User Centered AI Products
Last week I re-watched Andrew’s Ng “The State of Artificial Intelligence” at EmTech. The former chief scientist at Baidu provides one of the clearest frameworks to understand the state of AI. In his lecture, Andrew Ng explains that AI is particularly adept at performing tasks that take humans less than 1 second to perform. This insight prompted me to re-read Jobs Theory, and McKinsey’s study on workforce automation.
If AI’s strength is automating 1 second tasks, then AI product managers must examine which occupations have these type of tasks. Uncovering these activities represent an opportunity in building successful AI products. Needless to say this exercise alone will not result in successful ventures. Donna Romer, VP of product at IBM Watson summarizes other key takeaways for product managers from Andrew’s lecture in this post.
When it comes to those 1 second tasks, the principle “Just because you can doesn’t mean you should” should still apply. Buyers and users will need to see a real benefit for the technology to diffuse in the market. The experience should be superior as a result of using AI. Consider the self service meu-tablets connected to POS systems at airport restaurants. These devices enable customers to order and pay for their own food without the need for human help. This technology eliminates the need for waiters but it negatively affects the customer experience. If you hate these germ-infested menus as much as I do, chances are airports will be their first and last market. Adoption of AI will be slow if innovators ignore the users job.
AI is successful when it can do the job and augment the user experience.
A framework for applying AI to occupations is essential to the success of AI ventures and good design. McKinsey’s study provides an initial step to understand occupations. More work need to be done to figure out new ways of creating product in the AI era. However, one thing will not change: user-centered design is key to making AI products that people want to use.
