Building on its ongoing focus on artificial intelligence agents, Salesforce Inc. announced today the launch of AgentExchange, a trusted marketplace that allows enterprise customers to expand the capabilities of Agentforce AI agents.

Using Agentforce, companies can deploy highly customizable generative AI agents that operate autonomously, augmenting employee work. The platform enables developers to create AI “workers” that use business rules and prebuilt automation to perform tasks independently.

With today’s announcement, AgentExchange will build on the success of Salesforce’s AppExchange, the company’s existing app marketplace that has grown to more than 13 million app installations. Instead of apps, it will house skills and capabilities for AI agents and act as a marketplace for what the company calls “digital labor.”

AgentExchange launches with more than 200 initial partners and will contain hundreds of ready-made actions, topics and templates vetted by the company’s staff through rigorous security and customer review. The solutions available will help organizations quickly deploy AI agents to improve productivity and efficiency.

Partners adding their expertise and components to the exchange include such big names as cloud content management company Box Inc. and electronic signature company Docusign Inc. With Box, AI agents can extract insights from unstructured data and interact with data in Box using natural language. Agents using Docusign can generate documents and track signing authentication as well as signature status.

“AgentExchange unlocks new technology and markets for Docusign, expanding our reach and impact,” said Larry Jin, vice president of product management at Docusign. “We’ve worked closely with Salesforce to develop agent actions to seamlessly integrate them into our customers’ workflows.”

Salesforce recently inked a $2.5 billion deal with Google Cloud to allow its users to access Google Gemini and other tools across its marketplace. Google’s Gemini models will allow Agentforce AI agents to work with images, audio and video, handle more complex tasks and act using real-time insights as well as access Google Search.

When using the built-in Agentforce agent builder, users will be able to use the exchange to search for specific actions tailored to their industry from integration partners, ranging from flows and prompts.

Topics allow developers and business users to refine agent behavior further by grouping agent actions and instructions around a specific task or job. This lets them define business logic in natural language so the agent will deliver more consistent results – essentially a type of guardrail ensuring the agent stays on track.

Partners can also provide pre-written reusable prompt templates that make interacting with agents easier. This is especially useful when there is a consistent way that agents gather information and a particular goal is achieved but a specific input is changed — for example, a username, address or other information.

Finally, agent templates allow customers to customize agents from a patterned concept that makes sense to them. They work by combining multiple topics and multiple agent skills, complete with metadata and global instructions for a particular industry, allowing users to deploy and tailor an agent quickly without needing to start from scratch.

Image: SiliconANGLE/Microsoft Designer