How AI makes work better

Increasingly, knowledge work requires a blend of human and artificial intelligence. How will your organization meet this challenge?

Artificial intelligence increases productivity of labor

Early next year, millions of Bank of America account holders will get access to a new personal banker named Erica. She’ll check account balances, make money transfers, pay bills, and perform dozens of other tasks. Erica doesn’t get tired or rude, and she’s on call 24/7.

As you probably guessed, Erica is a next‑generation chatbot, or automated conversational agent. Unlike predecessor bots that primarily answered questions, Erica can perform a variety of basic tasks in response to voice commands.

True, you could handle most of those tasks by navigating the B of A website, using its mobile app, or walking into a branch. But that’s the point. Erica will make everyday banking easier, faster, and more efficient. Michelle Moore, Bank of America’s chief of digital banking, claims that Erica is designed not to replace jobs, but to help employees streamline complex tasks, such as building better relationships with customers.

In recent years, the emergence of intelligent agents like Erica has sparked fears that AI will replace human workers. This isn’t new. Forty years ago, for example, the rise of ATMs sparked fears that bank tellers were all headed for the breadline. In reality, as research by economist James Bessen has shown, the number of bank teller jobs actually increased after ATMs became ubiquitous in the early 1990s.

So what happened? The number of tellers required per branch fell because of ATMs. That reduced the cost of opening a new branch, with the result that banks rushed to increase their footprints. Overall growth in branch count drove aggregate job growth among tellers and other branch employees.

Annual change in productivity since 1950

Worker productivity has tailed off significantly since the tech boom of the late 1990s, despite mass adoption of cloud, mobile and social technologies in recent years. Some economists believe AI will reverse the trend and boost U.S. labor productivity by as much as 35 percent by 2035.

At the same time, teller job descriptions changed from dispensing and receiving cash to more complex services such as taking applications for lines of processing mortgage payments. Recent forecasts see more sophisticated bots and ATMs leading to fewer tellers, but that decline will likely be offset by increased demand for roles such as financial planning and personal account service. The U.S. Bureau of Labor Statistics predicts that teller jobs will decline by 8% from 2016 through 2026, while personal financial advisor jobs will grow by 14% over the same period.

Economists remain divided about the extent to which AI and related technologies will replace human workers. In a 2017 survey of economists by the University of Chicago Booth School of Business, 26% of respondents believed AI technologies would substantially increase the ranks of the long‑term unemployed in advanced countries. Another 24% were uncertain, and 18% thought AI would boost overall employment. And a recent study published by the National Bureau of Economic Research predicted that robots would eventually replace between three and six jobs per machine, with the burden falling more heavily on lower‑wage workers.

No matter where the dust settles on net employment between man and machine, it is the jobs themselves—and the nature of the work—that will be changing soon. For example, a recent McKinsey study found that fully half of all current work activities can be automated by adapting technologies that exist today.

Source: ServiceNow, State of Work Report

In many cases, intelligent assistants like Erica will help knowledge workers focus on more valuable and rewarding tasks. As MIT economist Erik Brynjolfsson (co‑author of The Second Machine Age) noted recently, “the most effective rule for the new division of labor is rarely, if ever, ‘give all tasks to the machine.’ Instead, if the successful completion of a process requires 10 steps, one or two of them may become automated while the rest become more valuable for humans to do.”

Take Amazon, for example. Three years ago, the company had just 1,400 robots operating in its warehouses. Today, it has over 50,000, yet its hiring levels have stayed the same. By and large, the machines aren’t displacing human workers. Instead they unburden employees from robotic parts of the job—running products back and forth from shelf to shelf—so they can focus on more valuable tasks.

Machine intelligence can help knowledge workers escape drudgery. In ServiceNow’s 2017 State of Work survey of executives at companies with at least 500 employees, respondents reported spending 16 hours per week in manual, repetitive tasks such as answering email and requesting support services. The survey also found that nine in 10 skilled employees spend too much time on low‑value manual tasks.

Mark Purdy, chief economist with Accenture Research, sees an employment symbiosis emerging. “With machines to do more of what I call the three Ds—dull, dirty, and dangerous work, this could free people up to do more interesting and fulfilling jobs, which could lift participation rates,” he says.

Adds Purdy: “There is also strong potential for AI to lift labor participation by promoting more flexible forms of working and bringing work to previously excluded groups or regions. People in remote areas with few traditional employment opportunities could in the future work in virtual offices or virtual factories, without ever having to leave their home region.”

Parallels in history

The blending of human and machine intelligence in knowledge work is part of an upward cycle of productivity that started with the Industrial Revolution. For example, farming once dominated the labor force, accounting for 38% of all U.S. employment in 1900. Then came the tractor and the internal combustion engine, along with better fertilizers and new irrigation technologies.

Today agriculture accounts for just 2.6% of the labor market, yet agricultural output has soared. Even as the amount of land and labor used in farming has declined steadily over the decades, total farm output more than doubled between 1948 and 2015. In 1930, the average American farmer could feed four people. Today, the same farmer can feed 155 people on average. Writing in the first AI Index Report, published in December 2017, Udacity founder Sebastian Thrun explains the history of modern agriculture this way: Technology simply “freed up 98 percent of us to find different jobs.”

With AI and other technologies emerging now, many believe a similar cycle of long‑term productivity will begin again, with technology replacing many manual tasks, freeing us up to create economic value in other ways. Case in point: Seventy‑five percent of the U.S. labor force now works in offices. Yet a significant portion of 21st‑century office work remains highly repetitive, even at the top tier or the org chart. As Thrun sees it, “AI technology can learn the patterns in our repetitive work, and help us do work faster.”

Think augmentation, not automation

What applies to chatbots and bankers also applies to cardiologists—and hundreds of other knowledge work roles across the labor force. By applying advanced machine vision and deep learning techniques to MRI imaging, the San Francisco‑based startup Arterys has built a system that automatically calculates how much blood is coursing through the human heart. That calculation might take nearly an hour for the typical cardiologist to do manually. Arterys can handle it in 15 seconds, and with greater accuracy. Arterys won’t replace cardiologists, but it will give these busy doctors more minutes to spend on higher‑value tasks such as interacting with patients or performing surgery.

“With this new revolution, I predict we will enter an era of unprecedented human creativity.”
Sebastian Thrun, founder, Udacity

The promise of general-purpose technology

We can expect to see the number of skills where AI beats humans increasing rapidly as new AI systems evolve and learn to perform more complex work. Because AI is a general‑purpose technology—just like electricity, the steam engine, and the combustion engine were in previous centuries—it will fuel similar waves of complementary innovation. Accenture estimates that AI could increase economic productivity by 40% and double economic growth rates by 2035.

How will companies capture their share of that growth? McKinsey cautions in a recent report that even though the U.S. government is forecasting steady overall employment growth over the next decade, employers must prepare to retrain large portions of their workforces to enhance skills for the more complex tasks that will remain after AI has subsumed rote tasks.

“Most organizations are still at an early stage of AI adoption,” says Accenture’s Purdy, “but it is likely that significant organizational changes will be needed to harness the full benefits of AI—in terms of human resources, finance, data management, and leadership.” The AI‑enabled company of the future, in other words, is clearly a work in progress, but one that will significantly alter our definitions of work itself.