Assistance for people and care teams
genusAI supports communication, information, drafting, and care coordination within agreed workflows.
How we use AI to support care
We use AI to help create useful content, explain care topics, support team learning, and connect people with resources.
Our goal is to save care teams time and help family caregivers find useful resources and stay connected to the person they support. These examples describe AI-assisted work; available features depend on the product and care program.
- Personalized content: drafting activities, stories, and plain-language guides shaped around a person’s interests and the care program’s purpose. For example, a team might prepare an activity guide around a favorite hobby.
- 3D animation and video: creating visual explanations, short videos, and orientation materials that make a service or learning topic easier to follow.
- Team training: preparing learning materials and practice scenarios to help staff talk through care situations and communication. The team reviews the materials before using them.
- Questions and resources: AI assistants help answer questions and find relevant information. For example, a family caregiver might look for support resources, or a staff member might need a guide from the program’s knowledge library. Important answers and resource details still need to be checked.
People first, with clear boundaries
Our AI requirements respect the person receiving care and the judgment of the people supporting them.
- Use AI only for the stated purpose and with information authorized for that workflow.
- The program is responsible for explaining where AI is used and which outputs require human review.
- Fairness, accessibility, and respectful language are responsibilities of the people designing, configuring, and using the service. AI can still produce biased or harmful errors.
- Access and sharing follow the program’s permissions and privacy obligations.
Information and model boundaries
We de-identify data before sending it to frontier models. Sensitive AI processing uses private genus models on our own hardware.
Frontier models are advanced AI models supplied by outside providers. We de-identify data before sending it to these models, removing or transforming information that could identify a person.
For AI processing of sensitive information, we use private genus models running on our own hardware at our office. The information used and who can access it follow the authorized care workflow.
The workflow and applicable agreements govern the information sent to a model and the permitted processing. Provider terms, including any training use and retention, are addressed for that workflow before sensitive information is introduced.
Removing direct identifiers alone does not necessarily meet legal de-identification requirements. Do not paste health information, passwords, or confidential records into a general-purpose AI tool or public website inquiry.
Information used for care programs remains subject to the applicable permissions, agreements, and privacy requirements. Use the program and privacy-request channels to request access, correction, or deletion of records.
Review before relying on or sharing output
The person responsible for using or sharing AI-assisted content must check its accuracy, audience, and fit for the situation.
- Check factual accuracy, source relevance, names, dates, and whether the output includes information you are not authorized to share.
- Clinical decisions belong with qualified professionals. AI output does not authorize treatment, determine eligibility, or replace professional judgment.
- Complete the workflow’s approval steps before sending AI-assisted content to a family, care team, or partner.
- Acknowledge uncertainty and seek a human explanation when an output is unclear or consequential. A response displayed by an AI tool is not evidence that a person has reviewed it.
Know when to pause
AI can make convincing mistakes, miss context, and reflect bias. It is not an emergency service.
- Responses may be incomplete, outdated, inaccurate, or inappropriate for someone’s circumstances. They are informational assistance, not a substitute for professional advice.
- A generated answer is not evidence that its cited source exists or supports it. Check important claims against an authoritative source.
- If the output seems unsafe or conflicts with a care plan, do not act on it without checking with the responsible care professional.
- For immediate danger or a medical emergency, call 911 or your local emergency number. Do not wait for an AI response, text reply, or website inquiry.
Report a concern or ask for a person
Use the program’s support channel or email our team with a brief description of the issue.
Include the product or program, what went wrong, and a way to reach you. Avoid sending identifiable patient details or a sensitive conversation transcript in your first email. Ask for an appropriate secure channel if more detail is needed.
We review reported AI concerns with the team responsible for the affected service. Contact your care team directly when the concern affects care. Questions about personal information follow the Privacy Policy and the applicable program’s request process.
NIST’s AI Risk Management Framework informs this policy’s approach to responsible use and human oversight. It provides voluntary guidance.