There is no doubt that AI agents will continue to be the fastest growing field in enterprise AI.
But as more companies look to deploy agents, they’re also looking for a way to help them make sense of the many steps these autonomous or semi-autonomous, AI-guided bots can take. Conflicts will be avoided.
To counter the potential proliferation of different AI agents deployed by consumers, service providers and enterprises alike are building another type of AI agent: the orchestrator agent.
Enter the orchestrator: These types of agents act as managers of other, more specialized agents, understanding each one’s role and activating each based on the next steps needed to complete the task.
Most orchestrator agents, sometimes called meta-agents, monitor whether an agent succeeded or failed and choose the next agent to achieve the desired result.
Good orchestrator agents exhibit certain characteristics that make these tasks different from other agents, and for enterprises, the elements make them work much better.
integration
Agent ecosystems will ultimately bring workflows together, even if that work involves talking to an agent outside of the existing platform. Orchestrator agents require strong integration with other systems. Otherwise, agents are left as an island that can only communicate with themselves.
Service Now Enterprises need to investigate whether the orchestration agents they are building or buying offer integration points for other systems, said Dorit Zilbershot, vice president of AI and innovation.
“Effective orchestration agents support integration with multiple enterprise systems, enabling them to pull data and execute actions across organizations,” Zllbershot said. “This holistic approach provides the orchestration agent with a deep understanding of the business context, allowing intelligent, contextual work management and prioritization.”
For now, AI agents exist on islands within themselves. However, service providers such as ServiceNow and Slack have begun to integrate with other agents. Slack announced that it offers integrations for agents from Salesforce, Workday, Asana and Coher. Full-stack AI company Writer connects its agents to Amazon and Macy’s APIs to sell products directly to customers.
On Don Schurman, CTO Pegaechoed the sentiment, saying that an ideal orchestration agent is “API-centric so it can work across all agents as well as human-centric channels to pull humans in when needed.”
Knowledge of enterprise processes
Like all agents, orchestrator agents need to know how the business works.
Orchestrator agents need a more holistic view of the best next step in moving the process forward. Zilbershot said a good orchestration agent is “capable of quickly analyzing context to determine both the best-suited AI agent and the best sequence of AI agent assignments to improve workflow and reduce latency.” Should be.”
It’s not just about having insight into a company’s data—although that’s another essential component to an agent ecosystem—it’s also about understanding the processes that enterprises take to run their business.
The writer CEO May Habib told VentureBeat in an earlier interview that enterprises that want an effective agent system should provide workflow for the orchestrator agent, not the other way around.
“If you don’t get the nodes in the workflow right, the automated workflow is just moving crap from one system to another,” Habib said. “Over time, we built an application that automatically knew, with AI, which tools to access based on the workflow.”
Reasoning skills.
By its very nature, orchestrator agents make more rational decisions than other AI agents. As AI agents are tasked with more complex tasks, so too are orchestrator agents to help manage them.
Larger language models reduce agent creation, and models with more reasoning capabilities can run different scenarios before triggering the next agent. Orchestrator agents must have strong reasoning skills to ensure that the workflow does not break down.
Smooth communication between agents and human employees
ServiceNow’s Zilbershot points out that orchestration agents are primarily responsible for interactions between humans and agents. Enterprises deploying AI agents will benefit from orchestrator agents with user-friendly interfaces and feedback networks, he said. Therefore, agents continue to improve based on how employees interact and use them.
“By acting as the connective tissue between specialized AI agents and human operators, orchestration agents make it easier to not only streamline operations but also increase the overall effectiveness of an organization’s agent AI systems,” he said.
While AI agents are designed to automate workflows, experts said it’s still important that the handoff between human employees and AI agents is smooth. Agent orchestration allows humans to see where agents are in the workflow and lets the agent find its way to completing tasks.
“An ideal orchestration agent allows for visual appreciation of processes, has rich auditing capabilities, and can leverage its AI to make recommendations and guide the best actions. At the same time, it can provide data A virtualization layer is needed to ensure that the orchestration logic is decoupled from the complexity of the back-end data stores,” said Pega’s Schurman.
Orchestrator agents already ship to many agent frameworks. It may even be the difference for many agent libraries in the future. As enterprises continue to experiment more with agents, orchestrator agents may improve.
Credit : venturebeat.com
