AWS and Salesforce Expand AI Agent Integration Across Quick, Slack, and Bedrock

The expanded AWS–Salesforce collaboration targets the daily-workflow layer where enterprise AI actually gets used. The centerpiece is making Salesforce data, context, and skills natively available through Amazon Quick, built on the Model Context Protocol (MCP), so an account team can surface pipeline health, account summaries, and open service cases inside one interface without app-switching. AWS also brought its frontier agents into Slack and expanded Agentforce model choice via Amazon Bedrock.

AWS's Matt Wood framed the thesis directly: 'The enterprises getting the most from AI are building on data they already own, in the tools they already trust.' The MCP foundation is notable — it's the emerging open standard for connecting agents to data sources, and both companies leaning on it signals convergence around interoperable agent plumbing rather than proprietary connectors.

The deal mechanics give Salesforce customers model choice through Bedrock (Claude, GPT-6 Astra, Nova) while keeping data governance in place, and give AWS deeper hooks into Salesforce's enterprise install base. It lands the same week Salesforce also unveiled its NVIDIA-based Koa model at Dreamforce — evidence Salesforce is deliberately multi-sourcing its AI stack across AWS, NVIDIA, and its own models.

The competitive backdrop is a three-way scramble (AWS, Microsoft, Google) to own the enterprise-agent surface, where the winner is whoever most seamlessly connects agents to systems of record. The caveat: integration announcements are easy; sustained adoption depends on whether these agents reliably act on live CRM data without governance failures. Watch for customer deployment case studies rather than launch demos.