We are all familiar with adages like, “Change for the sake of change, benefits no one” and “Change, for the sake of change, is a recipe for failure.” On the other hand, three business professors, Freek Vermeulen (@Freek_Vermeulen), from the London Business School, Phanish Puranam, from INSEAD, and Ranjay Gulati, from Harvard Business School, insist, “No one disputes that firms have to make organizational changes when the business environment demands them. But the idea that a firm might want change for its own sake often provokes skepticism. Why inflict all that pain if you don’t have to? That is a dangerous attitude. A company periodically needs to shake itself up, regardless of the competitive landscape.”[1] Change is hard, especially when that change involves new technologies, like machine learning. Nevertheless, many business gurus insist companies need to implement machine learning solutions if they are going to compete in the years ahead. Eric Siegel (@predictanalytic), a former computer science professor at Columbia University, writes, “Machine learning … is genuinely powerful and everyone oughta be excited about it.”[2]

When to implement a machine learning solution

Business leaders frequently confront conundrums, like the conflicting advice discussed above about the benefits and downsides of change for the sake of change. Normally, the safest course of action is to move forward when a solid business case can be made for change — even if there is no compelling external reason to change. Since many business experts insist survival in the future requires organizations to transform into digital enterprises, executives in every economic sector are trying to figure out the best course of action to accomplish that goal. Many of them are looking at implementing machine learning solutions as a first step in the transformation process; however, Daniel Faggella (@danfaggella) Head of Research at Emerj, notes, “Most technology executives often have trouble identifying where their business might actually apply machine learning (ML) to business problems.”[3] He suggests asking four pertinent questions before embarking on any machine learning project:

1 – Is the prediction you’re trying to make (or decision you’re trying to make) complex enough to warrant ML in the first place? Faggella explains, “If it’s possible to structure a set of rules or ‘if-then scenarios’ to handle your problem entirely, then there may be no need for ML at all. Also, if there is no precedent for any successful outcome applying machine learning to the specific problem to which you’re developing, it may not be the best foray into the ML world.”

2 – Do you have new data and clean data? To be effective, machine learning needs to train on the right set data. Faggella writes, “‘Clean data is better than big data’ is a common phrase among experienced data science professionals. If you have reams of business data from years ago, it may have no relevance today, particularly in fields where the basic business processes change drastically year-over-year, such as mobile eCommerce. If you have reams of unstructured and disjointed data, you may have too much ‘cleaning’ to do before you can ever get around to learning from the information collected.”

3 – Does your data have existing labels to help a machine make sense of it? Freelance writer Cynthia Harvey explains there are four types of machine learning.[4] They are:

• Supervised Learning: “Supervised learning requires a programmer or teacher who offers examples of which inputs line up with which outputs.”
• Unsupervised Learning: “Unsupervised learning requires the system to develop its own conclusions from a given data set.”
• Semi-supervised Learning: “Semi-supervised learning, as you probably guessed, is a combination of supervised and unsupervised learning.”
• Reinforcement Learning: “Reinforcement learning involves a system receiving feedback analogous to punishments and rewards.”

Faggella adds, “While unsupervised learning allows for a wide degree of applications in making sense of data without labels, it’s usually not advised for companies to ‘jump into’ ML with a first application in unsupervised learning. The low-hanging fruit for an ML use case is likely to spawn from its historical, labelled data.”

4 – Can your solution to this problem afford for some allowance of error? Executives looking for clear cut, black or white, indisputable answers from any form of artificial intelligence, including machine learning, are likely to be disappointed. Faggella explains, “ML might be thought of as a kind of ‘skill’, in the same sense that one might apply the word to human beings. A skill that’s alive, adapting, growing and informed by experience. For this reason, an ML solution will often be incorrect a certain percentage of the time, especially when it’s informed by new or varied stimuli. If your task absolutely cannot allow for any error, ML is likely to be the wrong tool for the job.”

Faggella goes on to quote a number of experts who suggest a few situations that are ripe for a machine learning solution: 1) When a business problem exists and you have hard data, variability, and a large number of examples; 2) When there are demonstrable savings to be made; 3) When you need predictions based many variables, having complicated non-linear relationships between them and in some cases are highly stochastic; 4) When there is a very large, historical data set that includes both rich features and some kind of direct feedback that can be used to build an algorithm that can be implemented and tested easily and will either decrease operational costs and/or increase revenue immediately.

Machine learning use cases

Sometimes we all need a spark of insight to get our creative juices flowing. If you are struggling with how machine learning can help your business, seeing how others are using the technology might help. Current use cases include:

• Face detection: Facial recognition systems very controversial and privacy concerns about their use are growing. Faggella discusses this use case because it demonstrates the important role complexity plays in making a decision about whether or not to use machine learning. Faggella notes, “It’s incredibly difficult to write a set of ‘rules’ to allow machines to detect faces (consider all the different skin colors, angles of view, hair/facial hair, etc.), but an algorithm can be trained to detect faces.”
• Combatting fraud. Nick Ismail (@ishers123), Editor for Information Age notes, “While unsupervised models might be the goal in the long term, supervised and semi-supervised models in which trained fraud experts provide input to better train and fine tune their ML models are critical to creating a fraud prevention system that can proactively identify fraudulent activity while reducing the incidence of false positives that can negatively impact the customer experience.”[5]
• Recommendation engines. Tech writer Gordon Gottsegen (@GGottsegen) points to Netflix as a great example of using a recommendation engine. He writes, “Using machine learning to curate its enormous collection of TV shows and movies, Netflix taps the streaming history and habits of its millions of users to predict what individual viewers will likely enjoy.” Organizations providing a large selection of products and services to the general public can probably benefit from a recommendation engine.
• Calculating customer lifetime value metrics. Gottsegen notes, “This metric estimates the net profit a business receives from a specific customer over time.” Loyal customers are generally better customers than newly acquired customers; yet, many companies spend more money trying to attract customers than to retain customers they already have.
• Improving efficiency and sustainability. More and more consumers are making purchasing decisions based on a company’s efforts to be more sustainable. Freelance digital journalist Keith Shaw notes, “Organizations have recognized that by applying ML models to the data they have on business processes, they can achieve new insights that help reduce waste and preserve natural resources.”[7]

Markus Noga (@mlnoga), Senior Vice President of machine learning at SAP, observes, “There [are] three main ways that organizations use machine learning today, and we like to group them into the areas of automation, conversation, and intelligence. If you think about automation, things like factory robots come to mind, and the same repetitive tasks that robots have automated in factories, automation technology for software can automate in desktop environments. This category of software is called robotic process automation. … The second big category is about the conversation. Not the conversation that you and I are having here, but repetitive large scale conversations around customer service, around procurement inquiries, around parts or order inquiries, where a small number of domains account for a large conversation volume. And by putting in the chat box and conversational agents, we enable the humans to focus on the challenging, the difficult to serve, or the value-adding parts of the conversation and letting bots take care of the routine elements and the repetitive tasks of this. … Last but not least, people are using AI to bring actual intelligence into business processes. This is about training little machine learning modules that interface with the data, that interface with the decisions, and that keep learning in the business process. This kind of model brings the biggest value whenever decisions are at stake and whenever you need to decide anything from which banner ad to serve, all the way to whether to offer a line of credit to this customer.”[8]

Obviously, I haven’t presented an exhaustive list of all that machine learning can do for an organization. I’m of the opinion that companies won’t have to look very hard to find ways that machine learning solutions could prove beneficial. Alfred P. West, founder and chief executive of SEI, has stated, “Change is not easy, [but] you can’t dodge it. It is with you. And you’d better embrace it.” If you haven’t implemented machine learning solutions, your competitors probably have. It’s time to embrace change.

Footnotes
[1] Freek Vermeulen, Phanish Puranam, and Ranjay Gulati, “Change for Change’s Sake,” Harvard Business Review, June 2010.
[2] Eric Siegel, “Why A.I. is a big fat lie,” Big Think, 23 January 2019.
[3] Daniel Faggella, “How to Apply Machine Learning to Business Problems,” Emerj, 25 April 2020.
[4] Cynthia Harvey, “What is Machine Learning?” Datamation, 3 January 2018.
[5] Nick Ismail, “Enabling machine learning to help combat fraud,” Information Age, 4 October 2017.
[6] Gordon Gottsegen, “15 Examples of machine learning making established industries smarter,” BuiltIn, 12 April 2019.
[7] Keith Shaw, “3 use cases for machine learning you probably haven’t thought of,” CIO, 17 August 2020.
[8] Macy Bayern, “How organizations can make the most of machine learning,” TechRepublic, 10 May 2019.