The AI Playbook / Eric Siegel

Back in the ancient days, I was developing data warehouses. The advice was then: first asks what kind of decisions the users will need to make, then ask what data they need for making the decisions. Everything else follows from there. 

Eric Siegel provides us with similar advice in the AI playbook. The book introduces a six-step process for making sure that a new AI based prediction system will be a success. By success Siegel means change in the organisation. The AI systems will change the way the organisation makes decisions, so if the organisation does not change, the new system has failed. 

Siegel puts lot of emphasis on the idea that AI is mostly used for making predictions, such as which customer will react to advertisements, or which diagnosis applies to a new patient. An AI system is, according to Siegel, very useful on in making predictions. And then action on the predictions is the change in the organisation. 

The six steps of Siegel’s are shown in the picture. They are quite easy to understand, but necessary. Especially, it is important to decide what is the deployment goal, and what the model will predict. These steps are too easy to overlook when being overly enthusiastic about AI. 

The books is well worth the 4-5 hours it takes to read. I find the emphasis on the value proposition very important.