Today we are in the early stages of building a new infrastructure for work; think of it as a new operating system for creating value and getting things done. A combination of next-generation networking, distributed computing, and artificial intelligence (AI) is laying the groundwork for this transformation, catalyzing the emergence of a worldwide digital coordination economy. In this economy, algorithms are being deployed to identify and match those in need of something with those who can fulfill their needs, including both human and non-human agents.
The impact of automated production is a big part of this story. We have seen this particularly acutely in manufacturing. One extreme example here is the early emergence of “lights out” manufacturing facilities like the one maintained by the robot manufacturer FANUC in Japan. Factories that run “lights out” are fully automated and require no human presence on-site. Thus, these factories can be run with the lights off. Automation, however, has increasingly moved from manufacturing to services and knowledge work. Legal analysis is increasingly done by machines, and companies like Narrative Science are building software to automatically convert raw data into stories that are sometimes hard to distinguish from those written by actual journalists.
A second important development, parallel to automation, is the widespread shift toward greater connectivity. As the world becomes more digital, objects, machines, and systems embedded with communication, sensing, and processing power will compete with the human workforce to drive new growth and create new kinds of value.
The implicit potential for using automation to grow prosperity is tremendous. With sufficient demand, automation facilitates the increases in productivity and efficiency that are fundamental to driving economic growth. But it also increases the risk of human dislocation. This is a familiar historical pattern. For example, while the Industrial Revolution greatly increased global prosperity over the long term, it also caused dislocation of rural populations. Design can—and must—make the difference.
The emergence of companies like Uber, Upwork, Honor, and many other coordination platforms that rely on algorithms to directly match human consumers and producers is just the first stage in the evolution of the new system for value creation. Many next-generation platforms could take humans out of the production role by delivering autonomous vehicles to people who need rides, for instance. From there, it’s easy to envision a system where economic value is often created entirely without humans. In this capitalism of things, smart objects and systems will exchange value and trade services autonomously.
How? While the automation of production and supply has been discussed quite a bit in recent years, the automation of demand may turn out to be an even more important, if much less explored, development. It is possible someday, for example, that self-managing and even self-owning autonomous vehicles may form “corporations” unto themselves. Using “smart” contracts, they may begin to incorporate themselves, seek investors, and pay dividends. If this sounds fantastical, consider that it is estimated that three-fourths of trades on the New York Stock Exchange are now automated, and that a computer recently taught itself to play the game of Go and beat the reigning human champion, something once thought nearly impossible for computers to do. The future is coming fast.
These self-managing systems may also invest in the deployment of algorithms that help them decide which services to provide when: commuter rides during the premium-fee rush hour, food delivery in the evening, and emergency medical package delivery late at night. They may even trade their algorithms or data with other self-managing vehicles—for a fee. Importantly, these vehicles would also contribute to aggregate demand by buying fuel, marketing themselves, and ordering their own repair services as needed.
As software takes an increasing role on both sides of transactions—ordering and producing—it promises to bring vastly more efficient coordination to these kinds of basic economic functions. This emerging digital coordination economy, with its efficient matching and fulfillment of both human and nonhuman needs, has the potential to generate tremendous economic growth.
However, as software engineers essentially author a growing segment of our economic operating system, it may take deliberate design choices in platform architecture, business models, new civic services, and public policy to prevent this increasingly seamless “coordination economy” from becoming highly inequitable as well. Already the growth of on-demand work has allowed investors and owners in some industrialized regions to reap substantial financial returns while many of the people using platforms to generate income streams are struggling to maintain their standard of living. Uber drivers, for example, have seen a drop in earnings in the United States over the last couple of years, even as the company continues to grow at a dramatic pace.
It is clear that the fundamental technologies driving the coordination economy are neither “good” nor “bad,” but rather offer a heady combination of opportunities and challenges. In order for society to thrive in this future, we will need a new design paradigm—a socio-technical framework in which the economic growth and societal benefits of an increasingly coordinated economy can be maximized. Such a paradigm could encompass: the technical design of platforms, regulatory frameworks necessary to both protect against inherent negative externalities and help distribute opportunities on a more equitable basis, efforts to foster the creation of new ecosystems of services, and public policies that support inclusive prosperity. Perhaps most importantly, it tries to create the most human value out of the big technological shifts that are advancing in stride. Let’s take a look at these shifts before we tackle the principles of such a design framework.
In the 1960s Paul Baran, a co-founder of the Institute for the Future (IFTF), proposed moving from a centralized communications infrastructure, where all signals flow through central switches, to a highly distributed system in which data are broken into small bits (packets) that move through distributed relay nodes in an ad hoc and emergent manner.
This model became a foundational ingredient of the Internet: a highly distributed communications infrastructure that has become intricately woven into every aspect of our lives. For the past 40 years, we’ve been putting this technology infrastructure in place, and today, it has become a kind of nervous system for our daily lives, our economy, and our culture. Over the next decades, this same infrastructure will connect more and more objects and machines as well as people—creating the so-called Internet of Things. But when we connect everyone and everything, the result is not just to make things faster, better, and bigger. Distributed communications make it possible to do things in completely different ways.
The iconic example is Wikipedia, a vast knowledge resource created with only a few hundred actual employees and millions of contributors from around the world. The distributed infrastructure of the Internet has also enabled investigative, data-based journalism, including deeply collaborative and cross-border efforts to uncover stories of global crime and corruption, as with the Panama Papers. It has engaged thousands of people in helping solve the mysteries of protein folding by collaboratively playing games on platforms such as Foldit.
This is a phenomenon we call socialstructing—creating value by aggregating micro-contributions of a large association of volunteers using networking tools and technologies to create value. It is the atomization of work, the breaking down of complex tasks into smaller and smaller pieces that can be performed collaboratively by more and more people (or things) and then aggregating them, often in new ways, to create new value.
Today, socialstructing is already reinventing media, education, science, and many other economic, civic, and social endeavors as the first steps in the evolution of the digital coordination economy. The Panama Papers would not be possible if 400 journalists from around the world couldn’t come together and use digital tools to collectively analyze an unprecedented amount of raw data—1.5 million documents totaling almost 3 terabytes! No one existing news organization could do this alone. In a different arena, Moon Zoo invites the public to help astronomers count and analyze craters and boulders on the surface of the moon. They collectively analyze images from NASA’s Lunar Reconnaissance Orbiter Camera and in the process help speed up the process of discovery. The next arc of network expansion will only reinforce the trend and take it beyond human agents.
At IFTF, we have already begun to experiment with the building blocks of automated organizations. A few years ago, in its quest to explore what happens when networking meets AI, IFTF developed prototype software informally dubbed iCEO. As the name suggests, iCEO is a virtual management system that automates complex work by dividing it into small individual tasks and then matching these micro-tasks to qualified workers registered on multiple online work platforms, such as Upwork, as well as various email and text messaging platforms. The IFTF team challenged itself to see if a piece of software could manage the process of writing a research report, a task that is intellectually challenging and core to many knowledge work business models. After a few practice runs, the team was surprised to find that the software could meet the challenge.
The process involved developing a “recipe” that divided complex tasks into a number of simpler ones and used a bit of machine learning and AI to find and match people in various on-demand platforms to complete different tasks. For instance, to create an in-depth assessment of how the material graphene is produced, iCEO asked workers on Amazon’s Mechanical Turk crowdsourcing platform to curate a list of articles on the topic. After duplicates were removed, the list of articles was passed to a pool of technical analysts from Upwork, who extracted and arranged the key insights from the articles. A different Upwork cohort then turned these insights into coherent text, which went to another pool of subject matter experts for review. Finally, iCEO passed the paper to a series of Upwork editors, proofreaders, and fact-checkers. While humans were involved throughout the process, the hiring and management was all turned over to software that was able to handle the task of coordinating individual contributions much more fluidly than a human manager alone could manage.
In this particular experiment, iCEO routed tasks out to 23 people from around the world. It generated 60 images and graphs, as well as text, in a print-ready format. The IFTF team rarely needed to intervene, even to check the quality of individual components of the report as they were submitted to iCEO, because quality assurance and human resources were already automated. The high-quality result—which typically would take several weeks to produce—took only three days. And now that the recipe has been created, it can be reused on the same types of projects many times over.
The iCEO software is just one example of what happens when the resources of easily accessible networks are combined with the power of automation. In many cases, it enables us to do previously unthinkable things on seemingly impossible schedules. In his book The Inevitable, Kevin Kelly, the noted futurist and technology observer, argues that the next generation of startups will be building just this kind of future—adding AI to every task. Today’s businesses are already adding AI to virtually every part of operations, from marketing to business development, communications to HR. They are using logic similar to what the IFTF team used to create iCEO: Design a recipe for an activity by dividing it into smaller steps and then use a combination of task-routing tools and AI to distribute work across large networks of people, carefully matching skills to the task and then sequentially adding layers of tasks to complete the larger job. The result is greater speed, reduced costs, and more thorough results. This is the horizon drawing near.
Designing for Inclusive Prosperity
The digital coordination economy, augmented with AI, promises to put economic growth on steroids by much more fluidly matching supply and demand.
However, it also poses significant challenges to the markets and models that have shaped human culture for the past century. The coordination economy is certainly at odds with traditional notions of jobs: Its core logic is the atomization of work, with the ability to divide larger tasks into smaller ones and then distribute them across a wide network. Atomized and networked labor, in turn, creates blueprints for further automation. With Uber, for example, automated routing is used to bring drivers and passengers together and then to plan their route. The driver is really just controlling one step in a larger chain of tasks, and it’s no coincidence that Uber’s CEO Travis Kalanick is on record saying that the use of human drivers is just a transitional step before fully automated vehicles hit the market. As distributed computing and AI continue to mature, markets will only tend to reinforce this logic.
The broadened American middle class of the last century emerged from a massive unionized workforce and a large middle-management tier, and both were strengthened by grand political bargains like the New Deal and Great Society. But there is no a priori reason to assume that any of these economic supports will necessarily be carried over into the context of a digital coordination economy. If this is the case, then the American middle class could be in a position analogous to the taxi-service owners confronted with Uber and Lyft: a group that has benefitted from the legacy structures, social conventions, and regulatory provisions that have historically given them an advantage, but have not well positioned them for the new era. The emerging coordination economy is unlikely to be friendly to the middle class by accident.
However, as a society, we can support broader prosperity by design through interventions at multiple levels:
Code for positive platforms:
The starting place is to build so-called “positive platforms”—platforms that embed protocols and practices designed not only to deliver profits to their owners and investors but also to help maximize incomes and create positive work conditions for those whose livelihoods depend on them. In a paper entitled “10 Strategies for a Workable Future,” IFTF synthesized the views of more than 60 diverse experts to highlight a spectrum of design choices regarding platform ownership, transparency, privacy, sustainability, marginalization, identity, collective bargaining, portable benefits, learning, new skillsets, and a “good work” code. We need to explicitly consider such choices in the development of every platform and algorithm. On most platforms, for example, online reputations serve as avenues to more and better work opportunities. On such platforms, it is important to make it transparent how such reputations are created and what workers can do to improve them.
Create integrated ecosystems of services:
In a digital coordination economy, every platform and algorithm operates in a technical ecosystem. But these innovations also exist in an ecosystem of secondary and tertiary services that were designed for traditional employment patterns. Today, there’s a mismatch between the historical ecosystem of worker services, such as skills training, career counseling, collective bargaining, and even employee discounts, on one hand, and the needs of on-demand and platform workers—for example transparency, stability, and basic fairness—on the other. We need to foster development of these ancillary services with the same kind of in-depth user awareness that has guided the design of user interfaces in computing and mobile devices.
Build out the twenty-first century social safety net:
Our social safety net was designed for an era in which most workers could rely on benefits from large formal employers like GM and US Steel. Although currently only a small percentage of people rely on incomes from the “gig economy,” this share is expected to rise substantially in the next ten years (with freelance workers growing to nearly half of the workforce by 2020, according to some estimates). Nascent conversations about portable benefits, a universal basic income, and cooperative ownership of platforms are steps in the right direction, but a more holistic approach is necessary. As a nation, we need to engage in a large-scale conversation about what the new social safety net should look like, given changes in work, ongoing advances in health and medicine, growing caregiving needs, and the changing educational and training requirements of a digital economy. What can we imagine for ourselves ten to 20 years from now, and what can we do today and tomorrow to get there? Can we imagine providing caregiving benefits given the realities of a rapidly aging population?
New opportunities for value creation, such as Uber-style transportation or Feastly-style food service, are already bumping up against regulations that never anticipated the potential for crowdsourced services, the dislocation of workers from the traditional labor market, or massive data-tracking for workers and consumers. At the same time, the new digital labor market presents a host of new risks that regulations can certainly mitigate. As a part of rethinking our regulatory infrastructure we need to ask questions like these: What do we need to regulate for, and what do we need to stop regulating for in the digital coordination economy? Do we need the current slew of professional and trade certifications in a world where people’s reputations and skill levels are easily visible on many platforms? On the other hand, do we need to regulate at the level of algorithms, since so many of these serve as filters for earning opportunities? Do we need to worry about personal data collected on many work platforms and how it is being used? Networks and AI have their unique affordances and potential for discrimination that are different from formal organization—for example, there is evidence of systemic racial biases in customer feedback on platforms—so what are sensible remedies for these biases?
The emerging digital coordination economy brings with it the seeds of great economic prosperity. However, large swaths of the population may not reap the benefits of such prosperity unless we embark on a large-scale effort to purposefully and thoughtfully design for it. We need to bring to this endeavor the best of technological expertise, but also the best thinking from disciplines such as economics, political science, governance, and others. The stakes are high and the time to start building the socio-technical infrastructure that will ensure equitable prosperity in the digital coordination economy is now.