Key takeaways
- Zero Copy federates access across AWS Glue Iceberg tables and Amazon Aurora, while Hyperforce on Google Cloud prepares for broader enterprise migration in Q4 2026.
- New operational integrations include the AWS DevOps Agent in Slack, bidirectional voice over WebSockets with Amazon Connect, and commerce purchasing via the Universal Commerce Protocol.
At Dreamforce 2026, Salesforce expanded strategic integrations with Amazon Web Services (AWS) and Google Cloud to eliminate enterprise data silos and operationalize Salesforce AI agents across multi-cloud environments. The partnerships allow third-party agents, such as AWS’s Amazon Quick and Google’s Gemini Enterprise, to access live customer relationship management (CRM) records and execute business actions natively via the open Model Context Protocol (MCP). Concurrently, Salesforce is broadening foundation model options for its Agentforce platform—incorporating models from Anthropic, OpenAI, and Nvidia through Amazon Bedrock alongside Google Gemini—while expanding zero-copy data federation across AWS Glue, Apache Iceberg, and Google Cloud Hyperforce.
Bringing live CRM context to Amazon Quick and Google Gemini
Enterprise AI deployments often encounter friction when autonomous systems hit siloed databases and proprietary application boundaries. Under the expanded partnership with AWS, Salesforce business context and actions are exposed directly to Amazon Quick, AWS’s business agent, through the Model Context Protocol (MCP). This standardized interface allows Amazon Quick to inspect real-time CRM data and perform autonomous tasks on behalf of sales and customer support teams without requiring data migration or bespoke API connectors.
The AWS collaboration also embeds operational automation into enterprise communications. Specialized frontier systems, starting with the AWS DevOps Agent, are being integrated directly into Slack. Engineering teams can triage and remediate cloud infrastructure alerts within Slack channels, avoiding the context switching typically required across cloud administrative consoles.
Salesforce is establishing an equivalent architecture with Google Cloud. Utilizing a headless framework built on MCP, Google’s Gemini Enterprise platform connects directly to Salesforce environments. Without configuring manual data pipelines, enterprise teams using Gemini Enterprise can execute pipeline reviews, summarize accounts, and triage service cases across sales, support, and finance workflows. The integration also brings Tableau analytics into Gemini, grounding conversational answers in verified enterprise metrics.
Additionally, Salesforce Commerce Cloud will connect to the Universal Commerce Protocol (UCP). This framework enables shoppers to browse product catalogs and purchase items directly within Google Search or the Gemini interface, while checkout processing and security controls execute through Salesforce’s backend infrastructure.
Zero-copy architecture and multi-model flexibility across hyperscalers
A central objective of the multi-cloud initiative is reducing the security risks, latency, and storage overhead of continuous data duplication. Salesforce is extending its Data 360 Zero Copy integration to support AWS Glue-managed Apache Iceberg tables and Amazon Aurora databases, enabling AI models to ground actions in transactional records without moving underlying datasets.
| Integration Surface | Amazon Web Services (AWS) | Google Cloud |
|---|---|---|
| CRM Data Grounding | Data 360 Zero Copy with AWS Glue (Apache Iceberg) and Amazon Aurora | Headless MCP architecture and Tableau analytics within Gemini Enterprise |
| External Agent Access | Native Salesforce context and actions inside Amazon Quick via MCP | Direct pipeline reviews, case triage, and account summaries in Gemini |
| Agentforce Model Catalog | Amazon Bedrock foundation models (Anthropic, OpenAI, Nvidia) | Native Google Gemini model deployment within Agentforce |
| Workplace & Channel Execution | AWS DevOps Agent in Slack; bidirectional audio via Amazon Connect | Universal Commerce Protocol for buying inside Google Search and Gemini |
For organizations running Salesforce Agentforce, the AWS partnership introduces direct access to Amazon Bedrock’s foundation model catalog. Enterprise architects can configure task-specific models from Anthropic, OpenAI, or Nvidia based on operational latency, cost, and accuracy requirements. Similarly, Google’s Gemini models are being brought natively into Agentforce, offering model flexibility on Google Cloud infrastructure.
AWS Vice President of Data and AI GTM Rahul Pathak stated that enterprises require data access without migration, model choice at sustainable economics, and AI capabilities situated where teams work. However, as autonomous agents execute tool sequences across cloud boundaries, governance frameworks become vital. Policies such as the AWS open-source Dogwood policy highlight how technical leaders must establish strict boundaries on automated tool calls before granting agents execution access to enterprise datastores.
Hyperforce on Google Cloud, bidirectional voice, and operational risk
To support distributed agent workloads, Salesforce is expanding its public cloud infrastructure deployment. Hyperforce running natively on Google Cloud is already processing live production customer traffic, with broader customer migrations scheduled for the fourth quarter of 2026.
Salesforce President and Chief Trust and Infrastructure Officer Meir Amiel stated that running Hyperforce on Google Cloud maintains identical security, compliance, and resilience standards while providing cloud flexibility. Google Cloud President of Global Partner Ecosystem Kevin Ichhpurani noted that bringing business context into Gemini Enterprise allows customers to deploy trusted agents that reason and act securely.
Real-time voice interactions are also expanding through Agent2Agent bidirectional audio streaming over WebSockets, linking Agentforce Voice with Amazon Connect. This configuration allows customer service operations to handle voice routing and agent assistance with lower latency across contact centers.
For enterprise architects, managing distributed agents across AWS and Google Cloud introduces operational complexities. Organizations must govern identity federation, monitor WebSocket stability across voice channels, audit IAM role boundaries on MCP endpoints, and manage API inference costs across disparate foundation models.
What happens next
Salesforce confirmed that broader customer migrations to Hyperforce on Google Cloud will begin in the fourth quarter of 2026. IT directors and infrastructure teams planning adoption should structure their rollout across three immediate steps:
- Audit Data 360 Zero Copy datastores: Review Amazon Aurora configurations and AWS Glue Iceberg schemas to verify compatibility before connecting external agents for runtime data retrieval.
- Establish Model Context Protocol controls: Implement role-based permissions and strict boundary filters on MCP endpoints before granting tools like Amazon Quick and Gemini Enterprise access to CRM objects.
- Review messaging and voice administration: Establish workspace governance guidelines prior to enabling the AWS DevOps Agent in Slack, and evaluate network capacity for Amazon Connect WebSocket audio.
