Human + Machine: Reimagining Work in the Age of AI by Paul R. Daugherty and H. James Wilson. Harvard Business Review Press, 2018, 249 pp.
The simplest artificial intelligence (AI) implementation would be turning on a new system and instantly obtaining answers for all problems. Unfortunately, current AI developments, on average, match to toddler mentalities. Daugherty and Wilson, in Human + Machine, explore this challenge through proposing ways humans and intelligent systems can work separately or collaboratively to increase business value. The book’s first section explores independent work while the second section explores “the missing middle,” where humans complement machines or AI gives humans superpowers. Throughout, the work builds from a clear, five-principle framework: mindset, experimentation, leadership, data, and skills (MELDS). The authors superbly color in the missing sections, review market successes, and demonstrate a way forward for organizations hoping to include AI solutions.
The secret to adapting AI solutions to empower workers and create workplace superheroes lies in Daugherty and Wilson’s MELDS concept. The first element, mindset, means adopting the corporate culture to reimagine work around the missing middle. This happens when one moves behind machine automation to create a distinctive, action-oriented workflow. Modern designs favor DevOps (development and operations) principles where one concentrates on flow, feedback, and continuous experimentation to remove constraints. DevOps necessitates cultural changes accelerated by experimentation—the second element. Experimentation and DevOps 3rd Way arise from continuously looking for integration opportunities. The third element, leadership, sets the organizational tone and delivers goals to the other areas. The authors make a critical point in emphasizing that leaders also set the tone for ethical AI use. In describing the fourth element, data, the authors liken petroleum to information in the digital age, as it is a key factor to efficiently running operations while steering design and experimentation. The final element, skills, explores areas where new human-machine fusion skill types will be necessary to achieve desired outcomes. Every element appears constantly throughout the book, providing a smooth glide path for understanding.
The exploration of AI begins with how systems are being used today. Unfortunately, too many companies separate the human from the machine and limit their gains. The authors repeatedly use news articles as sidebars to highlight recent cases of AI use. Cases are divided into four sections: factory floor, corporate functions, research and development, and customer service. Automating factory floor operations has existed for years and assigns machinery to tasks too complicated, labor intensive, or repetitive for humans. Corporate functions and customer service can be blended into one category, starting with robots learning corporate processes and moving to service-directed chatbots communicating with customers. The book advocates that corporations should abandon massive call centers and move to machines politely communicating to point users to desired solutions regardless of how angry a customer may get. As odd as automated innovation sounds, the answer lies in employing processing power across many systems to interpret larger data sets and find, for instance, new medicines, improved manufacturing, and sales solutions. Seeking better answers leads to the next step, integrating man and machine to create empowered workers. This topic has been studied for generations, such as in Norbert Weiner’s 1948 book Cybernetics: Or Control and Communication in the Animal and Machine, detailing how his contemporaries believed machines could augment success.
The second section examines developing responsible AI, achieving new productivity levels, and revisiting the MELDS format. Responsible AI involves a learning model where humans and machines reinforce action with explain ability, accountability, fairness, and symmetry. Daugherty and Wilson emphasize transparency in the use of AI and reinforcing operational success with continuous experimentation. One example suggests that humans prefer lifelike robots, up to a point, and then fear and mistrust begin. Alleviating mistrust requires transparent algorithms.
Once corporations achieve transparency, productivity can be unleashed. The chapter “Super Results from Everyday People” stands as the best in the book. It references three types of augmentation. The first, amplification, envisions AI agents providing data-driven insights. One key example is the Elbo chair, where machine-interpreted parameters led to a more comfortable end design easier to make—and not envisioned by human designers. Collaboration allows AI agents to facilitate interaction with scalable services, such as tailored personal assistants like Alexa and Siri. Finally, embodiment seeks to improve humans as in Amazon warehouses, where robot appendages and automated carts expedite millions of deliveries daily. Expanded AI appears in a Mercedes example where automating processes disappears in favor of allowing every customer to invent a personalized car.
The last section highlights eight new skills before exploring the MELDS framework in businesses today. Fusion skills are areas where old work disappears to investigate interaction benefits. Of the eight, my favorite fusion skill is reciprocal apprenticing. This model uses an AI to train the human while using the human to train the AI. One would potentially see this skill in maintenance activities, where a machine might teach a human the proper repairs to make. Sometimes, things that humans perceive might impact outcomes, such as contextual references about usage or preferences that are not apparent to the machine. At the same time, the human benefits from the AI’s stored knowledge.
This work expertly explores the synthesis of man and machine. While Daugherty and Wilson skillfully lay out parameters, my biggest objection is that the work was not sufficiently technical. From a business perspective, many proposed options appear simplistic: buy the machine, train the algorithm, and watch the cash mount up. Many AI solutions are incredibly complicated; it requires years of practice to create the right algorithm and a sufficiently lightweight model to make the solution usable. One sees the expanded use of graphical processing units (GPU) to model tools in the academic world and industry. While it is possible to rent, borrow, or otherwise temporarily acquire these GPUs, the systems remain exceedingly expensive. The average programmer can build, run, and train a generic ML agent within 15 minutes today. Aligning those systems to precisely fit the right solutions, acquire the data, and deliver useful answers still requires a great deal of technical acumen.
Overall, Human + Machine offers a comprehensive look into how today’s business practices must shift to incorporate new technology. Moving beyond the 3rd Industrial Revolution of digital into a fourth revolution based on AI, ML, and on-demand capabilities requires setting a framework and actively pursuing goals. The MELDS system offers a way for businesses or other organizations to set goals and chart the path ahead. All improvements first require understanding where one is today and where the future path lies. Daugherty and Wilson set the tone, provide examples, and offer their vision for tomorrow.
A quick read and an excellent summary, I strongly recommend Human + Machine for strategic leaders, futurists, and those hoping to employ AI and machine learning (ML) tools to improve everyday and unique tasks in their organizations.
Dr. Mark T. Peters, Lt Col, USAF, Retired