Chapter 14 AI Orchestration: Shouldn’t we just let AI run everything?

Chapter 14 from Duignan, P. (2026). Surfing AI: 30 New Concepts for Getting Your Head Around AI Shock.

One way of looking at how we make things happen in the world is that it is a matter of coordinating the actions and interactions of different people, objects, information flows, and many other types of systems. As a shorthand, we call this process orchestrating, ‘coordinating’ or ‘integrating’ the interaction of people, objects, and systems.

Such coordination is central to what is done by any company, organization, or initiative. The scale and potency of human action have increased as we have become better at such coordination. Improved IT and communications have ramped up our ability to undertake joint coordinated activity involving many diverse components. This success is reflected in the operations of large governments and transnational companies. It is also seen in major construction projects, transport systems, social institutions such as healthcare, and long and complex logistics and supply chains. However, AI is in the process of creating a step-change in our ability to coordinate activity.

At the moment, some of us think in terms of the potential for AI to be ‘embedded’ in particular objects or systems. Thinking in this way leads us to imagine a single AI system embedded within a single entity of some sort. This way of conceiving embodied AI comes naturally to us. We draw an analogy with how humans are built—as a single intelligence embedded within a single human body.

However, we need to ensure that this analogy does not narrow our thinking about the scope of AI’s potential interaction with people, objects, and systems. While we can think of individual entities acting on the world, we also need to be aware of the ability of networks to coordinate action. There are limits to a human’s ability to function as part of a network. These stem from humans’ physical form and because rapid communication between separate human brains is much slower and more narrow-band than communication within a single human brain.

However, despite our limited networking capability, we have still managed to leverage enormous gains from collaboration and coordination, reflected in our social and cultural achievements. However, AI is now very rapidly pushing wide-scale collaboration further than humans can because of its more advanced networking and information-processing ability. Increasingly, there are instances where an individual AI system is embedded within a robotic or physical system. However, there are also many situations where AI’s role is extending far beyond this. In such cases, a single or tightly federated AI system is be embedded within multiple different objects and systems.

“AI can push wide-scale collaboration further than humans can because of its more advanced networking and information-processing ability”

For instance, such an AI system can potentially collect information from the infrastructure of a building through cameras and multiple sensors. It can then adjust some of the settings of that infrastructure, such as heating and lighting levels. At the same time, it can be operating machines, robotics, mechatronics, and various vehicles working within the building. It can also collect sensor data from the humans working in the facility using their smart watches and communicate with them about what it wants them to do. And, of course, there is no reason for this coordination only to be limited to just one facility. AI coordination can be extended into the outside world, and there is no reason for such coordination not to be extended globally.

This networked approach has obvious efficiencies. These come from the activity of the different people, objects, and systems involved, which are being tightly orchestrated to achieve whatever the overall system seeks. Having new terms to describe what is happening here is essential because it can enable us to think more clearly about how things are developing in the age of AI.

Three terms that can describe this situation are AI orchestration, AI integrability, or AI choreography. A fourth term that can also be used is hydra-AI. This comes from the Hydra, a mythological creature with multiple heads. These terms describe a central AI system or AI federation coordinating the actions of a wide range of other entities and systems.

Given the potential gains from AI orchestration, in what direction are things evolve? There is a tendency for such AI systems to expand, taking more and more people, objects, and systems under their control. The more information such an AI system has fed into it, and the more entities and systems it can control, the better it can coordinate action to achieve its outcomes. This will allow more sophisticated AI orchestration of the people, objects, and systems over which a particular AI or AI federation oversees.

We can call the tendency towards such centralization Eye of God AI. This is the concept of AI orchestration being more effective the larger the span of control put under AI’s management. It is interesting to combine this concept with the automatization imperative discussed earlier. Given that imperative drives towards eliminating human decision-making, we are likely to end up with massive AI-run multiple-component systems almost completely under AI’s autonomous control.

As this happens, those involved in designing such AI-orchestrated systems will likely draw on concepts from human organizational development. Similar principles are likely to apply when coordinating AI system components as when managing people within organizations. These principles include hierarchical structure, span of control, decision-making, outcomes specification, strategic prioritization, performance monitoring, impact evaluation, and alignment between activities and the outcomes being sought. We have already discussed outcomes diagrams, outcomes theory, the outcomes organization and the outcomes society in the chapter on Outcomes-Transparent AI. These tools and concepts will play a role in thinking about and having oversight of large-scale AI-orchestrated systems.

However, there may be some pushback against AI orchestration. One will be the internal operating efficiency of increasingly large centralized AI systems. The bandwidth of communication between components in an AI-integrated system may also be a current constraint. Additionally, AI system designers will presumably prefer to avoid creating monolithic AI systems that can potentially become excessively powerful. Apart from anything else, the development of AI-orchestrated systems will presumably raise concerns regarding the development of monopolies arising from such systems.

There will also be institutional and legal considerations regarding who owns and controls what. For instance, will a building owner want their infrastructure fully integrated with an AI owned by a different company that controls the type of work done in the building? There are also advantages from AI systems competing with and monitoring each other. This is likely to see the emergence of AI watchdogs, AI systems designed to ensure that other AI systems act safely.

Despite any pushback, we will inevitably see the movement towards larger and larger Eye of God AI and AI orchestration because of its advantages. These are the benefits of AI federations emerging that include central control and processing combined with delegated peripheral control. We can see this happening at the moment with AI agent swarms and hierarchical sets of AI systems undertaking delegated tasks.