Air-Gapped Model
Propensity AI operates with a strict separation, or "air gap," between our databases and the AI itself. This includes any data customers sync to Propensity.
The AI does not have direct access to our databases or to synced customer data. Internal processing steps sit between any AI request and the database, and the AI has no visibility into that layer.
If a user types specific data points directly into a chat, that information becomes part of the conversation. This is different from the AI independently pulling or accessing data from a customer's database, which it cannot do.
Customer data is never used to train or fine-tune the AI model.
What the AI Accesses
The AI never accesses:
- Customer data
- Contact-level data
- Personally identifiable information (PII)
- Propensity's databases directly
The AI uses:
- Publicly available data
- A defined list of intent signals, used to generate recommendations such as campaign copy or strategy suggestions
Matching accounts to signals and identifying related contacts happens through separate internal processes, entirely outside the AI.
Data Retention
Data from user interactions with Propensity AI is retained for one year.
This data is used to:
- Execute the requested task
- Maintain a history of the conversation, so users can return to it later
Propensity may also review interactions internally to identify where AI agents are struggling to handle specific requests. This review is used to refine the prompts and configuration of those agents.
This internal review is separate from model training. Customer interactions are not used to train or fine-tune the underlying AI model, only to improve how Propensity's own agents are configured.
Audit Trail
All interactions with Propensity AI are logged, including:
- User chat history
- AI response history
- The tools and agents used at each step of the conversation, along with their responses
This audit trail is stored in Propensity's database. Access is restricted to Propensity's internal team.
CRM Sync vs. AI Processing
CRM sync and AI processing are completely isolated from each other.
When a user syncs CRM data to Propensity, that data does not flow into what the AI processes.