The growth of AI technology in various industries continues to refl ect its ubiquitous outbreak. However

some performance limitations of AI also constrain the application scenarios of related products. In terms of perception

few algorithms can reach a 100% precision and high recall rate; in terms of recognition

few knowledge acquisition algorithms can reach a coverage of 100%. Therefore

strategic plans for AIbased applications in business should be cautiously designed. This paper fi rst introduces several of our AI product development practices then summarizes research and development lessons. On this basis

we give an in-depth discussion on the feasibility of AI products in terms of two dimensions: mission-critical application and non-mission-critical application;four AI product types: decision support/augmentation

intelligent agents

decision automation and smart products; three kinds of AI product business values; and two types of application assumptions. We hope these findings be helpful to decision makers in applied AI.