AI model choice and private AI for your organization
TL;DR
Pantheon's assistant answers questions over your own conversations: who to follow up with, what a thread was about, what someone said last time. You choose the model behind it, from OpenAI or Anthropic. Your data is never used to train third-party models. Other models, such as Grok or a model you host yourself, are added on request, and so are models that run inside your own deployment.
On this page
What the assistant does
Ask it the questions you would otherwise answer by scrolling: who did I promise a follow-up to this week, what did we agree in that group, when did this person last write. It answers from your own conversations, notes and records.
Your model, your choice
Pick the model that answers: one from OpenAI or one from Anthropic. Firms often have an approved list already. If yours names another model, such as Grok or a model you host yourself, we add it on request. Whatever model you pick, your data is never sold or used to train third-party models.
AI that stays inside your deployment
Some organizations cannot send conversation content to an outside model at all. For them, the answer is a model that runs within their dedicated deployment, alongside the policy for which features may use it. That is built on request.
What is standard and what is on request
Standard means it is part of every Pantheon Enterprise deployment today. On request means we build it for your deployment once you ask, and support it from then on.
| Capability | Availability | What it means |
|---|---|---|
| Assistant over your own conversations and notes | Standard | Answers from your own conversations, notes and records. |
| Choice of model | Standard | OpenAI or Anthropic models. |
| Other or self-hosted models | On request | Such as Grok, or a model your firm hosts. |
| No training of third-party models on your data | Standard | Never sold, never used for training. |
| Approved-model list for your organization | On request | Limit the choice to the models your firm has approved. |
| Models inside your deployment | On request | Conversation content never leaves your deployment. |
Which AI models can we use?
OpenAI and Anthropic models as standard. Others, such as Grok or a model you host yourself, are added on request. An organization can also limit the choice to its approved list on request.
Is our data used to train AI models?
No. Your data is never sold or used to train third-party models.