Alzheimer’s disease (AD)-related neurodegeneration can begin several decades before the first symptoms appear, and AD biomarkers have only recently been identified. Our research focused on identifying AD-related biomarkers in healthy subjects. We analyzed Biocard data using a group of 150 subjects in different AD stages, with 40 AD patients compared to normal subjects. Using the granular computing method, we identified sets of attributes related to various stages of the disease. We applied this classification to the psychophysical test results of normal subjects to determine if some cases might show granular pattern similarities to groups with AD. Earlier detection can uncover patterns invisible to neuropsychologists in the cognitive like executive or immediate/delayed memory functions. These patterns may help to determine if a patient is getting amnestic or non-amnestic early symptoms that lead to dementia or not. Our analysis of Biocard data identified granules that classify cognitive attributes with disease stage scores (CDRSUM - Clinical Dementia Rating Scale Sum of Boxes). By applying these rules to 21 normal subjects with a CDRSUM of 0, we predicted that one subject might develop mild dementia (CDRSUM > 4.5) with a mixture of executive and cognitive test results. Another patient might get very mild dementia (CDRSUM > 2.25), but with dominating memory problems which might lead to serious amnestic symptoms. Other patients might develop questionable impairment (CDRSUM > 0.75). Our AI-powered method can identify patterns in cognitive attributes in normal subjects that might indicate their pre-dementia amnestic or non-amnestic stages, which may be invisible to neuropsychologists.