During the last years some decision tools on the detection of cancer have been published. Examples are the Q-cancer risk scores for all patients in practice and the Caper-studies on the risks of patients with alarm symptoms. The impression however is, that these instruments are not used on a regular basis and, if used this happens mostly unconsciously. Meanwhile survival increased, probably not only because of better therapy, but also of earlier detection (by introduction of screening and by increased symptomatic detection). However, this increase is not very impressive, about 10% in 40 years which led to current survivor rate of 60%. In this round table we will approach this world wide problem on three ways. Firstly we will explore why general practitioners underuse current detection algorithms and we will present a future, where artificial intelligence might help the GP to refer patients with cancer earlier. The question however will be: on what costs? We will discuss this both clinically as well as from a societal and ethical perspective. To end with we will try to gain understanding on how we make decisions, how much uncertainty do we think is acceptable and are formulated referral thresholds helpful. We hope to end with some conclusions and some sort of an agenda for the future.