Thomas H Davenport is currently the President’s Distinguished Professor of Information Technology and Management at Babson College. In his book The AI Advantage: How to Put the Artificial Intelligence Revolution to Work, he provides a factual guide to what AI is, how businesses are using it today, and how it could impact business in the future.
Davenport first defines what AI is. He defines AI as a group of cognitive technologies that apply “knowledge, insight, and perception to solve narrowly defined (with the current state of the technology) tasks” (Pg. 9). Given this definition, I find it easiest to conceptualize cognitive technologies as technologies that given a defined task can do that task with little or no human intervention. Within this group of technologies, there are seven specific technologies. The first is statistical machine learning, this automates the process of fitting models to data and is used to analyze big data and draw conclusions from it. The next is neural networks which “uses ‘artificial’ neurons to weight inputs and relate them to outputs” (Pg. 11) and is used to identify credit fraud. Expanding on neural networks is deep learning which is essentially many neural networks with many different variables and features. Deep learning is used in image recognition. The fourth cognitive technology is natural language processing which is used to understand human speech. Rule-based expert systems follow a set of rules defined by humans in things like credit approval. Finally, physical robots automate a physical activity like in warehouses and robotic process automation (RPA) does the same thing but for digital tasks. These seven technologies make up the group of defined cognitive technologies and they each have a unique application to business.
After defining the current capabilities of AI, Davenport describes how these technologies are currently being put to use in businesses. He describes three main processes by which cognitive technologies can be used, automating structured and repetitive work processes, gaining insight through analysis of large structured data, and using natural language processing to engage customers and employees. Out of these three capabilities Davenport first describes how companies are automating structured processes as this is the most common use of AI. Some examples of this include replacing lost credit or ATM cards and transferring data from email and call centers into databases and records. This is usually done using RPA’s. The next most common use of AI is to gain cognitive insight from large sets of data. This is usually done by using machine learning on massive sets of data for things like predicting whether a particular customer is likely to purchase a product, as well as identifying credit fraud in real time. Finally, Davenport describes how companies are creating cognitive engagements through processes like intelligent agents who offer 24/7 customer service (similar to Bank of America’s Erica), and product and service recommendations. While all of these processes are currently in use at various companies, Davenport points out that AI technologies are still difficult to develop and even more difficult to implement and scale using an organization’s existing enterprise software and systems.
After describing the current state of AI Davenport makes two very important arguments. The first being that even given the disruptive nature of cognitive technologies these technologies still are not fully capable of changing current business models. He gives a variety of reasons for this trend including that, “AI only picks off the easiest parts of the process” (Pg. 81), and AI does not possess the common sense that humans put to use when conducting business. The second argument that Davenport makes for the future of AI is that contrary to popular concern, he believes that AI will not automate jobs, rather augment them. He provides five reasons for this, (1) AI can only automate tasks and not entire jobs, (2) when surveyed current managers do not want to use AI to reduce the number of jobs, (3) history suggests that even with automated technology human jobs persist (think bank tellers and ATM’s), (4) people tend to find new jobs and tasks to perform, (5) a lot of entirely new jobs will be created. Essentially, Davenport argues that while AI is a powerful technology that cannot be ignored it is still very difficult to implement, and also will not eliminate the need for humans in the workplace. Davenport then concludes by addressing that AI also raises ethical concerns such as algorithmic biases and data security that must be addressed as AI continues to permeate society.
One topic that I found very interesting and I believe will be relevant to our trip is the use of AI to plan and optimize operations. This is one of the key business activities that Davenport argues AI will be used to create competitive advantage. One example of this possibility is demonstrated by the steel manufacturing company Big River Steel. Big River Steel currently uses sensors, control systems and a variety of statistical machine learning models to improve its business in six major areas. The first is demand prediction, by feeding a machine learning model macroeconomic data, historical steel demand, manufacturing activity, and the activity of large steel consumers, Big River Steel is able to forecast its demand and decrease overproduction much better than its competitors. Big River Steel also uses machine learning for sourcing and inventory management. It uses a machine learning model to predict the availability of scrap steel and when best to buy it. Next, Big River makes use of scheduling optimization by using an optimization model that makes the most efficient use of energy to minimize energy costs. Another way that Big River Steel makes use of AI is in production optimization. A large problem that plagues steel mills is unplanned events like when molten steel breaks out of molds during casting or when hot rolled steel escapes its rollers onto the mill floor. These events stop production and are very dangerous. By making use of its array of sensors and machine learning, Big River can predict when they are most likely to occur and therefore minimize the chances that they actually occur. Another use of Big River’s sensors come in the form of predictive maintenance to identify the best times to maintenance plant machines and equipment. Finally, Big River Steel uses AI to minimize delivery costs and give a clearer delivery window to consumers. Big River Steel represents how a non-technology company can make use of AI to gain a competitive advantage.
In conclusion, I found The Ai Advantage to be both informative and very interesting. Davenport is able to clearly articulate how cognitive technologies can truly be applied to business and also what their limitations are. Given that there is significant hype surrounding AI I feel it is very important to be aware of what is truly possible and what is just hype. Davenport also makes reasonable projections about how AI can be applied to current organizational processes that further explain how AI can affect business. Finally, Davenport makes a clear argument for augmentation over automation that I found to be very compelling. To conclude I would recommend that anyone interested in developing a competitive advantage through technology read this book.