日本語 
Amazon Web Services (AWS) and Salesforce have announced an expanded collaboration aimed at helping businesses use AI across workflows. These workflows include sales, customer service and operational tasks. The expanded partnership connects Salesforce CRM data and business context with Amazon Web Services (AWS) tools and AI agents. It also expands model choices and data integration capabilities.
The companies said the collaboration addresses several challenges businesses face as AI adoption expands. These challenges include the need to access CRM insights wherever employees work. They also need to connect data stored across cloud environments without moving it. Additionally they want to extend AI agents into areas such, as real-time voice interactions and team collaboration.
Under the expanded collaboration, Salesforce business data, context, and skills can be accessed natively through Amazon Quick without requiring custom integrations. Sales teams can use Amazon Quick to review information such as pipeline health, customer profiles, service cases, and business analysis. The integration allows users to ask the AI assistant for current opportunity status, open service cases, and recent activity while retaining Salesforce context.
AWS is also bringing its frontier AI agents into Slack. The integration initially supports the AWS DevOps Agent, with AWS Security Agent, AWS FinOps Agent, and AWS Partner Central agents planned for expansion. Teams can interact with these agents directly within Slack conversations rather than switching to separate consoles when investigating incidents or reviewing security findings.
The companies are also expanding model options available through Amazon Bedrock for Salesforce Agentforce. Agentforce customers can access a broader selection of foundation models, including models from providers such as Anthropic and NVIDIA, directly within Salesforce agents and agentic workflows. OpenAI models already generally available through Amazon Bedrock are also scheduled to become available through Agentforce.
According to the announcement, customer data is protected through security and governance measures across the two environments. Amazon Bedrock applies zero data retention, while Salesforce’s Trust Layer provides capabilities including dynamic grounding and harmful-content detection. The companies also cite support for compliance requirements including HIPAA, PCI, SOC 2, and ISO 42001 for regulated industries.
The partnership is further expanding zero-copy data access, allowing systems to access enterprise data without moving or duplicating it. Salesforce Data 360’s Zero Copy capabilities are being extended to AWS Glue-managed Apache Iceberg tables, S3-based Apache Iceberg tables, Amazon Aurora, Amazon RDS, and Amazon SageMaker Lakehouse. These capabilities build on existing support for Amazon Redshift and Amazon S3 through private connectivity.
The companies are also expanding Informatica’s headless capabilities for Amazon Bedrock AgentCore and Amazon Quick. These integrations include functions such as Informatica platform management, catalog discovery and enrichment, data-quality scores, and master-data retrieval. The goal is to enable AI agents to work with clean, searchable, trusted, and governed enterprise data.
In customer-service applications, Salesforce and AWS are introducing interoperability between Agentforce Voice and Amazon Connect Customer. Using Agent2Agent (A2A) capabilities, the two platforms will support real-time, two-way voice communication through WebSockets, allowing AI agents across the two environments to work together more directly. This capability is scheduled for autumn 2026.
Also Read: Rebellions and ai& Partner to Bring Energy-Efficient AI Inference Infrastructure to Japan
Salesforce AI Research has also worked with AWS to optimize Salesforce’s proprietary voice model for AWS Trainium chips. The resulting Slack voice-input capability is designed to provide low-latency transcription, allowing users to record and share ideas through voice without leaving Slack threads. Optimizing the model for AWS-designed chips is intended to reduce inference costs.
The expanded agreement builds on 10 years of collaboration between AWS and Salesforce. Salesforce Clouds are currently available through AWS Marketplace in 33 countries, including Japan, allowing eligible customers to apply AWS commitments toward Salesforce products, including Agentforce, while consolidating procurement and billing.
Several elements of the expanded integration are already available, including Amazon Quick’s Salesforce integration, the AWS DevOps Agent in Slack, Agentforce model selection through Amazon Bedrock, expanded Data 360 Zero Copy capabilities, Informatica’s expanded headless functionality, and Slack voice input using AWS Trainium. Additional AWS agents and the Agentforce Voice-Amazon Connect Customer A2A integration are scheduled for autumn 2026.
English