Do we use client data to train models?
No, customer data is ever used to train IMPLAN or public models.
Do we retain client data in our AI systems?
We do not retain client data within our AI systems. We temporarily retain some usage information to feedback tracking and to improve system prompting, but never for model training.
Those items and areas are:
- Customer prompt inputs along with user identifiers are temporarily stored for 2 weeks within Langsmith (an agentic workflow manager), and 4 weeks within Datadog (a real time product monitoring platform).
- User submitted feedback is stored within our Aurora databases.
Do we run our own models?
We access base models from Anthropic that are hosted & provided via AWS Bedrock. The model runs within AWS cloud but is a private instance of an Anthropic offering.
Our conversational analytics Data Library feature is accessed via an embed that connects to a private GCP cloud instance of Gemini, which in turn reads from our private redshift databases.
Where are our models hosted?
We host our models in AWS and Google Cloud Platform in the US,
Can we disable AI features?
Yes, AI features must be enabled and can always be disabled in the account settings. When AI features are disabled, there is no system or user interaction with AI systems in any way.