Alzheimer’s Disease (AD) is a neurodegenerative disease that primarily affects cognitive functions and memory loss. The maximum effectiveness of medications rely on early detection of the presence of AD, which requires invasive extensive blood tests and/or brain imaging. In this study, we present a speech-based tool to detect the presence of mild cognitive impairment (MCI) and probable AD (pAD). Through the Speech Vitals application, in conjunction with the Bio-Hermes study, we administer a set of speech and language-based tasks targeting motor and cognitive impairments. We extract metrics from speech and language tasks to generate models that discriminate between healthy controls and individuals with MCI or pAD with an AUC of 0.83, and between controls and individuals with pAD with an AUC of 0.90. These models highlight the promise of a speech-based tool for screening cognitive impairment and AD, which could help inform individuals when they should seek additional diagnostic tools.