How it works
A nurse who's read everything. Not a doctor. Not a substitute for one.
Last updated 2026-05-03
KeptWell uses AI to do three things: turn medical paperwork into something readable, surface what changed since last time, and answer questions about your family member's record. Here is exactly what that looks like, and where it stops.
What the AI does
- Reads documents.
- Every PDF, scan, and photo you upload is processed by Claude (Anthropic's most capable model). For text documents (Word, RTF, transcripts), we extract the text directly. For audio recordings, we transcribe first.
- Pulls out the facts.
- Diagnoses, medications, dosages, lab values, test dates, doctor names, key findings. These get organized into a structured record you can search and filter.
- Writes a plain-English summary.
- Every document gets a title, a one-paragraph summary, and a list of key findings, all written in the voice of an experienced nurse explaining things at the kitchen table.
- Answers questions with citations.
- When you ask the chat a question, every claim it makes links back to the exact page in the exact document it came from. You can always check its work.
What the AI doesn't do
- It does not diagnose.
- It will tell you what a pathology report says. It will not tell you what your loved one has, or what stage, or what to do.
- It does not prescribe or recommend treatment.
- It will explain what a medication does. It will not tell you whether to take one.
- It does not replace your medical team.
- Use KeptWell to walk into appointments better-prepared. Don't use it to skip them.
- It does not work in an emergency.
- Call 911. KeptWell is for the slower, more confusing parts of care, not the urgent ones.
Where it gets things wrong
AI extraction is highly accurate but not perfect. It can:
- Misread a handwritten note (we flag low-confidence extractions for you to verify).
- Miss context buried in an unusual document layout.
- Occasionally summarize a complex finding in a way that loses nuance the original report carried.
Every document keeps the original file, the extracted text, and a confidence score visible. If something looks wrong, you can edit it directly. Your edits override the AI's interpretation.
The honest summary
AI is genuinely useful for the work of being a caregiver: reading hundreds of pages, surfacing what matters, answering questions at 11pm. It is not a substitute for the people who care for your loved one. We do not pretend otherwise.