Introduction

Artificial intelligence (AI) plays a critical role in optimizing the value gained from digital transformation. Across different business segments, companies seek to leverage new technologies for increased revenue or lower cost. But AI is much more than an accelerator for taking the digital transformation journey to another level and making it possible for teams to work smarter, do things faster, and turn previously impossible tasks into routine.

Artificial intelligence has started to be seen as a key business enabler across more and more industries. Corporations are starting to view AI as a technology for future-proofing their business way beyond organizational efficiency. It's a revolutionary approach where AI becomes the foundation of the commercial portfolio, whether it's products, services, or some type of "as a service" setup. By embracing the full potential of AI, every company and organization in some sense becomes a technology company, whether or not that is the goal. But are companies in general ready for this massive transformation?

I would argue that most companies are not ready for this massive transformation, but it's important to remember that neither are their customers. Remember that major technology shifts like the one that AI imposes hold a lot of promise but require a fundamental transformation to take place in order to gain the expected return on investment (ROI). This fundamental shift will not happen overnight and will definitely not proceed in a synchronized manner across different markets and business segments, nor across the public sector with all its various service functions.

However, it's worth noting that the COVID-19 pandemic has accelerated the need for, and understanding of, the benefits of a fully digitalized workplace and society. However, keep in mind that just because the digitalization journey is speeding up, that doesn't necessarily mean that adding AI capabilities will be the next natural step to take.

It's not as easy as it may seem to effectively deploy and leverage AI in the enterprise. To be successful, you can't only focus on the technical pieces-you need to also address aspects such as strategy, people, and ways of working as well as how your AI solution is intended to run in production. This is crucial to break down barriers between AI in development and AI in production, and to quickly and seamlessly be able to move AI models and operate increasing numbers of models on a continuous basis in a live setting.

There is no easy fix for this, but by learning how to balance your AI investment while keeping an operational mind-set throughout, you will be more likely to succeed.

This book is centered on the fact that operating AI is not the same as operating software. That is not just a statement, but a principle that has many implications for what it means to embrace AI in your company or organization. By reading this book, you will gain insights on how to approach AI in your enterprise with operations in mind, and by doing so you are much more likely to succeed with your objectives. An operational approach should be taken directly from the start when you build your AI foundation with reproducible model pipelines. In the development phase, consider potential operational factors such as modeling the target environment or the actual use case, and you will be better positioned to build a solution that will meet its objective when it's running in live operations.

Another important aspect in this book involves truly addressing the data perspective as part of your strategic investments in AI. Remember that without the data, your AI solution cannot run. Understanding and caring for your data is vital, as well as making sure you have the data rights needed, which can sometimes be the hardest thing to manage as part of an operational setting. What you don't want to do is find out too late that the data you need isn't accessible or is owned by another party, or perhaps that the data pipeline you have invested in will not scale in production.

This book will also focus on how to successfully deploy your models as well as operate your AI solution in live environments. You will learn how different model target environments can influence aspects through the whole AI life cycle, not only which deployment options you have but which data you need to train your model on, which AI technique you will benefit most from using, how to scale your solution over time, and how and why you need to monitor and maintain your model when it's operating in production.

Finally, it's important to remember that AI is all about trust. In order for a company to rely on the AI solution to take over parts of its operations, make decisions, and let the AI system take action based on identified insights, both management and employees must trust the AI solution enough. To ensure that trust, from the start you need to think about the operational context, legal rights, and transparency and reliability aspects. This is especially valid for commercial usage of AI. In order for your customers to trust your AI-based products and/or services, you must be able to explain how your AI solution works and what is actually going on. The less your customers understand of how an AI-based solution works, the more insecure they will feel about trusting it. Customers hate to buy a black box solution. Although more complex AI techniques like deep learning can be hard to explain even for the data scientists who are building the solutions, there are ways to work with explainable AI (XAI), which will be further explored in this book.

Since the main objective of your AI investment is to realize a business value, internal or commercial, it's fundamental to understand what can be expected from your AI investment. Most companies understand the difficulties involved in reaching their objectives, but they may not fully grasp how to best navigate these challenges given a specific industry or for a specific business model. The book helps you connect these pieces, apply an operational mind-set to the business perspective, and set you on the path to success.

What Does This Book Cover?

This book covers the following topics:

• Chapter 1: Balancing the AI Investment   There is no simple answer to how to succeed with your AI investment, but there are some fundamental aspects that should be driving your objectives and realization plans, and that includes a balanced approach to AI. In this chapter you will find out what that means and why it is important for your business. The chapter will start by defining AI and by sorting out what AI is in relation to other related concepts such as machine learning (ML), automation, and robotics, just to name a few. This chapter will also address why you need to put more effort into making your AI model operational than you put into developing your AI model and how to embrace an operational mind-set for AI.
• Chapter 2: Data Engineering Focused on AI   Treating data as a valuable business asset should be the main priority in any company, and it's the key to staying on top of what is going on in your company. Leveraging data will help you understand what is not working and why, as well as enable you to see what is coming. This chapter will present a structured way for you to get to know your data and focuses on the importance of working with production. Furthermore, you will learn which data quality metrics are important and how to scale your data to succeed with your AI investment, as well as key competences in data engineering.
• Chapter 3: Embracing MLOps   In ML development the problem is seldom to technically develop, train, or implement ML models; instead, the main problem is mostly related to poor communication and lack of efficient cross-functional team collaboration. It might sound like an easy task to correct, but the fact remains that most AI projects do not make it to production due to this communication gap between the data scientists and the business. This chapter will introduce the most successful approach to tackle these problems: MLOps practices. You will learn that shifting the focus from building individual ML models to building ML pipelines is a game-changer. The chapter will also explain the importance of adopting a continuous learning approach. This chapter will also describe how to approach your AI/ML functional technology stack and ensure you have the right competences and toolsets for successful MLOps practices.
• Chapter 4: Deployment with AI Operations in Mind   It's important to remember that it's not until you deploy your models in a production setting that the value of AI can fully be realized. However, moving your models from the lab to production is far from an easy task. Successful model deployment is about a lot more than just running your model in another execution environment. When deploying AI models in production, you need to consider various areas spanning from legal rights and data access to managing retraining and redeployment of models in a live production setting. In this chapter you will learn how to handle model serving in practice and the role of the ML inference pipeline in this process. Furthermore, key success factors for industrializing AI will be outlined, as well as why it's equally important to focus attention on the cultural shift that needs to happen.
• Chapter 5: Operating AI Is Different from Operating Software   Observing and monitoring AI models in production is often...