The Pulse
AWS Brings Bedrock Models to Salesforce as Agents Move Into Slack
AWS and Salesforce announced new integrations connecting Amazon Bedrock models, Salesforce Agentforce, Amazon Quick and Slack. The September 15 agreement also adds zero-copy data access, AWS agents in Slack and planned voice coordination be
AI.info Team ·
AWS and Salesforce announced on September 15 that they are connecting CRM data, AI agents and foundation-model options across Amazon Quick, Slack, Salesforce Agentforce and Amazon Bedrock.
The expanded collaboration is designed to let employees work with Salesforce data from Amazon Quick, bring AWS agents into Slack and give Agentforce customers access to models available through Bedrock. The companies also announced planned voice coordination between Agentforce Voice and Amazon Connect Customer, along with new zero-copy connections to AWS data services.
Salesforce published a matching account of the announcement on its newsroom, while AWS detailed the product availability and customer examples in its September 15 release.
Salesforce data moves into Amazon Quick
The first integration brings Salesforce business context and actions directly into Amazon Quick, AWS’s enterprise assistant. The companies say customers can use the connection to retrieve pipeline health, account summaries, service cases and operational analytics without building a custom integration.
The connection uses Salesforce’s headless architecture and the Model Context Protocol, allowing Quick to work with Salesforce records and skills. AWS gives the example of an account team asking Quick for the latest deal status, open service cases and recent customer activity, then receiving a meeting brief based on live CRM data.
Nick Johnston, Salesforce’s executive vice president for global partnerships, said the arrangement extends Salesforce context beyond the CRM interface. “Trusted data now reaches into the surfaces teams rely on daily, and our collaboration with AWS extends that reach even further, all with governance built in every step of the way,” Johnston said.
AWS agents enter Slack channels
AWS is also placing its frontier agents inside Slack, starting with AWS DevOps Agent. AWS says teams can investigate incidents and review findings in the Slack channel where the discussion is already taking place, rather than moving into a separate AWS console.
The agent can use the conversation’s context when responding, which matters for teams handling incidents across multiple messages and participants. AWS Security Agent, AWS FinOps Agent and AWS Partner Central agents are scheduled to follow in fall 2026, according to the release.
GoodRx is testing the DevOps integration. Muddassar Shaikh, the company’s senior vice president of engineering, said, “A lot of our engineering work and conversations already happen in Slack. Bringing the AWS DevOps Agent into those conversations allows my teams to spot and resolve issues in their development environments faster, without breaking their flow.”
Agentforce gets Bedrock model choice
Salesforce customers using Agentforce can now access models from Anthropic and NVIDIA through Amazon Bedrock within Salesforce agents and agentic workflows. OpenAI models, which AWS says are already generally available on Bedrock, are coming to Agentforce later.
The arrangement gives enterprises a way to select different models for different workloads while keeping the agent experience inside Salesforce. The announcement does not specify which individual Anthropic or NVIDIA models are included, or provide pricing for the new access.
AWS says Bedrock applies zero data retention, meaning customer prompts and outputs are not stored or used to train models by providers. Salesforce says its Trust Layer adds dynamic grounding, toxicity detection and audit trails; the release also lists support for HIPAA, PCI, SOC 2 and ISO 42001 requirements for regulated customers.
Joshua Stern, director of GTM systems at Engine, said access to Bedrock’s model selection gives the company “the flexibility to choose the right model for each use case.”
Zero-copy access expands beyond Salesforce
The companies are extending Salesforce Data 360 zero-copy capabilities to more AWS data services. Zero-copy connections allow systems to query data where it resides instead of first moving it into another repository.
AWS and Salesforce present that arrangement as a response to a common enterprise constraint: companies want AI systems to work across cloud environments but do not want to undertake a full data migration. The announcement provides no list of every newly supported AWS data service, but says the Data 360 expansion is available now.
The same release says Informatica’s headless capabilities are also expanding. Salesforce Clouds are available through AWS Marketplace in 33 countries, allowing customers to apply eligible AWS spending commitments to Salesforce products, including Agentforce, and consolidate procurement and billing.
Voice coordination is scheduled for fall
The most ambitious part of the announcement is not yet generally available. AWS and Salesforce plan to connect Agentforce Voice with Amazon Connect Customer through the Agent2Agent, or A2A, protocol.
The proposed system will allow the two voice platforms to exchange real-time, bidirectional audio over WebSockets. The companies describe it as a way for agents operating on each platform to coordinate during customer calls, while giving customers more choice over the systems they use.
Salesforce AI Research and AWS are also working on voice dictation for Slack using AWS Trainium chips. The companies say the speech models are being optimized for low-latency transcription, with the feature available now.
For customers, the announcements divide into products available immediately and products still on the delivery schedule: Amazon Quick with Salesforce’s headless architecture, Agentforce access to Bedrock models, Data 360 zero-copy expansion and Slack voice dictation are available now; additional AWS agents for Slack and the Agentforce Voice integration are planned for fall 2026.