People & Machines: Partners in Innovation
13 de junio, 2019
Tiempo estimado de lectura: 15 minutos
The greatest impact of intelligent technologies won’t be from eliminating jobs but from changing what people do and driving innovation deeper into the business.
Thoughtful adoption of intelligent technologies will be essential to survival for many companies. But simply implementing the newest technologies and automation tools won’t be enough. Success will depend on whether organizations use them to innovate in their operations and in their products and services — and whether they acquire and develop the human capital to do so.
In a recent Deloitte survey of 250 executives familiar with how their companies are thinking about intelligent technologies, nearly three out of four said that they expected AI to substantially transform their organizations within three years. Of course, the workforce will be deeply affected by all this change. Yet even as AI eliminates some jobs in the coming decade (it most certainly will), it may create as many positions as it kills and open up vast new opportunities for collaborations between humans and machines. Earlier talk of large-scale job loss has subsided somewhat. In the Deloitte survey, for example, reducing head count through automation was the lowest-ranked objective for AI — only 7% of the respondents selected that as their first priority. Indeed, many observers are shifting their expectations away from job loss to job change, as humans find ways to work closely with machines.
Given the likelihood that many jobs will change rather than disappear, organizations need to understand the new skills required. In a recent McKinsey survey of executives at companies with revenues of more than $100 million, 66% of respondents said “addressing potential skills gaps related to automation/digitization” within their workforce was a “top 10 priority”; 64% of the U.S. respondents and 70% of the European respondents said they needed to retrain or replace at least a quarter of their current workforce. Significantly, just 16% of the business leaders responded that they were “very prepared” to address potential skills gaps, raising serious questions about their readiness to compete. Other recent surveys suggest that the high expectations executives have for intelligent technologies exceed their skills and experience in integrating such technologies into their companies.
Although we have observed and worked with many large companies and startups on AI issues, we know of very few that have begun significant job redesign, re-skilling, or retraining programs. Moreover, most individuals aren’t being adequately re-skilled or retrained for automation-enabled work. (See “About the Analysis.”) Smart organizations will take steps not just to adopt intelligent technologies but also to recruit and retrain people for skilled roles, redesign tasks and jobs, and use AI as an enabler of innovation in products, processes, and business models in what we call innovation based on intelligent automation. This will be a job-by-job, task-by-task transformation, but we can already see places where major advances in technology are being undermined by insufficient attention to integration and human capital. Surgeons, for example, are increasingly using robotic technology to assist them in routine surgery; the new technology provides better vision, more-precise incisions, and neater sutures. However, few hospitals and medical schools have developed effective approaches for training surgical residents on the technology; surgeons in training get no hands-on experience.
Although the potential for AI-enabled innovation exists in virtually every aspect of business and society, it is largely unrealized. A study of internal audit organizations, for example, found that less than one-third of the audit teams had road maps for incorporating new technologies. Technology vendors are conceiving and producing innovations ranging from self-driving cars and trucks to the “self-driving enterprise,” but very few would-be adopters have begun to envision how AI will change jobs in their companies and what new skills must be developed. Because many new AI technologies are appearing now or will be here in the near future, organizations have no time to waste in planning for them and creating work design innovations that parallel the technological innovations.
A Spectrum of Intelligent Automation
When intelligent technologies support individual workers, allowing them to do their jobs better or more efficiently, what we’re really talking about are tools rather than automation. A good example is a taxi driver who uses GPS for driving directions. Automation goes a step further: It allows tasks or processes to be carried out without human assistance or participation, but humans may supervise the work or perform adjacent or complementary tasks. For instance, intelligent diagnostic systems can read X-ray images, but radiologists are still needed to define the imaging to be performed, relate imaging results to other medical records and tests, discuss findings with patients, and perform other activities.
Although the earliest applications involved manual and systematic (structured and repeatable) cognitive tasks, we are moving toward nonsystematic cognitive tasks that include creativity and job variability, which until recently seemed beyond the scope of automation. And we are progressively adding greater autonomy to products and services. (See “AI at Work.”)
In fact, we are beginning to see autonomous systems that can perform tasks without any human involvement at all, using carefully prescribed guidelines. Consider automated financial trading. Because it depends entirely on algorithms, companies can complete transactions much faster with it than with systems relying on humans. In a similar fashion, robots are performing narrow tasks autonomously in manufacturing settings. In 2015, for example, Changying Precision Technology, a Chinese company involved in the production of mobile phones, replaced 90% of the workers in one of its plants with robots. In doing so, the company says it was able to more than double its output and slash defects by 80%. More commonly, however, AI and robotics change jobs rather than eliminate them. Amazon, for example, has hired more than 300,000 people since its 2018 purchase of Kiva Systems, a maker of warehouse robots. One distribution center employee, who “babysits” several robots and ensures they have bins to load, commented on her job: “For me, it’s the most mentally challenging thing we have here. It’s not repetitive.”
While today most AI systems augment only existing workers, many people believe it’s just a matter of time before complex systems will be able to operate by themselves in unstructured and dynamic environments. For example, in the next two or three years we will have self-driving vehicles capable of operating in limited spatial areas or under special circumstances (classified by the Society of Automotive Engineers as Level 4); by 2030 or so, many anticipate vehicles that operate without human intervention at all (classified as Level 5).
Furthermore, there is the growing possibility that in the not-too-distant future we will have machines that can operate according to their own goals. An example that’s as immediate as it is frightening is autonomous weaponry that will be able to decide where, when, and against whom it uses its capacity for destruction. This application represents the negative Mr. Hyde aspect of fully autonomous systems. But eventually, we can also expect to see a Dr. Jekyll side, with applications that have the potential to make life better.
Minds Working With Machines
Just as semiconductors enabled us to reduce the cost of calculations and apply arithmetic to new areas — first to scientific and military applications and later across all professional and social spheres — innovation based on AI will unleash an avalanche of both improved and entirely new products and services. The impact on the world of work will be unprecedented.
Human versus machine matchups in chess illustrate how humans will need to continuously change their roles relative to smart machines. Back in 1996 and 1997, IBM’s Deep Blue competed against world champion Garry Kasparov and became the first computer ever to beat a world champion in a six-game match. As with other chess programs, Deep Blue’s strategy blended computing power and strategic knowledge of the game provided by human experts. People could sharpen their skills by playing against it and studying its moves, but they wouldn’t learn anything new, per se. But now the competence of chess programs has risen to the point where many chess masters use them to improve their own level of play. At the end of 2017, a new chess milestone was achieved when AlphaZero software, developed by Alphabet’s DeepMind, learned how to play solely on the basis of its knowledge of the rules. In less than one day of playing against itself, AlphaZero learned enough to crush Stockfish, which had previously been the leading chess program. Among chess experts, one of the most surprising things about AlphaZero is that it has learned strategies that extend beyond how humans play. Humans taught Deep Blue to play chess, but AlphaZero developed its own approach — one that humans could learn from.
Such changes in the human-machine relationship will emerge in the workplace, too, as AI becomes increasingly intelligent. It will not be a spontaneous process but will be induced by designers and users of intelligent technologies and, of course, by companies that innovate on the basis of such technologies and have the right human resources in place to make it happen. However, major changes in jobs and skills don’t coalesce overnight, even when the approach involves hiring new employees instead of retraining existing ones. Once companies identify the needed changes, implementing them will take time.
In the future, organizations will need to place both adoption of technology and human capital development at the center of their innovation strategies. As time goes on, how companies deploy technology and human capital will have a tremendous impact on their competitiveness and their very survival. We see four basic scenarios playing out in the organizations we have worked with:
1. Minimal investment in automation technology and people. For a variety of reasons — including cost and lack of vision or knowledge, especially among executives — some companies delay making the kinds of fundamental decisions and commitments that will make them viable AI innovators in the future. In this scenario, they underinvest in the necessary technologies and human capital. Such reluctance to enact changes will inevitably lead to a loss of competitiveness and an inability to maintain a sustainable business. These companies will have higher labor costs, fewer intelligent products and services, and lower levels of customer service than their competitors. In wealth management, for example, companies without intelligent robo-advisers are already losing business to competitors such as Vanguard and Charles Schwab that offer low- or no-cost advice.
Take chatbots, which many companies are using to handle relatively simple customer service tasks. Starbucks, for example, uses chatbots to notify customers when their orders are ready; Mastercard uses them to make it easy for customers to get information on their transactions. (For more complicated problems, human agents typically take over.) To the extent that such companies reconfigure jobs or processes and help workers learn how to work with the technology, the chatbots can provide synergies, or at least a better distribution of tasks. Unfortunately, automation doesn’t always work this way. For example, in 2017, Tesla invested heavily in robots for manufacturing and underinvested in skill development for human workers. When it realized that the robots weren’t doing enough to help the company meet ambitious production goals for its Model 3 cars, management backed away from its reliance on robots and hired and trained humans to perform the necessary tasks. But for the final vehicle assembly, Tesla took a more nuanced, integrated approach, assigning humans to the complex tasks and using robots for specialized tasks such as moving goods around the factory, lifting heavy components, and testing seats. The result was, as one observer put it, “a delicate dance of human workers and robots on the production line.”
3. Incremental changes in jobs and skills with little investment in intelligent technologies. Many companies that prioritize incremental process improvement (for example, using Six Sigma or “lean” programs) don’t invest enough in new technology — in part because the methods don’t include a role for technology. In addition, it can be difficult to adopt broad, cross-organizational changes in jobs and technologies at the same time because the impact of AI and other technologies on jobs tends to be specific to particular jobs. Although it’s true that hiring and retraining skilled workers can generate short-term improvements, that approach alone won’t lead to meaningful change. Indeed, we have found that unless companies are willing to commit resources to AI technologies, they risk falling behind competitors in both productivity and quality. Eventually, moreover, they hurt their ability to hire and retain quality knowledge workers, who may see better opportunities elsewhere. Of course, there are particular settings in which an emphasis on people-oriented strategies makes sense. High-end restaurants, for example, are less dependent on automation than are fast-food establishments. The same goes for fashion and other luxury businesses. But even in these cases, intelligent automation should have a growing presence in back-end functions and processes such as supply-chain management and customer support.
4. Significant investment in both intelligent technology and human capital innovation. Organizations with a broad-based investment approach are best equipped to pursue innovation in both AI application and human capital development. Rather than simply looking at automation as a way to cut costs, these companies create innovative products, services, processes, and business models by implementing intelligent technologies, redesigning jobs, acquiring new skills through hiring, and training their existing workers. This approach is especially vital for companies that compete in markets dominated by global giants.
For example, GE — notwithstanding its current difficulties with its GE Power and Genworth Financial business units — is actively trying to use both AI applications and human capital to drive innovation. One way it is doing this is by studying the needs of different types of employee users, or personas, and then considering how they might be supported by technology. Personas are part of a widely used approach for understanding customer needs in marketing and product development, but they are rarely used for the development of internal systems and even less so to create AI systems.
One of the GE personas is made up of employees involved with buying or sourcing industrial materials. A key task for these employees is to ensure that the needed materials are available on the manufacturing line at the right time. Historically, they relied on their intuition to manage the delivery schedules, but machine learning models have the ability to learn from past deliveries and provide model-driven estimates. Users are being trained to understand how the models work and how they can be improved. Today, the models inform the sourcing manager, who makes the final decision about when to order. Eventually, GE expects the AI systems will be capable of making decisions on their own to optimize things like delivery schedules and in-process inventory. The role of humans will be to tweak the processes and address problems that occur.
Despite the power of AI and other new technologies, the likelihood that they will replace managers and professionals in the near term is minimal. Rather, many observers, including Erik Brynjolfsson and Andrew McAfee, codirectors of the MIT Initiative on the Digital Economy, believe that the change will be more gradual — that those “who use AI will replace those who don’t.” In our view, the challenge for companies is finding ways to ease intelligent technologies into their organizations, while simultaneously determining how to take advantage of what intelligent humans have to offer.
Think Before You Automate
There is no simple recipe for successful innovation based on automation. Different companies will have different opportunities to put intelligent technologies to work. However, in researching knowledge and technology transfer within companies and advising organizations on AI adoption, we have developed a set of guidelines:
Start with management education. The best starting point is to invest in training for the executives charged with making the strategy decisions about intelligent technologies. Based on our experience, executive ignorance often leads to two opposite but equally negative behaviors: If leaders underestimate the potential of these technologies, their companies will miss opportunities to benefit from them. On the other hand, if they overestimate it and initiate projects that are too ambitious and costly, they will waste resources and perhaps even generate a bias within the company against new projects, even those that are reasonable. To prepare leaders to make future decisions, a leading property and casualty insurance company, for example, held daylong sessions for top executives on what AI is, how best to manage it, and what it might mean for employees. Anthem Insurance Companies, a large health insurance corporation, and Bank of America have run similar sessions for their leaders and board members.
Develop a road map for future initiatives involving technology and people. As with any project, implementing an intelligent automation initiative requires having a road map that describes the objectives, the necessary resources, and the implementation schedule. A good road map should help the organization anticipate the potential benefits beyond the most obvious ones and should include a communication strategy, both internal and external, especially when intelligent automation projects might lead to a reduction in jobs. For example, Situm Technologies, a Spanish startup (of which one of us, Senén Barro, is a founder), developed technology that accurately tracks the location of people and assets via smartphones inside facilities such as hospitals, airports, and factories. The initial applications were fairly narrow — an early customer in the building-security business wanted to track the routes of its security guards. Eventually, however, the company developed a road map for using Situm’s technology within facilities in other ways — for example, to manage people during emergency situations such as fires or assaults. This enabled the company to offer a set of solutions that aligns the benefit of optimizing human resources with safety.
Focus on immediately valuable projects and be wary of initiatives that are too ambitious. Companies that lack significant AI experience should focus initially on low-hanging-fruit projects that will enable them to gain experience. Highly ambitious projects to treat cancer, provide individual investors with detailed investment recommendations, or eliminate drivers from cars have all either failed or taken far longer than researchers expected. Even Amazon has had challenges with its Amazon Go stores, and its drone delivery project is taking a long time to emerge.
Combining several manageable projects in a single business area often has a better chance of yielding significant results than trying to pursue one big one. At Amazon, for example, CEO Jeff Bezos says that many of the company’s investments in machine learning are focused on “quietly but meaningfully improving core operations.” If the company’s strategic focus is on using AI to enhance customer relationships, for example, the component projects might include chatbots or intelligent agents to answer questions quickly 24-7, machine learning models to capture the “voice of the customer” from call center operations, recommendation engines to pitch promotions only to customers with high interest, and so forth. This incremental approach also creates more time to redesign work and re-skill workers, since each AI-supported task will typically require only incremental change in jobs. The objective should be clear; even in cases where the goal is automating tasks previously performed by workers, key workflows should be designed or redesigned, focusing on the division of labor between humans and smart machines. The aim throughout should be innovative and effective work design, not just cost reduction.
Fuente: MIT Technology Review
Te puede interesar
It’s been a little over one year since I first published a list of 35 Latina tech founders who had raised more than $1 million in funding.leer más
Most corporate leaders understand that businesses have a key role to play in tackling urgent challenges such as climate change.leer más
A mere 20 years ago, the internet was in its infancy. Nowadays it seems the world is essentially built on connectivity.leer más