- “Conversational data is the gold of the agentic era,” Salesforce says
- New upgrades allow Slack to use your unstructured conversations
- Salesforce is also using Anthropic’s MCP server
Few companies have gone so publicly ‘all-in’ on AI agents as much as Salesforce, and the company is now hoping that using agentic AI within its Slack online collaboration platform will help it compete with the likes of Microsoft Teams in new ways.
The company explained that while AI agents rely on a strong data foundation, they thrive off further context, which they have lacked to a certain degree until now.
“Conversational data is the gold of the agentic era, yet it’s been locked away in unstructured messages and chats, largely out of reach for employees, let alone applications,” Salesforce declared.
Slack’s agentic AI is getting a huge boost
Salesforce is adding a new real-time search (RTS) API, a Model Context Protocol (MCP) server and improved developer tools to Slack to help it evolve from offering generic answers to producing more relevant and accurate responses.
The RTS API unlocks unstructured conversational data within Slack via discussions, files and channels, so that AI tools can interact with customer data without having to formalize it into a structured export and store it securely elsewhere.
With the MCP server (an open standard from Anthropic), Salesforce is also to standardize and simplify integrations across multiple enterprise systems for improved interoperability.
For developers, an updated CLI improves the AI app/agent building process.
“By building AI-powered applications directly in Slack, developers and partners help organizations unify their tech stack – making Slack the 2% of your IT budget that maximizes the return on the other 98%,” Salesforce wrote.
With the upgrades, businesses can unlock the unstructured data they’ve had all along for added intelligence, resulting in faster decision-making (37% quicker team insights and 36% faster customer responses) and boosted productivity (3x higher revenue growth per employee).
