Chapter 3. How to Overcome AI Failures and Challenges
In Chapter 2, we talked about why AI is the greatest opportunity of our time—namely, its potential to add almost $16 trillion to the global economy by 2030. We also learned that adoption is slow. But what are the reasons behind that slow adoption? Why would such a tremendous opportunity to expand the economy in unparalleled ways not inspire a rush to production?
As it turns out, implementing AI in organizations is hard, and it’s only in the past few years that technology, price, and funding have met in the middle. Some existing AI experiments have failed, and in a very public way—but like with most headlines, there’s usually more to the story.
In this chapter, we’ll talk about why the time for AI is now, some early examples of success in AI, and some very familiar failures of AI in business and what went wrong. We’ll then dive into the challenges to find out why adoption is slow today, even when the issues of data availability and horsepower are solved. And finally, we’ll look at some of the tools and services available to organizations that dramatically lessen the impact of those challenges.
Let’s dig in.
AI’s Emergence in Business Today
As O’Reilly detailed in its 2019 report AI Adoption in the Enterprise, the maturity of an organization’s AI adoption varies by industry (Figure 3-1). For example, 30% of those who work in finance describe their organization as having a mature AI practice, compared to just 16% from the public ...