Silvio Savarese, Chief Scientist at Salesforce and Adjunct Professor of Computer Science at Stanford University, has delved into the future development of AI agents, outlining a three-phase evolution: from individual task execution to multi-agent collaboration, and finally to enterprise-wide orchestration.
Three Stages of Enterprise Agents
In the first stage, we witness the rise of the Specialist AI Agent. These agents are focused experts in specific industries, excelling at predefined tasks. They revolutionize departmental workflows with their stability and speed in handling discrete tasks. For instance, in the realm of business, these agents have transformed inventory and account management, enhancing both efficiency and accuracy.
The second stage introduces the Composite Agent – the seamless collaborator. Here, expert agents within a company begin to work in concert towards a common business goal. The composite agent shines in complex business scenarios like customer service, where multiple agents collaborate silently to deliver superior service.
The third and final stage brings about the Integrated Agent – the enterprise orchestrator. This stage sees the emergence of complex agents engaging in cross-organizational interactions, giving rise to novel business models. AI agents act as intermediaries in work and transactions, enabling interactions from business-to-agent (B2A) and even business-to-agent-to-consumer (B2A2C).
Necessary Conditions for Deploying Complex Agent Systems: Trust and Responsibility
In the age of agents, every decision must adhere to the fundamental principles of trust and responsibility. Building trust requires accuracy, self-awareness of agents, and global agreements. Ensuring accountability calls for a clear chain of responsibility and a framework for oversight.
Looking to the Future
In the future, humans and AI will collaborate, leveraging their respective strengths. With the launch of Agentforce, agents have become powerful allies in enhancing work efficiency, enabling teams to accomplish tasks previously thought impossible.
