More than Human Resources

You are on chapter // 04 Intelligence

Human Intelligence (HI) spans a wide range of cognitive abilities, including experiential learning, language, abstract reasoning, and complex problem-solving. Artificial Intelligence (AI) can be defined as computer systems designed to perform tasks that typically require HI.

One way AI systems can be described is by the degree of autonomy they possess, from those requiring constant human supervision and intervention even in a closed environment to fully autonomous systems operating independently in open world environments. Artificial Narrow Intelligence (ANI) refers to AI systems designed to perform specific tasks or a limited set of tasks. These systems excel within their narrow domains but lack the ability to generalize.

“The best work still emerges from Human-AI teams, possessing distinctive features compared to Human or AI counterparts, balancing biases and augmenting cognitive abilities.”

Artificial General Intelligence (AGI) is a hypothetical development stage in which AI can understand, learn and apply intelligence across a broad range of complex tasks and domains, mimicking human cognitive capabilities. Artificial Superintelligence (ASI) is a theoretical concept describing AI that surpasses HI in all aspects, including those qualities traditionally associated with human cognition.

A common industry framing outlining the progression from ANI to ASI in economic terms is the five-step framework: (1) Chatbots are conversational interfaces for workers to use; (2) Reasoners are systems that achieve human-level problem solving in predefined tasks. They replace some human workers in an organization and augments others; (3) Agents are capable of autonomous action within human-defined boundaries. Agents can replace most human roles in an organization, carrying out the majority of operations without human intervention; (4) Innovators are capable of invention and innovation, creating new products or services, leading to the final step; and (5) Organizations, systems that could operate a whole organization. This represents the pinnacle of the economic AI model, where no roles remain exclusively human, even managerial ones.

Previous attempts to combine humans (H) and AI have focused on specialization, where each performs different tasks or subtasks. AI automates routine work and augments human expertise, freeing people for higher-order reasoning. AI performs well at pre-defined tasks while humans perform better at open-ended problems such as innovation. The best work still emerges from Human-AI teams, possessing distinctive features compared to Human or AI counterparts, balancing biases and augmenting cognitive abilities.

Dual-process theory distinguishes two interacting modes of thought: System 1 (fast, intuitive, heuristic) and System 2 (slow, deliberate, computational). Humans typically excel at System-1 thinking—intuition, empathy, sociocultural awareness—while AI traditionally scaled System-2 work—large-scale analysis and logical reasoning. In human-in-the-loop (HITL) workflows, continuous human feedback steers the system, aligning outcomes with human values while leveraging machine scale.

“Ensembling fosters an environment where H and AI are equal contributors, unlocking the full potential of AI to the benefit of the organization.”

However, today’s AI spans the two modes of thinking, making specialization a major constraint. The concept of ensembling introduces a third path, orthogonal to both automation and augmentation. By aggregating the strengths of H and AI without confining them to specialized roles, ensembling fosters an environment where H and AI are equal contributors, unlocking the full potential of AI to the benefit of the organization. This shift is visible in emerging artificial colleagues and even managers. These Non-Human Knowledge Workers (NHKW’s) are synthetic agents characterized by the convergence of four attributes distinguishing them from the common AI agent—information processing, knowledge work, task-level employment and comprehensive organizational integration.

AI adoption in business can consequently be divided into two primary approaches: augmentation and automation. Late-stage augmentation predicts one-person companies, where a single individual operates a company with AI assistance. In contrast, late-stage automation predicts zero-person companies, where AI systems are capable of operating an entire enterprise without human intervention. Beyond both lies ensembling—human and non-human knowledge workers as peers—pointing toward systems that are completely integrated.