While overt AI "flops" are less common than they were three years ago, the pattern of failure for AI projects is still much the same. Assessing project opportunities and having the right expectations about AI's adoption challenges is paramount to converting pilots into actual deployments. Especially for firms new to artificial intelligence, it's crucial to demonstrate progress and success with AI early on – in order to win approval, confidence, and additional budget to expand AI efforts productively. Nothing builds confidence like actually putting an AI application into production. These kinds of small wins are exactly what an AI catalyst needs to achieve in order to deliver actual transformative impact with AI. Because of this, assessing AI projects by their likelihood to reach deployment can be a useful lens – particularly for companies early in their AI journey and unaccustomed to the iteration and experimentation that AI requires. The remainder of article has been transcribed from one of our live sessions of the Catalyst Advisory Program.

A version of this article appeared in the Spring 2019 issue of strategy business. Robots will play a critical role throughout the modern workforce. A recent PwC study (pdf) of 29 countries estimated that on average, the proportion of jobs at high risk of automation will be roughly 20 percent by the late 2020s, and 30 percent by the mid-2030s. For employees around the globe, such shifts could mean preparing to share space with robots as colleagues or learning new skills as robots take over the most repetitive or dangerous tasks. When we talk about robots, we are essentially referring to two kinds: industrial robots and service robots.