Sage

AI supports fall recognition and care insights within a broader operational platform. Sage Detect interprets room camera input and sends apparent fall events to caregivers, while Sage Insight turns care activity into staffing and resident support information. Nurse call hardware, mobile task recording, and ordinary resident alerts also serve practical purposes without relying entirely on model inference.

The product therefore combines AI dependent features with conventional coordination infrastructure. Its value includes both recognizing events that residents may be unable to report and giving staff a common place to receive, act on, and record care requests.

Fall monitoring runs automatically, and alerts continue escalating until a caregiver is physically with the resident. Communities can deactivate monitoring, and consent determines whether a bedroom camera is active. Conventional call buttons and check ins remain part of care outside the camera's coverage. These are meaningful operational controls, but the current Detect page also promises that no fall goes undetected without qualifying that promise.

Insight similarly describes every recommendation as accurate. Such absolute language leaves missed events and incorrect recommendations inadequately bounded, even though staff retain responsibility for responding to alerts and deciding what care to provide.

Detect uses vision language models to interpret resident behavior from camera input. Its launch description connects this approach to fall detection, activity patterns, and information for care teams, going beyond a generic claim that the product uses AI. Insight separately analyzes observations, care tasks, and alerts to identify changing needs and staffing patterns. The precise production models, supplier identities, training populations, and version history remain unspecified.

As a result, the technical approach is recognizable, while the information needed to reproduce performance or distinguish changes between model releases is limited. Operational features alone do not establish the underlying model's validation.

The product describes proprietary detection technology built with vision language AI, but it does not identify the production model providers. The AI policy prohibits resident data from training external models, indicating that external model handling is relevant to the service's safeguards. The general privacy policy lists categories of technology and service providers rather than a product specific processor register.

Named inference suppliers, hosting processors, regional processing paths, and downstream retention terms remain unspecified. The available information therefore explains some restrictions on data use without providing a complete chain of organizations handling camera input or generated care insights.

A November 2025 CountryHouse Omaha case describes five months of Sage Detect use beginning earlier that year. Average fall response time fell from nearly 15 minutes to 6.6 minutes, alongside reported reductions of 50% in falls and 75% in related hospitalizations. Staff combined alerts with incident review and changes to individual care routines. This is a named, timed operational comparison with a response time baseline.

Resident counts, event denominators, and a concurrent comparison group are absent, limiting causal interpretation. The findings support that community's experience but do not establish independent validation or equal performance across other facilities.

Resident data cannot train external AI models under the April 2026 AI policy, while deidentified information may improve the service. Detect's launch materials specify automatic video deletion after 30 days and restrict incident footage to approved staff. The current resident explanation describes short clips, privacy blurring, and no human watching of live feeds. These controls address a particularly sensitive camera workflow.

They leave the retention schedule for other care records and the technical response to unsafe model outputs unspecified. Facility consent, limited access, and short video retention therefore provide concrete protections with an incomplete account of the wider data lifecycle.

The service agreement treats resident PHI as information Sage processes as a business associate under an executed BAA. It also requires BAAs with service providers receiving identifiable resident data. These are service specific obligations beyond general website privacy language. The April 2026 privacy policy separately excludes information governed by HIPAA and data processed for facilities, directing those relationships to their service agreements.

The actual customer BAA and its detailed coverage remain separate from the accessible master agreement. Consequently, the available terms establish the business associate relationship without resolving every product, trial, or subcontractor condition.

Sage's own site carries a SOC 2 Type II claim alongside healthcare privacy branding. That is a recognized attestation claim for the vendor, while the report itself and its system boundaries are not openly available in the reviewed material. The current audit period, covered products, exceptions, and report access conditions remain unspecified. Operational controls such as approved access to incident clips provide additional context but cannot substitute for the audit artifact. The scope matters because the platform combines connected devices, care applications, analytics, and camera monitoring, potentially involving different components and service providers.

Detect recognizes possible falls and supports timely intervention, while Insight highlights changes that may precede hospitalization or resident decline. Those functions extend beyond simple scheduling or record storage because their intended value includes safety monitoring and clinical attention. The service agreement reserves medical authority to the facility and describes information as general support rather than professional medical advice.

The regulatory status of the current fall detection and early warning functions remains unspecified. A contractual disclaimer alone does not explain how those functions are classified or identify a clearance, authorization, or device exemption applicable to the deployed models.

The April 2026 AI policy describes a staged release process involving clinical and engineering review, selected community pilots, testing across varied environments, and optional adoption. Communities can decline AI features without losing support, and staff retain clinical judgment over synthesized information. This gives a practical account of testing and rollout responsibility.

A named governance framework, measurable release thresholds, current subgroup results, and independent behavioral assessment remain unspecified. Testing in different environments may expose operational variation, but it does not establish comparable performance across resident demographics or provide a formal account of how bias findings change a production model.

Detect commits to alerting staff within 45 seconds after recognizing a fall. The starting point is recognition, so this commitment does not promise detection of every event or a fixed time from the physical fall itself. Service terms provide limited indemnity for Sage's contractual breaches or wrongful acts while disclaiming error free operation and prevention of falls or injuries. The facility retains authority over medical care.

These provisions create a measurable notification commitment and some contractual recourse, but they do not establish a general guarantee against missed detections, incorrect insights, or the resulting clinical harm.

Named integrations include ALIS, ECP, and Yardi. The November 2025 integration update describes resident profile and care level synchronization, with timing varying between systems. The current tasking product adds bidirectional EHR connections and point of care task recording, joining planned work with unplanned care activity. This supports more than a generic claim of interoperability.

A named production example tying a particular EHR to the complete read and write workflow remains unspecified. Older planned dates for PointClickCare and August Health also leave current availability unresolved unless supported by a present product or deployment statement.

Room hardware and staff phones connect to hosted care applications and management dashboards. Implementation includes configuration, integration with existing systems, and team training; the service agreement places network requirements in the statement of work. Camera monitoring adds a distinct stream of sensitive room information alongside ordinary care events and task records. Processing regions, tenant separation, and the location of video inference remain unspecified.

The deployment description therefore explains the facility equipment and staff workflow more clearly than the backend architecture, including whether any processing, support access, or stored incident data crosses national borders.

Purchases use a proposal and statement of work covering software, hardware, and additional features. The June 2026 agreement specifies $300 for an extra phone beyond the included allocation, plus its data charges. It also describes a 45 day initial trial, with trial fees becoming payable if use continues. These terms explain some purchasing mechanics and incidental costs. The base subscription, installation charge, and full cost for a particular community remain unspecified, so the phone amount cannot establish a representative platform price.

The platform serves independent living, assisted living, memory care, and skilled nursing organizations. Its workflows address resident calls, falls, wandering, planned care tasks, and staffing oversight rather than a broad range of hospital specialties. CountryHouse Omaha provides a named memory care example, with Agemark subsequently expanding Detect to additional communities. Other named senior living operators describe nurse call, documentation, and management use.

This supports breadth within senior care, although the strength and purpose of each reference differ. Comparable clinical outcome studies across independent living, assisted living, and skilled nursing populations remain unspecified.