AI Meal Insights
What exactly will my patient see?
A 1–2 sentence, non-judgmental insight under their logged meal (e.g., “Nice protein, add a veggie or chia to boost fiber for longer fullness.”) and a highlight + opportunity in their meal details. Patients can turn it off anytime in Settings.
What do I see in in the patient’s Journal?
A clinician mirror view under each meal: the exact insight the patient saw and the meal insights with a quick ability to leave feedback.
Are these insights replacing RD coaching?
No. They’re brief, patient-facing nudges to make logging useful between sessions, that are based on your session guidance and interventions. Your plan and judgment remain the source of truth.
What if the insights conflicts with my plan?
This should not happen, given the chart note is the main input for the insight. If you feel an insight is wrong or conflicting, you can flag the insight and leave more details on why it wasn’t appropriate. We regularly review and update the system based on this feedback.
Who’s excluded from the feature?
Patients with sensitive conditions (e.g., eating disorders, disordered eating, emotional eating) or anyone who disabled the feature.
How are you reducing risk up front?
RD/Patient feedback: RDs and patients can flag anything contradictory or off-tone; every flag routes to review and model tuning
Sensitive conditions & topics excluded: patients with ED, DE, and emotional eating do not receive these insights. Neither do some sensitive topics such as commentary on alcohol.
Safe phrasing: engineered to avoid judgmental language, respect dietary restrictions, and defer to RD chart context when relevant
Pre-screened: recommendations are screened with evaluation tools to catch inaccuracies/contradictions before rollout
Internal testing first: we’ve conducted extensive internal testing to validate usefulness, tone, and safety
Gradual rollout: slow, incremental launch so we can monitor performance and adjust quickly
Transparency + control: always visible as AI, and patients may opt-out at any time
How is feedback used?
Patients and RDs can flag a comment right next to the 'Insight.' We review regularly, remove unsafe insights quickly, and use aggregate signals to refine suggestions.
How should I interact with the insights?
This is designed to be an easy addition to support your work. You can use, ignore, or flag the insight as appropriate.
How is personalization handled?
The model considers the latest meal (foods/macros/time), patient goals/preferences, recent patterns (e.g., recurring protein gaps), medical history, and relevant RD notes (e.g., dietary restrictions).
Does any of this change my care plan automatically?
No. Insights never edit chart notes.
How can I improve the insights?
Share your feedback using the thumbs up or down! We try to incorporate common RD patterns and language into insights. We do weekly safety/quality reviews, and fixes will be launched continuously.
How can I disable this for a patient?
Patients can disable the feature by going to Settings and disabling Meal insights.
