Adaptive hydration prompts.
A reminder to drink water every two hours becomes wallpaper. Companion picks the moment, the wording, and sometimes the offer.
Hydration is one of the smallest interventions in long-term care with the largest downstream effect. UTIs, falls, kidney strain, and a lot of late-afternoon confusion all trace back to a resident who drank too little during the day. The clinical math is settled. The product question is what to actually say, and when.
A scheduled prompt every two hours sounds reasonable on a whiteboard. In a real room it fails three ways. It fires while the resident is already drinking. It fires while she's asleep. Or it fires for the fourth time today with the same wording, and she stops hearing it. The device becomes noise, which is worse than silence.
Prompt selection as a small policy.
Companion treats each prompt as a decision, not a timer. It looks at recent intake, time of day, what the resident just declined, and her current behavioral state — resting, watching TV, mid-conversation, asleep. The output isn't whether to prompt, it's which prompt, and whether to wait.
A resident who waved off water at 10am doesn't get the same line at noon. Different wording, a different offer — I have ginger ale in the fridge, or your daughter mentioned you liked iced tea — and a lighter tone. If she's napping at 2pm, the prompt slides to 3. The cost of a missed prompt is small. The cost of a rejected one is that the next one lands worse.
What the nurse sees.
Intake goes up, gently, because the prompts feel like a roommate offering rather than a machine reminding. The nurse sees a cleaner intake log on the floor dashboard, with the offers Companion made and which ones the resident accepted. The next thing we're shipping is per-resident preferences — preferred drinks, quiet hours, and a way for family to suggest the line that actually works.