Summary: Working with AI: Real Stories of Human-Machine Collaboration by Steven M. Miller and Thomas H. Davenport

Recommendation

The future of work is already here. Sectors as diverse as public transit and advertising are using AI today. In this eye-opening book, information technology experts Steven Miller and Thomas H. Davenport examine examples of successful workplace AI integration. They go beyond statistics to paint a detailed picture of the current effects of AI on the way people work and highlight questions about the future. Whether you are concerned about AI’s effects on your own work or considering spearheading AI adoption within your organization, you’ll find valuable insights in Miller and Davenport’s text.

Take-Aways

• AI automation involves rules-based systems as well as machine learning systems.
• The majority of sponsors interviewed believed in a model of AI adoption that increased the productivity of workers, rather than replacing them.
• AI is pushing deeper integration between business and IT processes, creating a demand for hybrid roles.
• Frontline workers’ deep knowledge of their roles is essential for successful AI integration.
• AI is having diverse effects on demand for entry-level workers.
• There are still many things machines can’t do.

Summary

AI automation involves rules-based systems as well as machine learning systems.

AI engineers build rules-based systems using multiple if-then statements. They train machine learning systems on labeled data. The machine learning systems are able to detect patterns in this data that they then deploy to interpret unlabeled data. In the workplace, AI performs numerous functions, including:

• Predictions
• Recommendations
• Rankings (such as for sales leads)
• Finding specific information in a document
• Process automation

“Many people today work with AI on a daily basis. We found this happening in big companies and small, in offices, in factories, on farms, and across a wide range of knowledge and administrative work tasks.”

AI automation and augmentation in the workplace is already widespread and observable in many workplaces today. At Morgan Stanley, AI provides personalized recommendations for products and programs based on a client’s specific profile and portfolio. The financial advisor decides on the relevance of the prompts based on their personal knowledge of the client. This allows for more efficient communication with clients, since AI can effectively tailor recommendations to each client to a degree that human advisors would not have time to do. Advisors report that, with the help of the AI system, they are able to give more personalized advice to a higher volume of clients.

At Jewel mall in Singapore, an AI system now partially automates the process of filing incident reports for the security team. The AI pre-fills reports with relevant data and makes it easy to include additional information such as photos and video from the CCTV system.

The majority of sponsors interviewed believed in a model of AI adoption that increased the productivity of workers, rather than replacing them.

The sponsors of an AI adoption initiative are its driving force in the modern workplace. Typically, senior managers define the vision that shapes the changes to business processes and sponsor new training for employees.

There are, broadly, two effects that automation can have on jobs: replacement of human work through full automation or augmentation of human work through partial automation. So far, augmentation has been the dominant trend. In 2018, Deloitte conducted a survey of US executives. 63% of those familiar with AI said they would replace workers with machines to save costs, but this was not executives’ main motivation for adopting AI. Executives were more interested in improving products or internal operations, helping in decision-making and freeing workers to focus on creative tasks.

“If you’re worried about the impact of AI on jobs, it should be good news that humans in many different roles are required to build, deploy, operate, and sustain these systems.”

DBS Bank, the largest bank in Southeast Asia, implemented AI to support their anti-money-laundering analysts. The AI helped with the more monotonous aspects of analysis, freeing analysts to more effectively study emerging threats. Instead of working toward the eventual replacement of employees by AI, the executive driving the change at DBS was interested in helping human workers to do their jobs better. This goal created a safe environment for employees working with AI; it encouraged them to give feedback, contribute knowledge and collaborate with AI.

Leading economies are facing an aging population and a workforce shortfall, so firms are likely to start making more use of automation to make up for a lack of supply, rather than to replace employees. While the overall number of workers needed by firms using augmentation may be lower, in many firms studied, the workforce did not shrink and even continued to grow thanks to business growth – which was itself due to AI-driven productivity increases.

Effective implementation of AI systems relies not just on new technology but also new business models, business processes and worker skills. While AI will replace or augment many roles, it also creates new ones involved in planning, designing, deploying, monitoring and improving AI systems.

AI is pushing deeper integration between business and IT processes, creating a demand for hybrid roles.

Historically, people in business and IT roles have not understood each other’s work. Business roles cover activities like human resources, marketing, finance and management. IT roles cover the creation or configuration of the IT systems people use, as well as data-driven roles like data scientists, analytics specialists, AI/machine-learning engineers and data engineers. There is now, arguably, an additional knowledge gap between AI teams and other IT professionals.

Cross-functional roles bridge the gap between IT and business knowledge. Sometimes this bridging involves high-level coordination, like the product manager for AI systems and services at Shopee, an e-commerce platform in Southeast Asia, who ensures alignment across the organization and finalizes multi-stakeholder decisions. Sometimes it means multidisciplinary teams more specifically focused on governance, compliance or ethics. The Salesforce ethical AI practices team performs outreach to get individuals thinking about the ethical implications of their decisions and supports other teams trying to solve specific ethical questions, such as avoiding bias in the data sets used to train AI.

The need for these specialized roles to act as a link between two knowledge sets that remain mostly separate only further underscores the gap that remains between most individuals in IT and those in business roles. It appears that this is beginning to change, however. Now, companies integrating AI tend to have at least one person in a business role who is deeply focused on solving business problems with data and technology. Although their formal training is in business, these people have learned enough about the technology they are working with to become conversant with IT professionals. This trend will likely continue as it is a natural result of the penetration of IT processes into every aspect of business.

“Many people in IT departments now have business backgrounds rather than technology ones.”

Additionally, many businesses now expect their IT specialists to have a firm grasp on the business problems they are working to solve. For example, online styling service Stitch Fix requires that their data scientists learn how to style clients. In-depth understanding of how stylists employ IT’s technical solutions allows the head of data science to better weigh the effects of her programming choices.

Frequently, ambitious people see an opportunity to integrate an IT or AI solution in their organization and learn the skills necessary to do it. Jennifer Schmich at Intuit, a provider of financial software, started as a copywriter before seeing a chance to start using Writer.com, an AI-powered writing assistant. Schmich designed her role as content architect with her boss and now heads a small team that can coordinate rules like style guides and standardized language for thousands of writers.

Competitive businesses have to embrace data. This means that the number of hybrid IT-business roles will only increase. Forward-thinking companies insist that IT professionals gain exposure to the work of their business units and vice versa.

Frontline workers’ deep knowledge of their roles is essential for successful AI integration.

The professional knowledge of frontline workers is essential in order to effectively integrate AI into the work they do. They are often called upon to evaluate AI suggestions or output, making their professional judgement more important than ever. For this reason, training employees who are not already very experienced in their role to work with AI could represent a challenge.

Frontline workers interviewed saw their role as essential for evaluating machine suggestions, integrating big-picture thinking and coordinating with other people. Many stated that AI reduced drudgery and made their jobs more intellectually stimulating. However, some found that the increased emphasis on communication and intellectual work made their jobs more demanding. In every case study carried out for this book, employee productivity increased with AI integration.

“Going forward, system design efforts, deployment efforts, and ongoing post-deployment operational support efforts will continue to work out better with participatory inputs and strong support from the frontline workers.”

Tasking individuals from across the organization to not just adopt automated solutions but to help design them allows for a fine-grained approach. Japanese international advertising and public relations agency Dentsu’s robotic process automation team found that the company, like many knowledge-intensive workplaces, didn’t rely on large-scale, easily automated business processes. Instead, the team found a long list of micro-tasks specific to the work of individuals, so they reached out to employees from across the company. They provided them with training and a tool that let them design their own automation routines for their repetitive tasks. Doing so saved around 3500 hours of work.

AI is having diverse effects on demand for entry-level workers.

In some industries the trend is to hire fewer entry-level workers. This is clearly the case for routine physical work such as burger flipping or weeding, as well as very clearly defined, routine mental work like visual inspections for quality control. As a result of AI integration, Haven Life/MassMutual now has a lower need for entry-level insurance underwriters because they tend to perform tasks the company has automated. They still need experienced underwriters for their ability to evaluate the output of the AI system and deal with non-routine cases, however.

On the other hand, AI-powered training can actually lower the barrier to entry for some jobs. For example, the PBC Linear machine shop uses an AI-supported augmented reality training system called Taqtile to fast track manufacturing workers’ training. Taqtile works with an augmented reality system that uses AI to adapt to each user, achieving a level of precision necessary for machine-shop training. The system provides a level of personalized training that would have previously required one-on-one time with an instructor, letting new hires learn at their own pace.

“The very same technology… can both reduce opportunities for entry-level workers through productivity increases and expand entry-level work opportunities through enhanced levels of embedded training, guidance, and performance support.”

AI systems can also help by improving job access for groups with capability disadvantages. Dentsu partnered with AutonomyWorks, which specializes in creating job opportunities for people on the autism spectrum and addresses companies’ repetitive process needs, to create special work tools for autistic employees.

The diminishing number of entry-level job opportunities is a systemic problem for the economy and society. While it remains to be seen what effect the widespread adoption of AI will have, the loss of entry-level positions in knowledge-based sectors is a serious threat.

There are still many things machines can’t do.

Human judgement remains an essential component in most work augmented by machine learning:

• AI can’t understand context – Programmers can’t capture context in data sets used to train machine-learning algorithms, nor can they capture such a broad concept in algorithmic rules. As a result, AI can’t use data to tell a coherent story, frame a problem, make subjective judgments, or consider the broader social and ethical implications of its actions.
• Complex systems also represent a stumbling block for AI – Many human systems are too complicated for AI. This means that AI can’t distinguish important alerts from unimportant ones in complex settings, like a public space under security surveillance. It can’t negotiate or coordinate decisions between groups with different or evolving priorities, nor can it persuade individuals to adopt new behaviors or remove organizational obstacles to drive organizational change.
• AI is still unable to understand emotional situations– Despite popular depictions of AI systems developing relationships with humans, like those in the movies Her and Ex Machina, AI can’t understand emotional needs, build relationships with humans, foster job satisfaction or employee moraleor analyze the tone of written communication. Even the Writer AI tool has trouble analyzing tone, and AI systems used for social media analytics are not good at recognizing sarcasm.
• AI is not self-reliant – AI still relies on human assistance to set up the physical systems or environments needed to capture data for analysis, fix AI when it malfunctions and transfer knowledge from human experts to AI systems.

For all these reasons, humans still need to have the final say when it comes to AI-generated recommendations.

“One of the great advantages of having humans and smart machines working alongside each other is that humans can confirm that an automated decision is ‘sensible.”

It’s important that humans working with machines understand the business processes the AI is trained to facilitate, so that they grasp the reasons for AI decisions and are able to determine their adequacy. Similarly, AI systems need functionality to provide explanations for the decisions that they make to human workers, so that humans have the information they need to evaluate those decisions. For example, DBS Bank’s anti-money-laundering system includes a dashboard that explains the risk scores generated by the AI. Explaining AI decisions also encourages workers to buy into AI integration in their roles.

About the Authors

Thomas H. Davenport is a Professor of Information Technology and Management at Babson College and a senior AI advisor to Deloitte Analytics. Steven Miller is Professor Emeritus of Information Systems at Singapore Management University.

Review

“Working with AI: Real Stories of Human-Machine Collaboration” by Steven M. Miller and Thomas H. Davenport offers a comprehensive and insightful exploration of the practical applications of artificial intelligence (AI) in various industries. It presents a collection of real-life stories and case studies that highlight the collaborative relationship between humans and AI systems.

The book begins by establishing a solid foundation of understanding AI and its capabilities. It explains the different types of AI technologies and their potential impact on the workforce and society. The authors emphasize the importance of embracing AI as a tool for augmentation rather than replacement, promoting the idea that humans and machines can work together to achieve superior outcomes.

Throughout the book, Miller and Davenport share numerous examples of successful human-machine collaborations in different domains. They showcase how AI has been integrated into fields such as healthcare, finance, manufacturing, and customer service to enhance productivity, improve decision-making, and drive innovation. The authors shed light on the challenges faced during these implementations and provide valuable insights into how organizations can navigate the complexities of AI adoption.

Key Takeaways:

• AI is no longer a distant dream but a reality that is transforming businesses and industries.
• The success of AI implementation depends on the ability of humans and machines to collaborate effectively.
• AI can augment human capabilities, but it cannot replace human judgment and decision-making entirely.
• The future of work will involve a combination of human and machine collaboration, with AI handling repetitive and mundane tasks and humans focusing on high-level decision-making and creative work.

One of the strengths of “Working with AI” is its focus on the human element in AI deployments. The authors emphasize the need for effective change management, addressing concerns about job displacement, and fostering a culture of collaboration between humans and AI systems. They highlight the importance of upskilling and reskilling the workforce to leverage the potential of AI technology fully.

Furthermore, the book delves into ethical considerations surrounding AI. Miller and Davenport discuss the potential biases and unintended consequences that can arise when employing AI systems, emphasizing the need for transparency, fairness, and accountability. They provide practical guidance on ensuring ethical AI practices and encourage readers to approach AI deployment with a responsible mindset.

The writing style of the book strikes a balance between technical depth and accessibility. The authors present complex concepts in a clear and concise manner, making it suitable for both technical and non-technical readers. The inclusion of real-life case studies adds credibility and allows readers to connect with the material, illustrating the potential of AI in tangible ways.

While “Working with AI” covers a wide range of industries and use cases, some readers may find that certain domains receive more attention than others. Additionally, the book primarily focuses on successful AI implementations, and although it briefly touches on challenges, a more in-depth exploration of potential pitfalls and failures could have added further depth to the narrative.

Strengths:

• The book is filled with real-life examples of AI implementation, making it easy for readers to understand the practical applications of AI in various industries.
• The authors provide a balanced view of AI, acknowledging its potential benefits and limitations.
• The book offers practical advice on how to work effectively with AI systems, including strategies for human-machine collaboration and tips for managing AI projects.

Weaknesses:

• Some of the examples and case studies may feel outdated, as the technology and industry landscapes are constantly evolving.
• The book could have benefited from more in-depth discussions of the technical aspects of AI, as some readers may find the subject matter too complex or unfamiliar.

Target Audience:

This book is ideal for anyone interested in understanding the practical applications of AI in various industries, including business leaders, entrepreneurs, managers, and professionals. It is also a valuable resource for students looking to gain insights into the real-world applications of AI.

Final Verdict:

Overall, “Working with AI” is a well-written and informative book that provides readers with a comprehensive understanding of the real-life applications of AI. The book’s practical advice and real-life examples make it an essential resource for anyone looking to leverage AI in their work or business. I would highly recommend this book to anyone looking to gain a deeper understanding of the impact of AI on the future of work.

Rating: 4.5 out of 5 stars.

In summary, “Working with AI: Real Stories of Human-Machine Collaboration” is a highly informative and insightful book that explores the practical applications of AI across various industries. Miller and Davenport effectively demonstrate the potential of human-machine collaboration and provide valuable guidance for organizations seeking to harness the power of AI. This book is recommended for anyone interested in understanding the real-world impact of AI and how it can be effectively integrated into different domains while considering the ethical implications.