Einstein Conversation Insights Data Processing and Storage

The Salesforce Platform is the primary underlying infrastructure that powers Salesforce products. With data directly on the Salesforce Platform, you can access more tools, get better data reliability, and utilize services such as Hyperforce, which allows you to have more control over where these processes occur.

Is data ingested, processed, and stored on the Salesforce Platform?

Data that isn’t directly ingested, processed, or stored on the Salesforce Platform (also referred to as a non-Salesforce Platform service) resides on Salesforce-owned services and repositories that are integrated with the Salesforce Platform, but aren’t directly a part of it. These include companies or products that were acquired by Salesforce or otherwise developed separately from the Salesforce Platform.

*For video call transcript ingestion, the data is ingested on a non-Salesforce Platform service that is on Hyperforce, but the geographic locations available for customers are limited to the US and Europe.

Hyperforce and Data Storage

Data on the Salesforce Platform is available on Hyperforce. Hyperforce is the Salesforce trusted cloud platform that allows customers to manage where their data is processed and stored. That’s because Hyperforce infrastructure is composed of code rather than hardware, and data can be processed in locations worldwide. With Orgs on Hyperforce, Customer Data is stored in the country where the Org is located, provided that the Services the customer is using are available in that Hyperforce country.

Hyperforce is currently available in Australia, Brazil, Canada, France, Germany, India, Indonesia, Israel, Italy, Japan, Singapore, South Korea, Sweden, Switzerland, United Arab Emirates (UAE), the UK, and the US for products built on the Salesforce platform. Data 360 is available on Hyperforce in Australia, Brazil, England, Germany, India, Japan, and the US. To learn more about Hyperforce, see Introducing Hyperforce.

Many ECI services are now available on Hyperforce. The exceptions include:

• Data not ingested, processed, or stored on the Salesforce Platform unless indicated otherwise.

• Generative AI features: Model routing isn’t currently set up for ECI generative AI features, and the default routing is used. This includes the Call Summaries, Generative Conversation Insights, Call Explorer, and Sales Signals features.

• First-party data centers (Salesforce servers in data centers) aren’t on Hyperforce.

Salesforce Data Migration

Salesforce is migrating more Einstein Conversation Insights (ECI) data to the Salesforce Platform on a rolling basis starting in May 2026. All the information in this document covers data storage and processing details after the data migration. All orgs with ECI enabled after Spring ’26 already have their data on the Salesforce Platform.

For more information about ECI migration in general, see the ECI Migration article.

Data Storage by Artifact

ECI Existing Objects

These objects existed in the Salesforce Platform prior to Spring ’26.

New ECI-Related Objects on the Salesforce Platform

These objects are newly available and replace the non-Salesforce Platform service storage.

Call Insights (Calculated Results)

How they’re generated:

• Non-generative insights (keyword, predictive) are generated by the non-Salesforce Platform service and sent to Salesforce Platform via CIInsightsEvent platform event → new async MQ handler writes directly to Salesforce Platform entities.

• Generative insights (LLM-based) are triggered on save of transcript to Salesforce Platform DB → handler fetches transcript and insight metadata from Salesforce Platform → calls LLM Gateway → writes results back to Salesforce Platform insight entities directly.

For migrated ECI orgs, insights are no longer stored in Non-Salesforce Platform Cassandra (riq2 / signals keyspace).

Call Structure

How they’re generated: A non-Salesforce Platform service publishes a CICallStructureEvent → Core subscribes and enqueues to MQ → MQ handler writes the Call Structure JSON (speaker segments, offsets, talk ratios, speaker-to-CRM-record mappings) to Core.

For migrated ECI orgs, call structure is no longer stored in non-Salesforce Platform S3 or non-Salesforce Platform Cassandra.

Voice and Video Call Recordings

Call recordings are not stored by Salesforce, with the exception of voice calls and manually uploaded video calls. These call recordings (voice call MP3 and video call MP4 files) are still stored in a non-Salesforce Platform service / AWS S3, but the path has changed for migrated ECI orgs. The new path is: 

{bucket}/[sfdc]/org/{orgId}/VOICE/<callStartDatetime>/<sfdc_call_id>/recording/

The prior path encoded the salesforce_owner_id and non-Salesforce Platform service event key, which created a dependency on the external_key_to_event_key mapping table. The new path uses the Salesforce Call ID directly, removing that dependency.

Call recordings aren’t stored in the Salesforce Platform. S3 remains the storage layer for the raw recording files that we store. Most vendor-based calls are streamed from the vendor storage directly.