chapter one

1 Why we need a new way to test AI opportunities

 

This chapter covers

  • The main reasons AI projects fail to deliver impact
  • The process we propose for testing AI ideas and avoiding failure
  • Why AI ideas should be tested before any development begins
  • The economics of testing AI opportunities early

Artificial Intelligence (AI) offers extraordinary opportunities across every sector and business function. The use of classical machine learning, generative AI, optimization, rules-based systems, simulation, reinforcement learning, or any combination thereof—which we simply refer to as AI in this book—can produce predictions, recommendations, decisions, or content that influence environments and generate significant benefits.

Yet, while organizations are racing to invest in AI, too many are failing to translate its promise into tangible value. If you haven’t begun your AI journey, you’re likely losing competitiveness and leaving value on the table. But if you have started, chances are high you’ve already encountered frustration: more than 80% of AI projects fail to deliver lasting impact.

1.1 How a good idea became a costly failure

1.1.1 What happened

1.1.2 Lessons from the Zillow Offers use case

1.2 AI failure rates and their main causes: What we know

1.2.1 AI failure rates

1.2.2 The causes of failure in AI projects

1.3 The three lenses of AI Road Testing

1.4 Turning known risks into structured decisions

1.5 The window of opportunity is before the proof of concept

1.5.1 A proof of concept won’t solve a problem-framing issue

1.5.2 A proof of concept won’t solve an approach issue

1.5.3 Decide your exit before you enter

1.5.4 A clear road test will give you exit flexibility

1.5.5 The AI project lifecycle and where the AI Road Test sits

1.5.6 Stage-gate does not mean not agile

1.6 Good planning is good business

1.6.1 An excellent return on investment

1.6.2 Faster time to value

1.7 Summary