Recommendations: Jumpstarting an AI-Augmented, Human-Centered Future of Work Recommendations: Jumpstarting an AI-Augmented, Human-Centered Future of Work

November 21, 2019

AI can’t work as a skilled teammate for humans, enhancing and complementing our talents, unless we choose to make that a priority. Here, Invested summarizes this issue’s key takeaways as this collaborative work begins.

Gabriella Chiarenza is managing editor for Invested and Regional and Community Outreach at the Federal Reserve Bank of Boston.

In this issue of Invested, we sought to learn more about AI, what it can and can’t do in the workplace, and what its path forward might mean for workers and businesses. We discovered that AI’s capabilities aren’t advanced enough that we should worry about a complete intelligent-machine takeover of our offices any time soon. But as we heard from those we spoke with for this issue, if we want to harness the transformative powers of AI in our workplaces in a way that is more equitable and productive and that spurs even more innovation, we have some work to do. AI does not have to be something humans should fear, and it shouldn’t be something we completely rely on. But it could easily be either or both of those things depending on who decides which way AI continues to develop.

Human abilities and characteristics remain as important as ever, even as intelligent machines learn to handle certain tasks as well as or better than we can. Our conversations in this issue suggest that it’s in our best interest as a society to treat AI as a tool that can be a skilled teammate, enhancing and complementing our human talents. It will only augment us rather than replacing us, though, if we choose to prioritize that approach when it comes to workplace uses, and develop the technology out that way.

If we leave AI to travel uninterrupted on its current path, we run the risk that its benefits will accrue only to those who currently hold the most control and power over technology. Yet subjecting AI to uninformed regulation or the backlash of a public that fears but doesn’t fully understand it would also be a mistake. We can use AI to the greatest and most innovative extent of its powers if we truly understand its capabilities and options for development. The conversation on AI must be more inclusive—and more informed. It will be important for researchers and technology developers to sit down with workers and businesses from all industries to discuss how AI can best be used in those environments. Working together, with review and assistance from ethicists and policymakers, these teams can then design, develop, and test specific approaches to AI teammates that usefully augment skilled humans of all kinds. Here are some of the key takeaways from this issue that we recommend keeping in mind as this collaborative work begins.

General purpose technologies (GPTs) like AI can definitely have an upside, but the transition isn’t easy.

AI is not the first GPT to come along: the internal-combustion engine and the personal computer are other familiar examples that emerged in the 19th and 20th centuries. But GPTs are still relatively rare, and when they do come into use, they have a tendency to change things dramatically. It’s important to remember that the job loss accompanying their uptake can happen much more quickly than worker reinstatement in other sectors or placement in new jobs that the GPT creates. Those who have studied previous industrial revolutions say that without complementary investments in worker training, transitions for workers can be very difficult or may not happen at all. The jobs GPTs help to create may require different skills, new credentials, and/or geographic relocation, which may be significant hurdles for those who are displaced. When a technology like AI is used solely for automation without thought to how it might be harnessed instead to complement human workers, its productivity benefits may be muted while the human costs may be devastating.

“In many situations, augmentation can customize AI’s powers to our own expert abilities to create a supercharged team that gets the job done better than either machine or human could do alone.”

AI can be a teammate, a tool, or both.

Adopting AI with an augmentation focus may prevent that bleak future. More importantly, the experts we spoke with suggest that augmentation may be the best way to squeeze the most benefit out of AI, because of the type of technology AI is. If developed in a way that recognizes what humans still (and perhaps always will) do better than machines, AI could complement us beautifully. It has the power to process huge amounts of information, venture into hazardous environments, repeat motions endlessly without complaint—all tasks that are currently dull, dangerous, or non-ergonomic for humans. Its abilities combined with our skill at reading emotions and situations, understanding social norms, handling physical challenges dexterously, and developing creative responses to problems can help us make more informed decisions and perform precise actions more quickly and efficiently. In short, in many situations, augmentation can customize AI’s powers to our own expert abilities to create a supercharged team that gets the job done better than either machine or human could do alone.

Technological hurdles remain, but AI continues to evolve.

AI has a lot of learning to do to help us become our best enhanced selves at work. Robots equipped with AI can struggle with physical motions that humans can often do without thinking, like opening doors, climbing stairs, and untangling cords. Navigating physical spaces with obstacles (such as human coworkers) is also difficult for robots without the assistance of sensors and cameras—which can then make the robot much more difficult to program and maintain. Even AI that doesn’t have a physical presence, like voice assistants, can’t always grasp the nuances of our requests well enough to meet our needs. Anyone who has ever received the response “I don’t understand” from automated customer service or an AI voice assistant can attest to the current limitations.

The technology will almost certainly improve, however, and will in time be able to deftly handle these and many other tasks. If anything, the fact that software developers and roboticists are still refining and expanding AI’s capabilities means we still have some time to work on equitable, beneficial augmentation applications—but the window to make those decisions is closing.

There’s a lot of misinformation out there.

No one knows for sure what it will look like when AI comes into a wider range of workplaces, but that hasn’t stopped a lot of people from speculating about doomsday scenarios. While it is important to be prepared for a future in which machines can do a lot more of the work that we currently think of as entirely human tasks, panicking about robot bosses won’t help us get there. The conversation around augmentation may not be as flashy as debates over whose jobs are going away, privacy concerns, or hacking, but it is just as important. Much of the public has never heard of augmentation and may not know about the alternatives of cobots and other machines that enhance and assist us, rather than taking over. When this information is lacking, the public only hears the sensationalist narratives—one side of a much more complex story being debated in technology and business circles.

We would all benefit from interdisciplinary conversations around augmentation.

As noted above, discussions around automation, augmentation, and the future of work need to include a lot more perspectives. Engineers, roboticists, and business leaders cannot and should not make decisions about how we will use AI more broadly at work on their own. Many are now asking for input from their colleagues in other fields like ethics, sociology, and economics. And to build out the powerful tools that augmentation-focused technologists want to make available, they will need assistance from many partners. For example, programs that help AI better understand human speech and intentions may need the guidance of linguists or psychologists, and robots intended to assist healthcare workers should be crafted with the input of experts in that field.

“When workers are involved from the beginning—when they have a chance to weigh in on what technological assistance would be helpful and effective, and when they understand how it will be used and how they will be affected—they are much more likely to be excited about AI than to fear it.”

Development and trials of new technology should also meaningfully involve workers. 

As we saw in this issue, workers are worried about what automation and AI will mean for them. A future that involves technology you don’t know much about and that could have unexpected impacts on your paycheck, your job stability, and your role and responsibilities understandably makes people nervous. But those we spoke with on the technology and business sides said that when workers are involved from the beginning—when they have a chance to weigh in on what technological assistance would be helpful and effective, and when they understand how it will be used and how they will be affected—they are much more likely to be excited about AI than to fear it. A key element of the augmentation design process is working closely with skilled practitioners in the field ahead of building the technology, while developing and testing it, and when deploying it, to ensure the priority of enhancing and helping human workers is upheld. Workers’ input and feedback is extremely valuable to roboticists and engineers as they build out products. Augmentation is more complicated than straightforward automation because it requires such a careful eye on all fronts to the interaction of humans and machines. But for that very reason, the technologies that evolve from augmentation may not only be more useful and productive in the long run, but also more trusted, embraced, and valued by the humans who work with them.

Resiliency is the skill of the future. AI can help with that.

Although augmentation aims to enhance rather than replace workers, people in all fields must be prepared for the possibility that their current occupations will change or even disappear as AI develops and more companies adopt this technology. In addition to learning more about AI and what it can do—we’ve included a link to a free online course that explains the basics in our resource section below—workers will need to become more and more resilient. Gone are the days of an employee working in the same role at the same company for an entire career. We will all need to become more flexible and adaptable as technologies evolve, which will also help us be better prepared for other unexpected challenges, such as financial shocks or policy changes. Workers will need to assess their own skill sets, think about adding more if that will make them more adaptable, and consider how skills used in their current roles might actually qualify them for work in a completely different sector.

AI can actually help us on this front. It has the ability to process huge amounts of data, and we provide plenty of that data every time we update our online work profiles. Using this data, AI may be able to help us connect with companies hiring employees with skills like ours. Creating more of these platforms for people in a wider range of fields and with different experience levels, and making related training and technology available at employment assistance centers for those without access to a computer or the Internet, could help ensure more people are swept along in new roles rather than left behind.

AI can also help with on-the-job training—the type of upskilling that workers prefer to classroom or external programs. AI powers augmented reality (AR) and virtual reality (VR), both of which are becoming more commonly available. For newer workers training to step into processes that involve expensive, dangerous, or complicated components or systems, both AR and VR would give them the chance to practice the process before they have their hands or eyes on the real deal. This kind of technology could make apprentice-style training more feasible in different environments and allow workers to experiment with different applications of their existing skills or building out new ones.

“It is important that at least some funding for research and development be focused on a neutral exploration of augmentation, which can help spur further innovation toward uses that prioritize stability, equity, and growth in the human job market as well.”

Neutral, supportive, and flexible investment in research and development will be necessary to make augmentation a reality.

The exciting AI developments occurring in university labs, start-ups, and private companies don’t happen in a vacuum. The funding and the clients, as well as current and potential future regulation, all affect that development. This is why it is important that at least some funding for research and development be focused on a neutral exploration of augmentation, which can help spur further innovation toward uses that prioritize stability, equity, and growth in the human job market as well. In the past, public programs and agencies played this role and helped to bring us such revolutionary developments as the Internet and nanotechnology. While similar programs around augmentation would require a careful balance between innovation, regulatory checks, and ethical and public reviews, they may be crucial in making effective and wider-reaching augmentation-focused AI a reality.

Businesses of all industries and sizes will need incentives and support to try out and maintain augmentation-oriented AI.

AI is still very expensive, particularly for smaller businesses and those that operate with tight margins. Augmentation-focused systems require a lot more knowledge, programming, and maintenance than the industrial robots widely used today, so they can be even more expensive. Companies that want to keep or expand their human workforces while adding augmentation technologies will need financial and technical support to do so. Business incentives could help ensure that these up-front costs are neither a financial barrier nor a major hassle for workers already busy with other elements of the work process. Staff education, cross-training, and new, specialized support staff who may be needed to keep the systems running should all be considered in developing such incentives.

Time to pick our lane on AI is waning, so it’s time to start the conversation.

AI isn’t ready for prime time yet in most environments, but it’s learning fast, and its capabilities are growing all the time. If new perspectives and different voices aren’t brought into the fold now, it’s entirely possible that AI will end up serving only the narrow segment of society that is already benefiting from it. Educating the public and policymakers, facilitating conversations across boundaries of practice and geography, and integrating AI technology into our workplaces takes time, so we must start now. Creating a community of workers, scientists, practitioners, policymakers, community members, scholars, and others who want to prioritize technology that augments us rather than replacing us is a great start. We hope this issue of Invested gets you thinking about how you might join the conversation around making AI more equitable, rewarding, and productive for workers and businesses of all kinds.

The views expressed are not necessarily those of the Federal Reserve Bank of Boston or the Federal Reserve System. Information about organizations, programs, and events is strictly informational and not an endorsement.