CATEGORY: METABOLIC & BLOOD SUGAR
The human coaches didn’t win. In a year-long trial of 368 adults with prediabetes, blood sugar running high but not yet in the diabetic range, 31.7% of those given a fully automated AI coaching app met their weight, activity or blood-sugar goals. Among those given trained human coaches, the long-standing gold standard, 31.9% did.1
The Same Course, With Nobody on the Other End
Everyone in the trial was referred to the Diabetes Prevention Program, the year-long lifestyle course descended from a landmark 2002 trial in which diet and activity changes cut new diabetes cases by 58%.2 A computer draw settled the format each person got. One version was delivered remotely by human coaches. The other ran entirely through a phone app paired with a smart scale, with the software tailoring lessons and nudges on its own. No person watched over it. Success was defined up front as clearing any one of three bars by month 12: losing 5% of body weight, losing 4% while logging 150 minutes of weekly activity, or trimming HbA1c, a blood test reflecting roughly three months of sugar levels, while staying out of the diabetic range.
Where the App Quietly Pulled Ahead
The researchers built the study to ask whether the app was any worse than human coaching. It wasn’t, and the result cleared their preset fairness margin with room to spare. The app had one real edge: more people actually started, 93% versus 83%, and more finished. That matters. A program with no waitlist, no schedule and no staffing limit removes several of the reasons the millions of adults with prediabetes never begin one at all. What a trial this size can’t settle is whether the two formats stay even over many years, or who does better with a human voice on the line.
Key Takeaways
- A fully automated, AI-led version of the Diabetes Prevention Program matched remote human coaching over 12 months: 31.7% versus 31.9% of participants reached their weight, activity or blood-sugar targets.
- Matching the coaches isn’t the same as beating diabetes; roughly two-thirds of both groups fell short of their goals, so the format question is settled more firmly than the prevention one.
References
- Mathioudakis N, Lalani B, Abusamaan MS, et al. An AI-Powered Lifestyle Intervention vs Human Coaching in the Diabetes Prevention Program: A Randomized Clinical Trial. JAMA. 2025;334(23):2079–2089. link
- Knowler WC, Barrett-Connor E, Fowler SE, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002;346(6):393–403. link