How Leo Handles Customer Data

Last updated: September 25, 2026

Leo is Paramark’s AI assistant for answering questions about your Paramark workspace. It helps you explore your marketing measurement data, reports, experiments, Paramark Live, and Scenario Planner by using approved Paramark tools and summarizing the results in a clear way.

This article gives a high-level overview of what data Leo can use and how Paramark keeps that data safely scoped to your organization.

What Leo can help with

Leo is designed to answer questions about your own Paramark workspace. Depending on what is enabled for your account, Leo can help with things like:

  • Understanding marketing mix model reports, model outputs, channel contributions, ROI/CPA, spend, revenue trends, and analyst notes

  • Reviewing incrementality experiment designs, power curves, and lift results

  • Summarizing published reports and report notes

  • Reading or updating Scenario Planner inputs and outputs when you explicitly ask Leo to do so

  • Using recent shared Slack context when the Slack surface is enabled and connected to your account

  • Using call transcripts and summaries associated with your organization

  • Referencing Paramark help content, product documentation, and white papers

What kinds of data Leo uses

Paramark’s core data is aggregate marketing measurement data: things like spend, impressions, conversions, revenue, channels, campaigns, and dates. Leo does not use consumer or end-user personal data.

The personal data involved in a Leo conversation is typically limited to business contact information, such as the name and work email of the person asking the question, plus any names or work contact details that appear in collaboration sources your organization has connected, such as shared Slack messages or call transcripts.

How Leo keeps customer data isolated

Leo is scoped to your account. Its tools are set up to work within the customer environment connected to the request, so it is focused on helping you with your own Paramark data.

In practice, this means:

  • Single-account tool access: every data tool Leo uses is bound to the customer tenant for the request.

  • Authenticated requests: Leo uses the signed-in user and account context to determine what information is available.

  • Account-scoped history: conversation history is keyed to the account and user, so Leo is given the asking user’s own conversation context.

  • Connected-source boundaries: when Leo uses Slack, scope comes from the linked customer channel or the verified sender. When Leo uses call context, it is matched to your organization.

  • No cross-customer browsing: Leo is designed to work inside your Paramark workspace, not across unrelated customer environments.

Where Leo runs

Leo runs inside Paramark’s application environment and uses approved infrastructure and vendors to provide answers, all within the SOC 2 framework.

  • Paramark systems: Leo works through Paramark’s application and data systems. Conversations are stored in Paramark’s database environment using standard cloud security controls, including encrypted connections and encryption at rest.

  • Google Cloud Vertex AI: Leo may use Claude through Google Cloud Vertex AI. These model calls are made from Paramark’s Google Cloud environment. Google does not use prompts or completions to train models, and does not share them with the model provider.

  • Langfuse: Paramark may use Langfuse Cloud US for tracing, debugging, and answer-quality monitoring. This helps the team understand when an answer needs improvement and keep Leo reliable over time.

How Paramark keeps data safe

Leo is built on the same security foundation as the Paramark platform. In practice, that means:

  • Account-level scoping: Leo is designed to use the data connected to your Paramark account, not unrelated customer environments.

  • Authenticated access: requests are tied to the signed-in user and their account context.

  • Purpose-limited use: customer data is used to provide, support, and improve the Paramark product experience.

  • Careful vendor use: Paramark uses approved vendors for the infrastructure needed to run the product.

  • Operational monitoring: tracing and logs help Paramark troubleshoot issues, improve answer quality, and maintain reliability.

Model training and analytics

Leo’s model calls through Google Cloud Vertex AI are not used by Google to train models. Paramark also structures product analytics so chat message text is not sent to product analytics vendors.

Paramark may use aggregated and de-identified usage information to improve product features. This helps improve the product while keeping customer-specific content protected.

Retention

Leo conversation history may be retained in Paramark systems so users can keep context and Paramark can support the service. Tool inputs and outputs are not stored as full conversation records; Paramark stores the message text and a display-level record of which tools ran.

Customers can ask Paramark to delete Leo history and associated traces. At account termination, Leo history is deleted along with the rest of the customer’s data, subject to any legal or operational retention requirements.

Compliance

Paramark maintains SOC 2 Type II compliance.

Leo operates within Paramark’s broader platform environment and follows the same security, access control, change management, logging, vendor review, and compliance practices used across the Paramark product.

In short

Leo is meant to make your Paramark data easier to understand. It works within your account, uses approved tools and infrastructure, and is supported by Paramark’s existing security and compliance program. The result is a helpful assistant that can answer questions while keeping customer data handled safely and responsibly.