviewed are deployed commercially.
• Adoption patterns illustrate a growing gap between digitized early AI adopters and others. Sectors that are highly digitized, such as tech, telecom, and financial services, are also leading adopters of AI, and have the most aggressive plans to expand use.
• Use case studies we conducted in retail, electric utilities, manufacturing, health care, and education highlight AI’s potential to improve forecasting and sourcing, optimize and automate operations, develop targeted marketing and pricing, and enhance the user experience.

The opportunity

Participants in our conversations saw the smarter exploitation of data using AI as a significant opportunity, and provided some anecdotal examples. Yet actual adoption in core operation is still at an early stage in many cases; our research suggests only about 20% of firms aware of AI are early adopters. Others are experimenting, or still watching (see exhibit). AI early adopters report profit margins that are 3 to 15 percentage points higher than their industry average.

Main areas of application

AI has applications across a broad spectrum of sectors and activities, and our conversation partners highlighted main areas of application including:
• Predict key patterns, ranging from potential maintenance needs to future demand
• Promote sales by anticipating customer needs
• Prevent or detect fraud and waste

Specific use cases

There was no shortage of anecdotal evidence that AI can and does make a difference. Among the examples:
• Detecting fraud in ATM usage, by finding discrepancies between data on cash demands processed and cash actually dispensed
• Using algorithms to replace thousands of hours of legal services
• Voice recognition in banking call centers
• Increasing capacity in urban subways by reducing the gap between trains
• Predictive cardiology
• Reducing waste in restaurants by analyzing quantities of food served and eaten
• Multiple applications in aerospace and defence sectors
• In finance, accounting, and administration, taking over a substantial share of back-office functions

A major barrier: Talent

AI remains at an early stage, and in their conversations with us, our AI-aware guests pointed to several barriers to its growth. The complexity of easing the usage and sharing of data on the one hand, and protecting stakeholders from cyberattacks and misuse, on the other, was frequently cited. For companies, adapting organizational structures and business processes to optimize AI adoption will also be challenging. For now, perhaps the biggest barrier is a lack of people with the skills to leverage the emerging technologies and the insights that AI can deliver; the large-scale investments by tech giants are partly aimed at recruiting top talent globally. There is a shortage not only of data scientists, but also of translators who can connect the AI solution and analysis to some kind of business value. Hardest of all: finding people who can transform operations to actually make use of those insights and value.

Difficult policy questions

Even before AI adoption becomes widespread, our conversations underscored the often complicated policy implications. They include: Data ownership, and liability for it: If a self-driving car has an accident, who is liable? The manufacturer, the OEM, the insurer, the algorithm developer, or the client himself? Social and user acceptance: In an AI world, how do you define human responsibility and judgment. “Will a pilot be allowed to bypass what the computer recommends or decides,” asked one participant. Employment: What will be the impact on jobs? Could AI break with the long-standing pattern, dating back to the Industrial Revolution, that technology creates more jobs than it destroys? Privacy: In a world where data is broadly shared and exploited, how does one safeguard privacy?

A new world of possibility is dawning

For all the hesitations and concerns that we heard, our conversations underscored the excitement about artificial intelligence and how it could transform business productivity and return on investment. The next steps seem clear: acquire a greater understanding of the technologies, recruit the talent who can leverage them, and start experimenting. A new world of possibility is dawning