<p>Up to 40% of hospitalized patients experience hyperglycemia, often caused by undiagnosed diabetes or stress-induced hyperglycemia. These patients commonly have multiple health conditions and face higher risks of hypoglycemia and blood sugar fluctuations, which can lead to worse outcomes. Algorithm-guided insulin therapy has emerged as a promising approach to enhance inpatient blood sugar management and tailor insulin dosing to individual needs. A narrative literature review was conducted across PubMed, Embase, and Scopus (as of January 2026), encompassing studies published in English or Spanish. Eligible studies addressed the development, validation, or clinical application of algorithmic tools—particularly those using artificial intelligence (AI)—for inpatient insulin therapy. Due to heterogeneity in study design, interventions, and outcomes, a meta-analysis was not performed; instead, a thematic synthesis was employed. The literature indicates an increasing interest in algorithm-guided insulin therapy to enhance inpatient glycemic control. Tools range from deterministic algorithms, based on fixed rules, to advanced AI-driven systems capable of dynamic dose adjustments. Most hospitals still rely on deterministic models, while AI-based systems are gradually being implemented. Machine learning–based approaches, though promising, remain in early phases of development. Initial studies suggest potential improvements in glycemic outcomes, time in range, and safety profiles; however, clinical evidence remains limited. Despite technological advances, the widespread implementation of algorithm-guided insulin therapy is hindered by regulatory limitations, concerns about sensor accuracy, integration challenges, and the need for staff training. Broader clinical validation, ethical oversight, and institutional support are essential for safe and effective adoption.</p> Graphical Abstract <p></p>