Blood serum clinical chemistry is an essential part of clinical pathology diagnosis. Diagnostic parameters such as bilirubin (BIL), triglycerides (TRIG), uric acid (UA), urea (UR), creatinine (CRE), and myoglobin (MIO) are currently measured using reagent-based methods and wet lab chemistry, either at the clinical pathology laboratory by automated clinical chemistry machines (ACC) or at the point-of-care (POC). The use of reagents and fluids in technologies, such as chemo/biochips, restricts the application of POC technology to controlled environments and the need to manage consumables. Furthermore, blood is still collected by traditional venipuncture, centrifugation, or filtration to remove the cellular fraction, which increases the complexity of the operation and makes it less attractive for emergency or harsh scenarios. The current state-of-the-art limitations imposed by wet chemistry can be overcome using Spectral POC technology. In this study, we assessed blood serum chemistry using ultraviolet, visible, and near-infrared spectroscopy to relate spectral information to blood serum composition using self-learning artificial intelligence. This preliminary study demonstrated the feasibility of the direct detection of BIL, TRIG, UA, UR, CRE, and MIO. These parameters were determined with significant statistical agreement in terms of correlation (R) and total error (TE) with the ground-truth methods used in the clinical pathology laboratory, where i. BIL: R=0.99 and TE=10.4%; ii. TRIG: R=0.84 and TE=22.6%; iii. UR: R=0.82 and TE=32.3%; iv. CRE: R=0.88 and TE=20.9%; and v. MIO: R=0.96 and TE=42.65%. These results show that it is possible to perform clinical chemistry in the blood serum without any reagents. Further research is necessary to enhance the detection of constituents with less absorbance and expand the range of parameters investigated to develop a system that can perform blood chemistry directly in a drop of blood without using reagents or with the need to remove the cell fraction.