You're trusting this with client records and Medicaid claims, so you should ask hard questions. Here are the ones we hear most, answered plainly.
No. It writes drafts. Your caregiver approves the visit note, your nurse signs the care plan, your office clicks the action. If nobody approves, nothing happens - and every approval is logged with who and when.
When it isn't sure, it says so instead of guessing. An ambiguous medication mention gets flagged for your nurse rather than smoothed into the record. Your staff can edit or throw out any draft.
No. Clock-ins, EVV records, and payroll math run on plain rules - there's no AI anywhere in that path, so nobody's paycheck ever depends on a model.
Your staff, and that's it. We don't train models on your records and we don't sell data. Trial workspaces are isolated from each other, and production deployments run single-tenant.
Real client data connects only under a signed agreement, including a BAA. In production, AI calls go through BAA-covered endpoints, access is role-based, and there's an audit trail of every approval.
We attack it on purpose - fake protocol names, instructions hidden inside note text, gibberish input - and keep the logs of every run. Ask us and we'll walk you through them.
Diagnose, prescribe, decide eligibility, or message a family or a payer on its own. The demo runs entirely on fictional people; your real data connects only after paperwork is signed. See how connections work →