Zesty AI¶
Last Modified: August 27, 2026
This policy describes Zesty AI: what it does, what it deliberately does not do, and how conversation data is handled. It forms part of our Privacy Policy.
What Is Zesty?¶
Zesty is an AI feature offered as an optional add-on to Wellzesta Life. It is not a standalone product. It works inside Wellzesta Life, and only for communities that have chosen to enable it.
An end-user interacts with Zesty conversationally. With respect to the API, Zesty acts as a proxy for the end-user. She possesses the same read and write permissions (and limitations) as the end user she is representing.
What Zesty Can Do¶
Zesty's scope is narrow by design. She helps with:
- Events — finding what is happening, and registering or unregistering
- Directory — looking up residents and staff in the community directory
- Messages — reading the resident's own inbox and marking a message as read
- Documents — finding community documents published in Wellzesta
Just as an end user can only access information for which they are authorized — in terms of community, group, relation, or person — Zesty, as a proxy for the end user, inherits the access permissions of the end user.
What Zesty Does NOT Do¶
- No medical or clinical advice. Zesty does not give medical, clinical, medication or health advice, and does not answer clinical questions. It is not a medical device, and it is not a substitute for the community's care team, a nurse or a physician.
- No decisions about care. Zesty does not make, recommend or influence any care decision.
- No emergency handling. Zesty is not a way to summon help. In an emergency, residents should use their community's normal emergency procedure.
- No opinions outside its scope. Zesty does not answer or offer views on subjects outside the scope discussed above.
- No access to other communities. Zesty cannot see or return data belonging to another community. Tenant isolation is enforced by the platform on every request, not by Zesty.
- No audio sent to Wellzesta. Speech is converted to text on the resident's own device — see Voice below.
- No use of conversations to train AI models. See Your Data below.
Voice¶
End users can speak instead of typing in Wellzesta Life. Speech is converted to text on the end user's own device. The audio is not transmitted to Wellzesta and is not retained anywhere. What reaches Zesty is text.
Your Data¶
Conversations are not used to train AI models¶
No data processed by AI at Wellzesta is used to train or improve AI models. Your conversations with Zesty are not training material.
We do keep conversation history, to improve Zesty as a product¶
We use conversation history for security and compliance and to improve our services. We do not use conversation history to train or improve AI models. Some examples of how we use Zesty conversations include:
- to understand when Zesty's response is inadequate
- to find and reduce incorrect or inaccurate answers
- to evaluate Zesty's "voice" and improve her social skills
| Training the AI model | Improving Zesty | |
|---|---|---|
| What it means | Using data as examples to teach the AI to answer better in general | Reviewing conversations to refine and improve the product |
| Does this happen with Zesty? | No | Yes |
| Who would do it | The AI provider | Wellzesta (the Zesty team) |
In short: end user ("your") conversations do not make the AI smarter in general. They do help our team make Zesty more useful and more accurate.
Only Wellzesta's own team can review a conversation, through an internal tool, for the product-improvement purposes described above. Access is restricted to the Wellzesta personnel who need it, under the same confidentiality obligations that apply to all customer data.
Conversation data receives the same protections as the rest of your information under our Privacy Policy, including the HIPAA safeguards described there.
Third parties¶
The AI model that powers Zesty is provided by Google, under a Business Associate Agreement. Google processes a conversation only to return an answer, and does not retain it. Google is named in our Subprocessors list, alongside every other third party that comes in contact with customer data on our behalf.
For security and compliance, Wellzesta logs all calls to the Wellzesta API. Zesty's calls to the API on behalf of an end user will be logged.
The Wellzesta API uses an error-monitoring subprocessor that is named in our Subprocessors list. If Zesty's interaction with the API triggers an error then the error will get logged.
Accuracy¶
An AI can state something that is not true. Zesty is built with automatic checks that compare what it says against what actually happened in Wellzesta — for example, whether an event registration really went through. When an answer does not match the real outcome, it is corrected before the resident sees it.
This reduces errors. It does not eliminate them. Residents and staff should confirm anything that matters — an event time, a registration, a message — in Wellzesta Life itself.
FAQ¶
Q: Can Zesty see medical records?
Zesty has no access to clinical or medical records held by your community.
Q: Can Zesty see another resident's messages?
No. Zesty reads only the inbox of the resident it is talking to.
Q: Does Zesty work without Wellzesta Life?
No. Zesty is an add-on to Wellzesta Life and is not sold or operated separately.
Q: Can a community turn Zesty off?
Yes. Zesty is optional and is enabled per community.
Q: What happens if Zesty gets something wrong?
At the end of each conversation, the resident can rate Zesty with a thumbs-up or thumbs-down and say whether something went wrong. That is the built-in way to report an incorrect answer. Incorrect answers are one of the specific things we review conversation history to find and fix. Anything that needs a response from our team can still be raised through the in-app support portal.