List of Figures

Chapter 1. Introduction

Figure 1.1. A machine learning algorithm produces a result that minimizes the evaluation metric. At its core, machine learning is data transformation from some predefined input to an answer in a predefined output format. It’s a human’s job to align the evaluation metric with the business goals.

Figure 1.2. Machine learning is a combination of formulation, optimization, and evaluation. The only part that the computer can do (mostly) on its own is optimization.

Figure 1.3. You can perform many more analyses than there are business actions you can take. Don’t spend a ton of time and money on an analysis just to figure out it was never an actionable one. (Earth image is from Wikipedia [20].)

Figure 1.4. Elements of CLUE. AI projects that don’t have good answers for all elements of CLUE experience difficulties.

Figure 1.5. High-level overview of the process that a successful AI project should use. This workflow shows you how to make sure that all elements of CLUE are part of your project.

Chapter 2. How to use AI in your business

Figure 2.1. The Sense/Analyze/React loop. Any successful analytical project must have all three elements of this loop.

Figure 2.2. Making money with AI is based on finding a business problem in which you can apply the Sense/Analyze/React loop to one of the actions you can take.

Figure 2.3. A workflow of a routine optometry exam. We’ll apply AI to automate part of this workflow.

Figure 2.4. General principles for applying AI to a business problem. The basic idea is to make sure that you’ll be able to implement all parts of the Sense/Analyze/React loop.

Figure 2.5. AI taxonomy based on the high-level role it plays in business. You could use this taxonomy to guide you in eliciting available business actions you can help with AI.

Figure 2.6. Taxonomy based on AI capabilities. This framework groups broad areas of AI capabilities so that you can quickly check if any of them are applicable to the business problem you’re addressing.

Figure 2.7. Gap analysis between skills the team has and the ones needed. This analysis allows you to create a plan for how to address missing skills.

Chapter 3. Choosing your first AI project

Figure 3.1. You’re in a rich hunting ground—plenty of rabbits and a big mammoth are in sight. Which animal should you try to catch first?

Figure 3.2. Start with easy projects. Success with those projects enhances your skills and reputation within the rest of the company, allowing you to attempt more difficult hunts later.

Figure 3.3. You’ve cornered the mammoth on your first hunt. What are you going to do now?

Figure 3.4. The C part of the CLUE allows you to create a list of viable AI projects and estimate their complexity.

Figure 3.5. The React part of the Sense/Analyze/React loop: finding business problems that AI can react to. Once you’ve identified the role AI plays in your business, ask the questions provided here.

Figure 3.6. Data science methods and data are interconnected and influence each other. Never discuss a method without asking where you can get the data needed to train it.

Figure 3.7. The translation of a business question into a research question. AI doesn’t understand business concepts. If you aren’t familiar with statistics, a research question formulation might be difficult to understand.

Chapter 4. Linking business and technology

Figure 4.1. Once a project starts, decisions have to be made. In which plane would you rather be?

Figure 4.2. Linking business and technology. You must make sure you have the right relationship between business questions and business metrics.

Figure 4.3. Technical metrics are known to escape into the wild. This is not a reason to invite them to business meetings.

Figure 4.4. A profit curve specifies the relationship between a technical metric and a business metric. It allows you to understand what the technical answer (in the form of a technical metric) means for the business terms.

Figure 4.5. Profit curve for bike rentals. Note that in the case of a business metric being a cost, the goal is to minimize the business metric.

Figure 4.6. A typical profit curve when you’re working in an academic environment. Such a profit curve allows you to focus only on the technical metric.

Figure 4.7. A profit curve is a bidirectional relationship. You can use it to answer a business question based on the technical metric.

Chapter 5. What is an ML pipeline, and how does it affect an AI project?

Figure 5.1. Example of an ML pipeline for a factory line. This ML pipeline oversees the automated ordering of consumable supplies from suppliers.

Figure 5.2. Modification of the ML pipeline from figure 5.1 for a situation in which supply levels are stored in the machine itself. You don’t need to worry about EDW or merging data and can assume that the historical data is correct, so there’s no need to clean the data.

Figure 5.3. An ML pipeline requires data engineering and data science expertise to construct. Specialists in one area might not know another area well, making mismatched elements of an ML pipeline more likely.

Figure 5.4. Example of an ML pipeline that’s used for an AI security system prototype.

Figure 5.5. A simple pipeline with just three stages, of which only the first two stages affect the end result.

Figure 5.6. The role of methods in the construction of an ML pipeline. The methods that implement a step in the pipeline often change during the lifespan of the pipeline.

Figure 5.7. The relationship of the business question, methods, data, and infrastructure. The order in which you make decisions is the business AI infrastructure.

Chapter 6. Analyzing an ML pipeline

Figure 6.1. Project management decisions are made under time constraints. Some projects are like the plane in this figure and must correct their course before they crash. Doing nothing can sometimes be as dangerous as taking the wrong action.

Figure 6.2. The Economize part of the CLUE process. MinMax and sensitivity analyses let you apply your efforts to the right parts of a business-viable ML pipeline.

Figure 6.3. A MinMax analysis consists of the Min part (–) and the Max part (+). It uses a profit curve to transform the technical metric into an expected business result. The result of this analysis determines the viability of your ML pipeline.

Figure 6.4. A simple ML pipeline for the automated parking meter that takes a picture of the license plate, checks if parking is legal, and issues a citation if not. The examples in this chapter use this pipeline for the analysis.

Figure 6.5. Here, the Min analysis doesn’t reach the value threshold, but the Max analysis exceeds it. It might be possible to improve this pipeline enough to make it business-viable.

Figure 6.6. Analysis of an ML pipeline. This analysis tells you what the current ML pipeline can achieve for the business. Both the Min and Max part of the MinMax analysis use the same process.

Figure 6.7. Various types of profit curves. The linear and monotonic profit curves are simpler to optimize than the non-monotonic profit curve. Avoid non-unique profit curves.

Figure 6.8. MinMax analysis of a non-monotonic profit curve. The Min analysis result is the best value of the business metric on the segment [0, Min Point]. The Max analysis result is the highest value of the business metric on the segment [0, Max Point].

Figure 6.9. The order in which you should perform a MinMax analysis on the ML pipeline. Perform the part of the analysis that can provide conclusive answers first.

Figure 6.10. An example of an ML pipeline. We’ll use this pipeline for the exercises in this chapter.

Figure 6.11. AI taxonomy based on the high-level role it plays in business. You could use this taxonomy to guide you in eliciting available business actions you can help with AI. (This figure is a repeat of figure 2.5.

Chapter 7. Guiding an AI project to success

Figure 7.1. An example of an ML pipeline. We use this pipeline as a base example for sensitivity analysis. (This repeats figure 6.10 for the reader’s convenience.)

Figure 7.2. Local sensitivity analysis. For small improvements in the response of stage B, this analysis assumes a linear response in the ML pipeline. If a gain of 0.5% in stage B results in a 1% improvement in the pipeline’s result, then a 1% gain in stage B would result in a 2% improvement in the entire ML pipeline.

Figure 7.3. Global sensitivity analysis uses a wide range of values. You perform global sensitivity analysis when there’s no reason to believe that only small improvements are possible in some stages of the pipeline (in other words, you believe drastic improvements are possible).

Figure 7.4. Dependencies between the parts of CLUE—subsequent stages depend on the correct implementation of the previous stages. Following CLUE enables you to work on the right business problem, choose the proper ML pipeline to solve that problem, and always work on improving the right stage of the pipeline.

Figure 7.5. Convexity in the ML pipeline’s response. The further you move away from the point at which analysis was performed, the more significant the error in your analysis. Never extrapolate far from the point from where your local sensitivity analysis was performed.

Figure 7.6. Sensitivity analysis with convexity present. Localized sensitivity analysis performed at only two points, x and x + 1%, has missed the convexity because you can always draw a line between two points.

Figure 7.7. Global sensitivity analysis in the presence of convexity. At a price of the increased complexity of the analysis, global sensitivity analysis could detect the presence of nonlinearity of response.

Figure 7.8. Sensitivity analysis performed in locations S1–S5 but missing the actual maximum at point M. It doesn’t matter; you’re above the value threshold, so you’re still making money. No cheap analysis performed on only a few points can avoid this problem.

Figure 7.9. If the underlying curve has a shape given in this figure (as opposed to figure 7.8), sensitivity analysis will find the actual maximum. You never know the shape of the underlying curve, so it doesn’t matter what the maximum of that curve is. What matters is that you’re above the value threshold.

Figure 7.10. A set of profit curves for a long-running AI project. Both the shape of the profit curve and especially the value threshold change with time.

Figure 7.11. Modification of the Understand part of CLUE to account for long delivery times. The project value changes with the time it takes to deliver it. You must use the profit curve corresponding to the time when you would release software to calculate the value of the improvement.

Figure 7.12. Improving the ML pipeline accounting for the time parameters given in table 7.1 and 7.2. The order of improvement of the ML pipeline stages should be first stage A, then stage C, then stage D, and finally stage B. That order allows you to release a viable product after only two months.

Figure 7.13. Running an AI project with the ability to put challenging projects on hold. This approach allows you to quickly cut your losses on projects that prove to be more difficult than anticipated and, instead, allows you to try a more straightforward project.

Figure 7.14. An example of an ML pipeline. We will perform sensitivity analysis of this pipeline. (This is a repeat of figure 6.10 for the reader’s convenience.)

Chapter 8. AI trends that may affect you

Figure 8.1. AI versus machine learning (ML). As you can see, AI is a superset of ML.

Appendix B. Exercise solutions

Figure B.1. ML pipeline for sentiment analysis of the customer feedback

Figure B.2. An example ML pipeline. This figure is a repeat of figure 6.10.

Figure B.3. An AI taxonomy based on the high-level role it plays in business. You could use this taxonomy to guide you in eliciting available business actions you can help with AI. This figure is a repeat of figure 2.5.

Figure B.4. An example ML pipeline. We use this pipeline as a motivating example for sensitivity analysis. This is a repeat of figure 6.10 for the reader’s convenience.

Figure B.5. A timing diagram that answers question 4 of chapter 7’s exercises