The high prevalence of type II diabetes in India necessitates novel public health tools to promote behaviors that lower the likelihood of developing diabetes and associated comorbidities. We introduce a public health text messaging system that learns from participant feedback to target each individual with messages that are most relevant to their current health behaviors. The participants who completed the AI-assisted text messaging intervention outperformed the control group in three out of eight categories. Highly engaged participants outperformed the control group in four out of eight categories and outperformed moderately engaged participants in seven out of eight categories, suggesting that engagement with the intervention is positively correlated with behavior change. We also perform a demographic analysis of the intervention results along the lines of sex, education level, and age. Recommendations for future work include personalizing message timing to increase intervention engagement and leveraging the plethora of data that is generated by the AI-assisted approach.