AI that gives nurses time back.

Family members aren't asking "is mom okay" because they don't trust your staff. They're asking because the alternative is sitting with the silence. A 60-word evening summary turns that anxiety into a daily exhale — without adding a minute to anyone's shift.

Tendara generates each summary by reading the day's structured chart entries — meals consumed, medications administered, vitals captured, activity participation, behavioral notes. The model then writes a warm, plain-English one-liner appropriate for a family member, not a clinician.

Every summary is reviewable before send. Facilities can opt into a 30-second nurse-of-shift glance-and-approve workflow, or fully automated send with a daily quality audit. The model never invents a fact — every claim traces back to a chart entry.

Most vitals-monitoring systems alert when a number crosses an absolute threshold — BP > 160, HR > 100, etc. That's far too late. The warning is in the trend, not the number.

Tendara's model learns each resident's personal baseline — sleep, BP variance, weight, oxygen saturation, activity — and flags meaningful deviations. A resident whose resting HR is normally 68 ± 4 starting to trend at 78 ± 6 is a clinically significant signal weeks before any threshold-based alert would fire.

Alerts route to the on-shift nurse with the trend chart attached and a one-tap "escalate / monitor / dismiss" action. Every dismissal trains the model on what's normal for that specific resident.

Every facility's biggest staff complaint isn't workload — it's interrupt-driven workflow. A family's billing question shouldn't pull a charge nurse off the floor. A family's clinical question shouldn't sit in a billing inbox for two hours.

Tendara reads each incoming family message and routes it: clinical questions go to the on-shift nurse for that resident's care plan, billing questions go to administration, scheduling goes to the front desk, and general check-ins go to the resident's care coordinator. Routing happens in under a second; the recipient sees a one-line summary plus the original message.

Edge cases (urgent multi-topic messages, unclear intent) escalate to the care coordinator with all context preserved. Misroutes are correctable in one tap and become training data.

Charting is the single largest time sink for direct-care staff. Most software assumes staff have time to type structured entries; reality is stolen-moments documentation, often after the shift ends.

Tendara's voice-to-chart accepts free-form spoken notes ("Just gave Dorothy her 2pm metformin, she had a small bruise on her left forearm — looks old, asked her about it, she said she bumped the bedside table yesterday, no skin tear, monitored for 5 min") and structures it into the right chart locations: medication administered, skin observation, resident interview, monitoring period.

All entries are reviewed by the staff member before commit; the model never auto-files anything that wasn't said. Average time to chart a 15-minute task drops from ~4 minutes of typing to ~30 seconds of speaking + 20 seconds of review.

Most preventable adverse events have a 1-4 week signature in the data: subtle gait changes, declining engagement, sleep fragmentation, weight changes. The signal is there; the human attention to spot it across 142 residents is not.

Tendara's risk model produces daily scores per resident across four dimensions: fall risk, hospitalization risk (within 30 days), pressure-ulcer development risk, and re-admission risk for residents who recently returned from acute care. Each score has a confidence interval and a list of contributing factors.

Scores are surfaced in the resident card, on the shift handoff report, and as a weekly DON-level summary highlighting residents whose risk has materially changed. The goal is never to replace clinical judgment — it's to make sure the data the clinician would have considered is actually surfaced.