Skip to main content

​
How Tempo treats your data

  • Read-first by default. Tempo reads from your systems to build context. The writes it makes back to your CRM follow the update behavior set on each field. See Fields and CRM objects.
  • Your data is yours. We do not train foundation models on it.
  • Scoped access. We ask for the permissions each connection needs. Every integration page lists the exact scopes Tempo requests and what it does with them.
  • Tempo does not send email. Tempo reads mail to build context. It has no code path that sends mail on your behalf. Google Workspace connections request the gmail.compose scope; see the Google Workspace page for what that scope is and is not used for.

​
What Tempo writes to your CRM without a review step

Most CRM writes go through a review. These do not:
  1. Fields set to Automatic. When an organization admin sets a CRM-synced, updateable field’s update behavior to Automatic, Tempo applies each proposed value and writes it to the CRM. No person reviews it first.
  2. Per-user Automatic. When an admin turns on Allow per-user field settings, a seller can set a field to Automatic for their own deals. Proposals on that seller’s deals then apply without review.
  3. Record creation. New Opportunity in the Tempo app and the create_crm_object tool over MCP create accounts, contacts, and opportunities in Tempo and queue the create to your CRM. There is no review queue for creates.
Two guards apply to automatic field updates: Tempo does not automatically apply a proposal to a closed opportunity, and it does not generate proposals from a batch whose evidence is all older than 30 days. Fields set to Suggest are written only when a person approves the proposal, in the Tempo app, over MCP, or from a Slack approval card. Fields set to Manual generate no proposals.

​
Where your data flows

Tempo uses the data synced through its integrations, such as emails, calendar events, call transcripts, Slack messages, documents, and CRM records, in prompts to large language models (LLMs). This is how Tempo forms observations and proposes CRM updates.

​
Models and providers

Tempo uses large language models from multiple providers to process the data synced through your integrations. We reevaluate which models we use as capabilities and costs change, so the specific set shifts over time. If your security team needs the current model list or our subprocessor list, get in touch and we will send it to you.
Data collected by Tempo is not used to train models.

​
What each integration accesses

Permissions are scoped per connection. Each integration page lists the OAuth scopes Tempo requests, who connects it, and what Tempo does with the access.

Salesforce

Two scopes, the organization reader and personal connections, and CRM write behavior.

HubSpot

Read, schema, owner, and write scopes, plus the optional sales email scope.

Google Workspace

Gmail, Calendar, and Meeting Transcripts scopes, including gmail.compose.

Slack

Bot scopes, per-member reader access scopes, and the three capture modes.

Granola

Granola-hosted consent and encrypted token storage.

Gong

Four read scopes, or an API key.

Fathom

One read scope, per member.

Notion

Page-level access you choose in Notion.

BigQuery

Read-only roles on the datasets you choose.

​
Questions from your security team

If your security team wants more detail before approving a connection, we are happy to set up a call. Get in touch.