The integration of artificial intelligence (AI) agents within enterprises is evolving from simple functionality to a critical focus on how these agents interact with one another. Businesses are now prioritizing the orchestration of multi-agent systems, aiming to enhance communication and collaboration among various AI solutions. This shift addresses the potential for misunderstandings, which can hinder performance and lead to security risks, according to Tim Sanders, Chief Innovation Officer at G2, in an interview with VentureBeat.
Enhancing Coordination Among AI Agents
Traditionally, AI orchestration centered on data management, but the focus is increasingly shifting towards actionable coordination. Sanders described the emergence of “conductor-like solutions” that integrate AI agents, robotic process automation (RPA), and data repositories. He likened this evolution to the development of answer engine optimization, which has transitioned from monitoring to creating customized content and code. “Orchestration platforms coordinate a variety of different agentic solutions to increase the consistency of outcomes,” he explained.
Notable early providers in this space include Salesforce MuleSoft, UiPath Maestro, and IBM Watsonx Orchestrate. These platforms serve as “phase one” observability dashboards that allow IT leaders to monitor agent activities across an organization, enhancing overall visibility and control.
Addressing Risks Through Advanced Management Tools
While coordination adds significant value, these orchestration platforms are expected to evolve into comprehensive technical risk management tools. They will likely incorporate features such as agent assessments, policy recommendations, and proactive scoring. For instance, organizations may gain insights into how reliable agents are when utilizing enterprise tools or the frequency of erroneous outputs, known as “hallucinations.”
As organizations increasingly adopt AI solutions, many leaders express skepticism regarding the reliability claims made by vendors. Sanders noted that IT decision-makers often do not fully trust vendor assertions about their agents’ performance. To address this concern, third-party tools are emerging to automate tedious processes, such as managing guardrails and escalation tickets.
For example, in the banking sector, the loan approval process can involve up to 17 steps. AI agents often interrupt human workflows by requesting approvals when encountering established guardrails. Third-party orchestration platforms can streamline these interactions, potentially eliminating the need for constant human oversight and fostering significant productivity gains.
“Where it goes from there is remote management of the entire agentic process for organizations,” Sanders stated.
The distinction between “human-in-the-loop” and “human-on-the-loop” roles is also critical to this evolution. Moving forward, human evaluators are expected to transition into designers, actively shaping agents to automate workflows. This shift is facilitated by the ongoing innovation in no-code solutions that enable users to create agents using simple natural language.
“This will democratize agentic AI, and the super skill will be the ability to express a goal, provide context, and envision pitfalls, very similar to a good people manager today,” Sanders added.
Strategies for Effective Implementation
According to Sanders, enterprises should implement “expeditious programs” to integrate AI agents into workflows, particularly in areas characterized by repetitive tasks that create bottlenecks. Initially, human oversight will remain crucial to ensure quality and effective change management. Serving as evaluators will enhance understanding of these systems, ultimately empowering organizations to engage more effectively with agentic workflows.
IT leaders are encouraged to conduct a thorough inventory of their automation capabilities, including rules-based automation, RPA, and emerging agentic automation technologies. A comprehensive understanding of the organization’s automation stack will enable optimal utilization of orchestration platforms as they develop.
“If they don’t, there could actually be dis-synergies across organizations where old school technology and cutting-edge technology clash at the point of delivery, oftentimes customer-facing,” Sanders warned. “You can’t orchestrate what you can’t see clearly.”
The ongoing evolution of AI agents promises to transform enterprise workflows, enhancing both efficiency and effectiveness. As organizations embrace these technologies, the need for clear communication among agents will be paramount to their success.
