As we organize human work and learning at fantastic scales, let’s strive for something inspiring.
Through our relationship with machines, humans are building a new, synthetic intelligence. Organizations are the home for this part-human, part-machine intelligence. The topic to which we now turn is the role organizations play in coordinating the immense scale of work and learning needed for this new intelligence.
The Functions of All Organizations: Learning and Work
All organizations have two common functions. Whether a multinational corporation, a mom-and-pop grocer, for-profit, or nonprofit, all organizations learn, and all do work.
The role of organizations in coordinating work is obvious. We have plenty of evidence of coordinated hunting, gathering, and trading from the rise of the human species. Egyptian, Greek, and Chinese civilizations pioneered remarkably sophisticated state institutions for coordinating human work at mind-boggling scales of complexity. By the Middle Ages in Europe, guilds, banks, and joint-stock corporations catalyzed a shift toward more commercial, market-driven approaches to organizing work. Organizations are, in fact, largely synonymous with work for most of us. When I say that I am “heading to work,” I mean both that I am going to do work and going to the place where my organization allows me to do it.
The Rise of Learning Organizations
What is less clear is the link between organizations and learning. Learning organizations are more than just schools and research labs. That’s because all organizations learn. Just as organizations tap people’s contributions of work, they also tap their contributions of knowledge. Sometimes that knowledge remains solely in the heads of employees and sometimes it is explicitly embedded in knowledge stores such as patents, databases, and policies and procedures.
As the field of organizational learning has grown, we’ve come to understand a lot more about how knowledge is created, transferred, and retained in organizations. The best-known book on the subject is Peter Senge’s, The Fifth Discipline. Senge was deeply inspired by W. Edwards Deming’s work on Total Quality Management, largely because of its focus on iterative learning. By the 1990s, many companies were using Deming’s work on continuous improvement (“Kaizen” in Japanese) to build feedback loops into their workflows to learn from and boost their performance over time. Many of these techniques were picked up by Silicon Valley startups, and turned into an obsession with data-driven, high-speed learning. The most successful of these companies learned to combine the explosion of data from the Internet with big data analytics to drive their organizational learning through their ongoing engagements with end users. In time, companies like Google, Amazon, and Facebook began to center their competitive strategy on this new form of organizational learning.
Automating Work and Learning
Today, these companies are revolutionizing organizational learning by automating it with machine learning. The result turns the way we do work on its head.
Work and learning have always been closely intertwined since the latter increases the efficiency and effectiveness of the former. Learning to fish with a net saved our ancestors time and increased their fishing yields. Today, machine learning algorithms are doing the same for information workflows, industrial automation, and countless other forms of work. Perhaps less obvious is the way work creates feedback for learning. The enormous streams of data flowing from end users working with the automated interfaces of Google and Amazon generate an invaluable feedstock for machine learning. Today, these companies pay nothing for that data.
The term increasingly used to describe this synergy between machine learning and automation is “intelligent automation.” Its effect on workflows and learning is fourfold: It increases scale, speed, and quality and decreases marginal costs. Intelligent automation may increase employment within market disrupters like Amazon and Google, as these firms hire highly educated technical and entrepreneurial staff to develop and run these technologies. For the existing companies whose markets are being disrupted, employment is often decimated. The net effect is a highly concentrated uptick in automation developers, coupled with a larger decrease in operating staff spread out across a range of established competitors. Automation thus increasingly eliminates employment through market consolidation, while concentrating it into fewer, and better-educated, hands.
Lowering the Cost of Organization
Economist Ronald Coase saw coordination within organizations (in-house work) as more efficient than markets (outsourcing) for most recurring workflows. It’s more efficient for you to write the weekly report you’re already familiar with than for your boss to put it out for bids each week. In contrast, the more efficient markets are at handling these “transaction” costs, the more it makes sense to outsource work to them.
Coase’s insights on the cost of organizing people are very important in a world of intelligent automation. Technology helps organizations outsource work by driving down the cost of coordinating with the market. Email and other digital communications thus played a critical role in coordinating contingent staffing, domestic outsourcing, and the offshoring of manufacturing and call center jobs to China and India in the 1990s. But the biggest boost to coordinating outside contributions of work came the following decade with the rise of platform businesses.
Shrinking the Cost of Work and Learning: The Platform Advantage
Coase saw that lower internal transaction costs enabled companies to grow the scale of internal operations, while lower external transaction costs increased outsourcing to the market. Platforms fuse business model and technology to simultaneously lower internal and external transaction costs. Today’s best-known examples, firms like Amazon, Google, and Uber, serve as a nexus for coordinating contributions from suppliers and end users. Platforms squeeze out the cost of attracting external contributions of work and learning, effectively blurring the line between organizations and their markets. The result is that platform organizations operate at phenomenally large scales as they draw on vast pools of suppliers and end users. General Motors and other industrial giants may once have tapped the contributions of hundreds of thousands of employees. But firms like Amazon, Google, and Facebook regularly engage millions and even billions of contributors.
This powerful cost advantage in lowering coordination costs sets platform organizations up for an even more powerful operating advantage. By pushing more work to suppliers, such as Uber contracting rather than employing its drivers, these firms avoid the cost of providing benefits to these workers, which typically amounts to 30 percent of total compensation costs. Even more astounding, however, is the way platforms coordinate volunteer contributions of free work and learning from their end users. We upload our work in the form of videos, pictures, product and restaurant reviews, and a variety of other forms of “user-contributed content.” We also do work for these platforms every time we serve ourselves without the assistance of any employees. The data from our engagement with the automated interfaces of these platforms fuels the machine learning now accelerating the explosion of intelligence in these platforms. And we do it all for free.
A Choice in How We Use the Power of Platforms
Platforms create insurmountable advantages by building feedback loops between automation and machine learning. It is no coincidence that today’s platform operators also lead the world in machine learning. Platforms automate stakeholder engagement and generate enormous quantities of learning feedback in the process. This new form of automated organizational learning is then used in a number of ways.
One way builds on the idea of continuous learning mentioned above. Data from our interactions with products and services is used to make them more useful. As these offerings improve, they attract more users, which increases the flow of data, learning, and improvements at larger and larger scales. As these platforms grow, the feedback between automation, machine learning, and stakeholder engagement results in an upward spiral of intelligence within a particular market. Ridesharing gets better at getting us places. Social media keeps us better in touch. Search platforms get us faster and better information. Online shopping gets faster and more convenient.
Each of these services can be extremely helpful and has the potential to genuinely improve our quality of life. But they don’t always do that; in fact, they often have undesirable side effects. This suggests that these platforms are not learning solely to continuously improve our end-user experience. As Shoshana Zuboff notes in her book, The Age of Surveillance Capitalism, much of the intelligence gathered by platforms trades the needs of shareholders against those of stakeholders. We become products, packaged and sold to the highest bidder, to make Google and Facebook shareholders enormously wealthy.
The Code within the Code
This leads to a final point, which is the need to understand the objectives undergirding organizational platforms. These immensely powerful organizing technologies connect humans and machines on scales unlike anything we’ve seen. They are the housing for a new synthetic intelligence, part human and part machine, that is revolutionizing work and learning on this planet. These systems represent an important jump in humanity’s ability to coordinate our work and learning at revolutionary scales. They could be awe-inspiring and wonderful as tools for addressing pressing societal challenges and helping us live to our highest potential. And yet, they are falling far short of this potential. The question is why.
All technologies are encoded to achieve some purpose within a particular context. Organizations shape that coding through bylaws, mission statements, and strategies, as well as culture, relationships, and other social norms. This “code within the code” deeply impacts what technology does and how it is used. To understand how it is that we seem to be squandering the true potential of these new technologies for organizing ourselves, we need to examine this underlying coding.
The business world is largely driven by a code within the code that sees maximizing returns for shareholders as the primary purpose of business. This “shareholder primacy” is what drives Facebook’s willingness to allow political lies on its service and what makes Google extract far more personal data from us than they need simply to improve their services. Until relatively recently, I saw shareholder primacy as the gravest challenge we face in governing technology. But the growth of China’s technology sector and simultaneous rise of global authoritarianism points to another extremely dangerous virus in the code within the code: the drive for power and control.
In subsequent installments in this series, we’ll look at new possibilities for the code within the code. The organizations we build over the next few decades will house the future of synthetic intelligence. They will be the medium through which humans and machines coordinate themselves in work and learning. And we must do everything we can to ensure that they are pointed in the right direction.