A clinical trial conducted in Kenya in 2026 found that a generative AI-enabled clinical decision support system, called 'AI Consult,' improved documentation and treatment planning in primary care settings but did not reduce rates of treatment failure.

The study, published in Nature Medicine on June 26, 2026, involved over 9,600 patients across 16 primary care clinics. Clinicians participating in the trial were randomly assigned to utilize an electronic medical record system either with or without the integrated 'AI Consult' tool.

The AI Consult system functioned as a large language model-based clinical decision support tool embedded within the existing electronic medical record system. It analyzed information entered by clinicians and generated diagnostic and treatment suggestions that aligned with Kenyan national clinical guidelines. The tool also flagged potential concerns using a color-coded alert system. Clinicians maintained full autonomy and were not obligated to follow the AI's recommendations, and the AI interface was not visible to patients during consultations.

Researchers determined there was no difference in treatment failure within 14 days between patients seen with AI-supported care, which stood at 2.2%, and those receiving standard care, which was 2.0%. The study also found no evidence of harm, with similar rates of hospitalization and death observed in both groups. An independent panel of blinded clinicians assessed that the AI tool led to improvements in the quality of clinical documentation and treatment planning, and patient satisfaction remained consistent across both the AI-supported and standard care groups. Antibiotic-related costs were lower in the AI-supported group due to more cost-conscious prescribing choices; however, overall antibiotic prescribing rates were similar between the groups.

Bilal Mateen, Professor of Machine Learning for Health at the University of Birmingham, stated, "What we found is reassuring but also sobering. The technology appears safe and clearly improves aspects of clinical decision-making, but translating those gains into measurable patient benefit is much more challenging, particularly in everyday primary care." Professor Alastair Denniston, Professor of Regulatory Science and Innovation at the University of Birmingham, added, "What this study shows is that AI can be integrated safely into real clinical workflows, without undermining patient trust or clinician autonomy – which is a critical foundation for any future impact."

The trial was conducted by experts at the University of Birmingham with support from the National Institute for Health and Care Research (NIHR) Biomedical Research center: Birmingham. Collaborators from the London School of Hygiene and Tropical Medicine and the KEMRI-Wellcome Trust Research Programme in Kenya also participated. The Gates Foundation funded the study, and PATH sponsored it.