The orchestration of several agents collaboratively executing their functions and coordinating their activities, such as creating communication and reacting in real time in a smart way, can often be more effective than using just a single agent; however, the overall orchestration of these agents can become increasingly complex compared to simply having multiple agents perform their functions and coordinate their activities.
For example, consider a supply chain where one agent forecasts demand, another agent manages inventory, and yet another agent manages logistics; all three agents perform their assigned roles independently but are connected to each other through a common objective to meet demand.
Managing the complexity of orchestration of multiple agents can make accountability and responsibility difficult, due to the dispersion of decision-making amongst agents. This is why there are many problems experienced with the governance structures that are officially assigned to support the performance of multiple agents.
Check out: Human-Centered AI Integration
In a pilot project experience, I had two systems that were optimized for separate goals (cost versus speed), which both reached high levels of performance individually, but combined, the two systems placed a great deal of stress on the entire system due to the competing objectives.
Ultimately, the ability of multi-agent systems to create significant scalability and efficiency advantages over single agents will be tremendous.
At JAMS Advisory we raise awareness about the importance of understanding the goal of creating orchestrated and coordinated systems of agents.
Ultimately, many intelligent agents operating within a multi-agent system to accomplish a common objective will not be able to create value unless those agents work together in a manner consistent with the overall business.
